1
00:00:00,000 --> 00:00:09,960
Welcome to Artificially Intelligent Marketing, a weekly podcast where we stay on top of the

2
00:00:09,960 --> 00:00:15,700
latest trends, tips and tools in the world of marketing AI, helping you get the best

3
00:00:15,700 --> 00:00:18,540
results from your marketing efforts.

4
00:00:18,540 --> 00:00:22,880
Now let's join our hosts, Paul Avery and Martin Broadhurst.

5
00:00:22,880 --> 00:00:27,560
Hello everyone, welcome to episode 20 of Artificially Intelligent Marketing.

6
00:00:27,560 --> 00:00:28,560
We're glad you're here with us.

7
00:00:28,560 --> 00:00:32,240
I'm also glad to be joined by my good friend, Martin Broadhurst.

8
00:00:32,240 --> 00:00:34,280
Martin, how are you?

9
00:00:34,280 --> 00:00:35,800
Fantastic, thank you.

10
00:00:35,800 --> 00:00:43,120
Glad to be back home after the adventures over to Cleveland for MAKON 2023.

11
00:00:43,120 --> 00:00:46,960
Listeners to last week's episode will have heard my report from there.

12
00:00:46,960 --> 00:00:53,520
It was a great conference, met some fantastic people and got some insights and thought provoking

13
00:00:53,520 --> 00:00:59,080
discussions around what AI means for organizations.

14
00:00:59,080 --> 00:01:02,000
So a little bit less on the technology front.

15
00:01:02,000 --> 00:01:06,760
I guess we're a bit inside baseball when it comes to the technology at this juncture,

16
00:01:06,760 --> 00:01:13,600
but you really had some key insights and thoughts on organizational change and how do organizations

17
00:01:13,600 --> 00:01:19,520
actually start to bring some of this generative AI, especially power into their businesses.

18
00:01:19,520 --> 00:01:20,840
Yeah, very much so.

19
00:01:20,840 --> 00:01:27,360
I think one of the most interesting discussions from the whole conference was a fireside chat

20
00:01:27,360 --> 00:01:32,040
with one of the senior marketing team from VMware.

21
00:01:32,040 --> 00:01:40,040
Now this is an organization with 35,000 employees globally, 650 marketing professionals in the

22
00:01:40,040 --> 00:01:45,640
team and they have created earlier this year, I think it was February this year, they put

23
00:01:45,640 --> 00:01:48,400
it together, an AI council.

24
00:01:48,400 --> 00:01:54,600
Now this is made up of 30 employees from the marketing team and they meet regularly to

25
00:01:54,600 --> 00:02:00,480
discuss how the organization is going to use AI across the marketing piece.

26
00:02:00,480 --> 00:02:07,280
That's everything from predictive analytics, personalization through to generative AI.

27
00:02:07,280 --> 00:02:11,840
And they're considering all sorts of things, including the ethics, but which tools should

28
00:02:11,840 --> 00:02:12,840
they use?

29
00:02:12,840 --> 00:02:19,240
And getting into the weeds to say, should we be using this generative AI copywriting

30
00:02:19,240 --> 00:02:20,760
tool or this one?

31
00:02:20,760 --> 00:02:23,320
They're really getting into that.

32
00:02:23,320 --> 00:02:27,760
I think that was something that seemed to be a bit of a trend.

33
00:02:27,760 --> 00:02:32,720
The companies at the leading edge of this are dedicating resource to it, whether it's

34
00:02:32,720 --> 00:02:39,400
someone's full-time job, which seemed to be a rarity or somebody was built for it.

35
00:02:39,400 --> 00:02:43,360
Somebody had the accountability for AI within the organization.

36
00:02:43,360 --> 00:02:48,120
That definitely seemed to be more of a trend for the forward-looking organizations.

37
00:02:48,120 --> 00:02:49,120
Very interesting stuff.

38
00:02:49,120 --> 00:02:57,040
I'm glad you touched upon how to choose what tools to do what with because I can very effectively

39
00:02:57,040 --> 00:03:00,480
segue into letting the listeners know what we're going to cover today because we have

40
00:03:00,480 --> 00:03:05,240
a slightly special episode this week because Martin was lucky enough to catch up with Brennan

41
00:03:05,240 --> 00:03:09,560
Woodruff, Chief Business Officer over at Go Charlie.

42
00:03:09,560 --> 00:03:15,320
So we've got a fantastic interview with Brennan at the end of this podcast, the sort of the

43
00:03:15,320 --> 00:03:19,040
sections that we normally do on the news, which we'll do in a moment, and then into

44
00:03:19,040 --> 00:03:20,040
that interview.

45
00:03:20,040 --> 00:03:23,160
So do look forward to that later on.

46
00:03:23,160 --> 00:03:28,200
In terms of the news stories, because we want to respect your time, lovely listeners, we're

47
00:03:28,200 --> 00:03:31,400
going to race through a couple of stories that we thought were the most important this

48
00:03:31,400 --> 00:03:35,480
week and then get straight into that interview.

49
00:03:35,480 --> 00:03:37,360
So the first story Martin is with you.

50
00:03:37,360 --> 00:03:38,720
It's about OpenAI.

51
00:03:38,720 --> 00:03:39,720
Tell us about it.

52
00:03:39,720 --> 00:03:40,720
Yeah.

53
00:03:40,720 --> 00:03:45,320
So OpenAI announced this week that they're providing a way to block its bot from accessing

54
00:03:45,320 --> 00:03:50,080
your content, which is important because as marketers, we pour a lot of time, effort,

55
00:03:50,080 --> 00:03:55,880
energy into creating all of that website content, whether it's blog posts or product pages,

56
00:03:55,880 --> 00:04:01,600
it's all strategically designed to attract and engage with different customers and staff.

57
00:04:01,600 --> 00:04:09,480
We don't necessarily want all of that IP getting slooped up by the AI bots without our permission.

58
00:04:09,480 --> 00:04:14,520
And that's certainly the case with the major publishers such as newsrooms.

59
00:04:14,520 --> 00:04:20,600
So OpenAI has given us the tools now that we need to protect our sites so we can control

60
00:04:20,600 --> 00:04:31,480
which pages the GPT bot can access and analyze and then incorporate into the training systems

61
00:04:31,480 --> 00:04:34,440
of their large language models.

62
00:04:34,440 --> 00:04:41,520
So what they've done is they've provided specific user agent string and some IP address ranges

63
00:04:41,520 --> 00:04:48,520
for their crawler so that we can now easily block the bot at a technical level by updating

64
00:04:48,520 --> 00:04:52,120
our robots TXT file and our firewall rules.

65
00:04:52,120 --> 00:04:57,680
So very much like we can block search engines from indexing certain pages, if we've got

66
00:04:57,680 --> 00:05:01,400
maybe we've got thank you pages or pieces of content on our website that we don't want

67
00:05:01,400 --> 00:05:06,920
indexed in search engines, it's the exact same theory as that.

68
00:05:06,920 --> 00:05:13,560
So that's going to help us safeguard any proprietary documentation that might be accidentally spied

69
00:05:13,560 --> 00:05:20,160
on by the crawler so we can, whether it's PDFs, we can de-index those from the crawler

70
00:05:20,160 --> 00:05:21,160
and all sorts.

71
00:05:21,160 --> 00:05:25,760
It's just going to help us keep proprietary information proprietary.

72
00:05:25,760 --> 00:05:31,640
So another interesting thing to consider here is that there are loads of websites where

73
00:05:31,640 --> 00:05:38,080
there might be personally identifiable information and user generated content, stuff that people

74
00:05:38,080 --> 00:05:42,320
might be using our website, but they don't want, they're not saying that everyone in

75
00:05:42,320 --> 00:05:45,560
the world can access, it's just maybe website users.

76
00:05:45,560 --> 00:05:52,440
So now we can delist those pages to make sure that GPT-5 or whatever it's going to be in

77
00:05:52,440 --> 00:05:57,720
the future isn't trained on all of this user generated content as well if we don't want

78
00:05:57,720 --> 00:05:58,720
it to.

79
00:05:58,720 --> 00:06:03,720
So yeah, at the end of the day, this is a great move for product marketers and just

80
00:06:03,720 --> 00:06:11,920
for web masters and digital marketers to have more control over what is indexed, analyzed

81
00:06:11,920 --> 00:06:16,520
and used in coming large language model training.

82
00:06:16,520 --> 00:06:22,560
And yeah, hat tip to OpenAI for the transparency and acknowledging the concerns of publishers.

83
00:06:22,560 --> 00:06:28,640
I think they have had their feet held to the fire a little bit with some of the lawsuits

84
00:06:28,640 --> 00:06:33,080
that are coming up from the likes of New York Times and Sarah Silverman.

85
00:06:33,080 --> 00:06:39,000
So yeah, this is good for digital marketers to have more control over their content.

86
00:06:39,000 --> 00:06:40,000
Yeah.

87
00:06:40,000 --> 00:06:41,560
I think it's an interesting one.

88
00:06:41,560 --> 00:06:44,640
On the one hand, the horse has already bolted on this.

89
00:06:44,640 --> 00:06:50,320
So all your old content, would I go through and try and de-index my old content from OpenAI

90
00:06:50,320 --> 00:06:51,320
at this point?

91
00:06:51,320 --> 00:06:52,920
No, we already scraped it.

92
00:06:52,920 --> 00:06:56,480
Or at least, you know, especially if I was a publisher and had a lot of high quality

93
00:06:56,480 --> 00:06:59,520
content on my website like in New York Times.

94
00:06:59,520 --> 00:07:03,860
But it's good to know that we have control over restricting that information from going

95
00:07:03,860 --> 00:07:05,640
into the model from here forward.

96
00:07:05,640 --> 00:07:12,000
So in terms of building a strategy for this, it's probably a new content production, build

97
00:07:12,000 --> 00:07:15,880
this into your workflow if you don't want it to be scraped by bots.

98
00:07:15,880 --> 00:07:22,320
In terms of OpenAI bots for training large language models, in terms of a strategy, it's

99
00:07:22,320 --> 00:07:23,320
not clear, is it?

100
00:07:23,320 --> 00:07:28,440
Yeah, I don't think whether or not in some cases we might want our content indexed, especially

101
00:07:28,440 --> 00:07:32,720
if it's quite brand heavy in terms of it's talking about what we can do specifically

102
00:07:32,720 --> 00:07:34,440
as a business and how we do it.

103
00:07:34,440 --> 00:07:39,760
Because actually, that might be the type of information and recommendations we want bots

104
00:07:39,760 --> 00:07:43,560
to be giving if it's branded versus say general knowledge info.

105
00:07:43,560 --> 00:07:52,040
Well, I saw a screenshot in a Facebook group that I'm part of and it was someone's sketch.

106
00:07:52,040 --> 00:07:58,000
There was a chat GPT conversation and within it, it mentioned a particular brand and it

107
00:07:58,000 --> 00:08:00,600
had a link to their website.

108
00:08:00,600 --> 00:08:03,740
And the comment was, how has this company done that?

109
00:08:03,740 --> 00:08:10,600
And the person sharing it was asking it almost as if it was like AI optimization, like in

110
00:08:10,600 --> 00:08:13,720
search engine optimization, it'd been optimized for this.

111
00:08:13,720 --> 00:08:18,200
Now, clearly it hadn't, that's not the way chat GPT works.

112
00:08:18,200 --> 00:08:22,320
It wasn't a conversation with plugins or anything, it was just a raw chat.

113
00:08:22,320 --> 00:08:27,480
So within the model, clearly the topic that they were talking about had this company that

114
00:08:27,480 --> 00:08:36,360
was referenced within its knowledge base and it shared the website address of this company.

115
00:08:36,360 --> 00:08:41,440
I think that's going to be, if you can get lots of content that says you are aligned

116
00:08:41,440 --> 00:08:48,600
to a particular topic, that's going to appear more in the models going forward.

117
00:08:48,600 --> 00:08:52,080
Now that's assuming that the architecture of these models stays the same as it is today,

118
00:08:52,080 --> 00:08:55,440
which I highly doubt going forward.

119
00:08:55,440 --> 00:09:01,280
On GPT-5, I imagine things are going to look slightly different, but yeah, I think people

120
00:09:01,280 --> 00:09:06,440
already have this perception about we need to optimize for AI models, even if that's

121
00:09:06,440 --> 00:09:09,320
not necessarily the case of how it works in reality.

122
00:09:09,320 --> 00:09:11,440
Yeah, it's funny you should say that.

123
00:09:11,440 --> 00:09:19,120
I remember pre-GPT-4, I asked chat GPT to give me five life science marketing agencies

124
00:09:19,120 --> 00:09:23,700
and I was pleased to see that Biostrata was in the list because certainly I can imagine

125
00:09:23,700 --> 00:09:28,320
that trying to make sure you end up in that list is not necessarily something that's going

126
00:09:28,320 --> 00:09:32,180
to be easy to control, but it is the type of question that somebody might reasonably

127
00:09:32,180 --> 00:09:36,960
go and ask a chat bot over a search engine in the future.

128
00:09:36,960 --> 00:09:41,680
Expect a whole scramble to try and think about how we can engineer the content on our sites

129
00:09:41,680 --> 00:09:46,720
to try and ensure that the models learn that we're a relevant brand in whatever areas that

130
00:09:46,720 --> 00:09:47,720
we're playing in.

131
00:09:47,720 --> 00:09:50,960
Right, let's look at our next story then.

132
00:09:50,960 --> 00:09:58,160
There's been a couple of stories this week about generative image AIs and protecting

133
00:09:58,160 --> 00:10:02,800
artists' work and how do we tell the difference between real images and fake images.

134
00:10:02,800 --> 00:10:05,540
There's been some interesting stories here.

135
00:10:05,540 --> 00:10:11,460
We heard about some initiatives, one of which was called C2PA, which stands for Content

136
00:10:11,460 --> 00:10:13,360
Authenticity and Attribution.

137
00:10:13,360 --> 00:10:19,980
It's kind of like a Twitter blue tick verification system, but in this case for images.

138
00:10:19,980 --> 00:10:26,360
The tool uses cryptography to tag content that's been generated by AI so that as humans

139
00:10:26,360 --> 00:10:31,920
or users of images, we can identify if an image is AI generated or not.

140
00:10:31,920 --> 00:10:38,360
Adobe has even incorporated C2PA, which kind of sounds like a character from Star Wars

141
00:10:38,360 --> 00:10:44,480
to me, into Photoshop and Firefly, making it accessible to users.

142
00:10:44,480 --> 00:10:49,740
It is a system that you opt into though, so it's not as if every image that you generate

143
00:10:49,740 --> 00:10:53,880
artificially is going to get tagged, at least not as it stands.

144
00:10:53,880 --> 00:10:58,400
Another initiative we saw on this kind of topic this week is Photoguard, which has

145
00:10:58,400 --> 00:11:01,680
been developed by the bods over MIT.

146
00:11:01,680 --> 00:11:08,880
This tool protects your original images from being repurposed by AI and used as part of

147
00:11:08,880 --> 00:11:11,440
image AI training runs.

148
00:11:11,440 --> 00:11:15,480
What they've done is they've found a way to tweak an image's pixels so that AI models

149
00:11:15,480 --> 00:11:20,000
like stable diffusion basically can't recognize them.

150
00:11:20,000 --> 00:11:26,000
It's all code driven at the moment, but there are plans we hear to have Photoguard's API

151
00:11:26,000 --> 00:11:33,560
baked into different platforms so that you can easily overlay your images with this watermark

152
00:11:33,560 --> 00:11:39,080
or whatever that stops these tools from being able to index and use your images as part

153
00:11:39,080 --> 00:11:40,300
of their training runs.

154
00:11:40,300 --> 00:11:44,820
In essence, because even though humans can see what the image looks like, it's kind

155
00:11:44,820 --> 00:11:48,680
of garbled for the AI tools.

156
00:11:48,680 --> 00:11:50,720
Why does this matter for marketers?

157
00:11:50,720 --> 00:11:56,280
Well, first and foremost, these are pretty interesting tools, but they're still a little

158
00:11:56,280 --> 00:11:58,680
bit unproven at the moment.

159
00:11:58,680 --> 00:12:00,960
How easy will it be to sidestep them?

160
00:12:00,960 --> 00:12:03,880
Nobody knows currently as far as I can tell.

161
00:12:03,880 --> 00:12:09,160
I think also you could argue that some of the larger AI models, mid-journey, stable

162
00:12:09,160 --> 00:12:13,040
diffusion have already consumed a vast amount of images, so maybe that damage is going to

163
00:12:13,040 --> 00:12:15,040
be a bit hard to undo.

164
00:12:15,040 --> 00:12:19,120
It's a little bit like the first story from Martin in terms of probably images you've

165
00:12:19,120 --> 00:12:20,360
already put online.

166
00:12:20,360 --> 00:12:25,640
That ship has sailed, but your new images perhaps look at some of these tools.

167
00:12:25,640 --> 00:12:29,520
I think as marketers, we really need to be aware of all these different developments

168
00:12:29,520 --> 00:12:35,040
so that we can be informed about the risks and the potential solutions when we as brands

169
00:12:35,040 --> 00:12:41,440
are creating content, but also just the overall complexities of how these tools work and whether

170
00:12:41,440 --> 00:12:46,400
when we're using them, we're going to be infringing on the copyrights of other creatives and anything

171
00:12:46,400 --> 00:12:50,240
that can be built into this whole system to give it more credibility, I think is going

172
00:12:50,240 --> 00:12:56,000
to give brands more confidence in using these tools to generate images as part of their

173
00:12:56,000 --> 00:12:58,160
creative workflows.

174
00:12:58,160 --> 00:13:07,880
There's definitely a theme here, isn't there, of how do you stop AI consuming your content?

175
00:13:07,880 --> 00:13:14,760
This is very much front and center of people's thinking at the moment.

176
00:13:14,760 --> 00:13:18,760
I was intrigued by the Photoguard solution.

177
00:13:18,760 --> 00:13:27,000
I'd like to get into the technical details of that to see how it's basically baffling

178
00:13:27,000 --> 00:13:32,480
the diffusion models in the pixel level obfuscation.

179
00:13:32,480 --> 00:13:36,280
I think my biggest challenge I've had with it in the conversations that I've had online

180
00:13:36,280 --> 00:13:42,400
or voyeur'd online, some of which I've been proactive, some I've just read, is just the

181
00:13:42,400 --> 00:13:45,920
belief that these things are going to be quite easy to sidestep.

182
00:13:45,920 --> 00:13:49,400
They have been so far and they will be in the future.

183
00:13:49,400 --> 00:13:53,920
Whether this ends up being non-news or not because it gets sidestepped very quickly,

184
00:13:53,920 --> 00:13:55,320
I don't know, but I think you're right.

185
00:13:55,320 --> 00:13:59,800
I think this is where we're at in terms of people going, oh, you took our content.

186
00:13:59,800 --> 00:14:01,160
You shouldn't have.

187
00:14:01,160 --> 00:14:02,160
We want some money for it.

188
00:14:02,160 --> 00:14:05,160
And by the way, we want a mechanism to stop it happening in the future.

189
00:14:05,160 --> 00:14:07,520
That's kind of where we're at.

190
00:14:07,520 --> 00:14:11,880
Then hopefully that will get resolved very quickly and doesn't pull any of the tools

191
00:14:11,880 --> 00:14:15,560
down in terms of stopping us being able to use them.

192
00:14:15,560 --> 00:14:20,760
I think some of us have become somewhat reliant on maybe using them in our workflow.

193
00:14:20,760 --> 00:14:24,680
We've probably got to be mindful of that, but that hasn't happened yet, obviously.

194
00:14:24,680 --> 00:14:27,160
What about our next story then, Martin?

195
00:14:27,160 --> 00:14:31,160
Tell us about A Meta's personas.

196
00:14:31,160 --> 00:14:35,440
Meta are at it again announcing some new products.

197
00:14:35,440 --> 00:14:41,880
So they've developed some AI-powered chatbots with unique personas to deploy across platforms

198
00:14:41,880 --> 00:14:48,960
such as Facebook and Instagram and no doubt in the future, Threads.

199
00:14:48,960 --> 00:14:49,960
If it's still going.

200
00:14:49,960 --> 00:14:57,040
I saw it had an 80% drop off in active users in a news story this week.

201
00:14:57,040 --> 00:15:03,600
This isn't a social media podcast, but yes, Threads asterisk.

202
00:15:03,600 --> 00:15:06,800
If they can just add search to that tool, I will use it.

203
00:15:06,800 --> 00:15:08,120
I'm in there frequently enough.

204
00:15:08,120 --> 00:15:09,980
I just want to be able to search.

205
00:15:09,980 --> 00:15:11,600
Let me click on a hashtag.

206
00:15:11,600 --> 00:15:13,880
I don't ask for much.

207
00:15:13,880 --> 00:15:14,880
Anyway.

208
00:15:14,880 --> 00:15:21,120
Anyway, so these bots, these personas aim to boost user engagement through personalized

209
00:15:21,120 --> 00:15:26,540
recommendations and conversational features, much like we'll have seen through the likes

210
00:15:26,540 --> 00:15:29,880
of Inflections Pi, if you've been using that.

211
00:15:29,880 --> 00:15:37,360
So it will allow Meta to also gather more user data, which you guessed it, can be used

212
00:15:37,360 --> 00:15:46,240
for improved ad targeting, which I think has got some privacy advocates, got them a bit

213
00:15:46,240 --> 00:15:50,400
concerned maybe, but you know, watch this space.

214
00:15:50,400 --> 00:15:56,000
So I think we're going to see more details about this at their conference in September

215
00:15:56,000 --> 00:15:59,120
at the Connect conference.

216
00:15:59,120 --> 00:16:01,160
What can we kind of expect as marketers though?

217
00:16:01,160 --> 00:16:02,600
What's important for us?

218
00:16:02,600 --> 00:16:09,100
Well, I think, you know, opportunities and challenges are very much at play here.

219
00:16:09,100 --> 00:16:15,500
On the one hand, these persona based bots should be able to give us a better understanding

220
00:16:15,500 --> 00:16:20,440
of our target audiences and deliver more relevant personalized experience.

221
00:16:20,440 --> 00:16:25,960
I imagine those experiences will be genuinely engaging if we can plug them into our, our

222
00:16:25,960 --> 00:16:33,400
corporate brand tone of voice and you know, into things like knowledge bases and things

223
00:16:33,400 --> 00:16:34,520
like that.

224
00:16:34,520 --> 00:16:38,560
When you plug that into the likes of Llama and the recent update with Llama 2, this is

225
00:16:38,560 --> 00:16:44,160
going to be quite a powerful tool, I would imagine.

226
00:16:44,160 --> 00:16:48,080
So yeah, lots of new chatbot driven marketing opportunities.

227
00:16:48,080 --> 00:16:52,720
I think Facebook Messenger chatbots always felt like they were slightly underpowered

228
00:16:52,720 --> 00:16:53,720
historically.

229
00:16:53,720 --> 00:16:58,080
I don't know whenever I've tried to interact with them or even launched one for a client

230
00:16:58,080 --> 00:17:02,040
and I've done several times, they always just felt a little bit weak.

231
00:17:02,040 --> 00:17:06,880
So this potentially could open up some really new and exciting avenues.

232
00:17:06,880 --> 00:17:12,480
However, on the flip side, as I mentioned, privacy is going to be something of a concern

233
00:17:12,480 --> 00:17:13,480
for people.

234
00:17:13,480 --> 00:17:20,760
So I think we'll need to closely monitor how Meta plans to leverage this additional user

235
00:17:20,760 --> 00:17:22,800
data.

236
00:17:22,800 --> 00:17:27,360
This is obviously a massive consumer concern at the moment.

237
00:17:27,360 --> 00:17:34,120
So we must ensure that whatever approach they take respects both user consent expectations

238
00:17:34,120 --> 00:17:35,120
and boundaries.

239
00:17:35,120 --> 00:17:38,880
Yeah, so that's an interesting new product.

240
00:17:38,880 --> 00:17:43,440
I personally am quite excited to roll this out for some clients.

241
00:17:43,440 --> 00:17:52,320
I think when you see consumer interactions on Facebook Messenger at scale, the existing

242
00:17:52,320 --> 00:18:02,040
chatbots are quite limited and I think there's massive opportunity here to create more authentic

243
00:18:02,040 --> 00:18:04,000
feeling brand experiences.

244
00:18:04,000 --> 00:18:06,160
Yeah, I agree.

245
00:18:06,160 --> 00:18:10,880
And I think it'd be really interesting to see how it plays out really.

246
00:18:10,880 --> 00:18:17,520
Will it be like high, but maybe a bit more commercialized in terms of products and services

247
00:18:17,520 --> 00:18:26,440
type conversations or is Meta actually trying to leverage this very large base of users

248
00:18:26,440 --> 00:18:33,040
to actually over time switch those users to relying on its chat and persona interfaces

249
00:18:33,040 --> 00:18:39,040
rather than chat GPT-Pi or others, which I think is a TBD at the moment and maybe we'll

250
00:18:39,040 --> 00:18:43,000
get a bit more information in their conference in September.

251
00:18:43,000 --> 00:18:48,000
While we're talking about Meta, they have released some more news this week around some

252
00:18:48,000 --> 00:18:51,440
of the big AI bets that they are making on Instagram.

253
00:18:51,440 --> 00:18:57,360
So outside of this persona's tool and work that they're doing, they've been cooking

254
00:18:57,360 --> 00:19:02,440
up some new AI powered features for marketers and creators on the Instagram platform.

255
00:19:02,440 --> 00:19:09,080
So they are reportedly developing tools that will label AI generated images, summarize

256
00:19:09,080 --> 00:19:16,320
DMs and enhance editing for your images with AI driven super clever brushes.

257
00:19:16,320 --> 00:19:22,160
There are no info really yet on when these new capabilities are going to be launched,

258
00:19:22,160 --> 00:19:29,960
but certainly the news is chock-a-block with Meta AI developments.

259
00:19:29,960 --> 00:19:36,200
For us as marketers, certainly the AI labeling could help manage misinformation and build

260
00:19:36,200 --> 00:19:39,440
trust in terms of what images are real and what are not.

261
00:19:39,440 --> 00:19:43,200
So again, back to the thread of this entire podcast so far.

262
00:19:43,200 --> 00:19:47,840
Some of the other applications have very practical benefits.

263
00:19:47,840 --> 00:19:54,680
So the DM summaries could help brands and influencers manage their direct message conversations

264
00:19:54,680 --> 00:19:55,680
at scale.

265
00:19:55,680 --> 00:19:59,800
And it looks like some of the new editing tools could make it easier for content creators

266
00:19:59,800 --> 00:20:03,800
to create really high quality content quickly and easily.

267
00:20:03,800 --> 00:20:09,720
So for us as marketers, we know that AI is playing a bigger role in lots of different

268
00:20:09,720 --> 00:20:10,720
social platforms.

269
00:20:10,720 --> 00:20:15,360
LinkedIn are doing a fair bit of this as well, but we just need to keep an eye on this because

270
00:20:15,360 --> 00:20:20,160
if you use Instagram as a channel, really understanding how you can supercharge your

271
00:20:20,160 --> 00:20:27,340
approach using these AI driven tools, whilst at the same time being mindful of other aspects

272
00:20:27,340 --> 00:20:30,120
of how the platform works that you might need to be mindful of.

273
00:20:30,120 --> 00:20:35,360
For example, for those of you out there generating AI images and passing them off as your own

274
00:20:35,360 --> 00:20:40,480
photography, something like this automatic image labeling tool is probably going to get

275
00:20:40,480 --> 00:20:42,480
you in a bit of hot water.

276
00:20:42,480 --> 00:20:48,760
And we perhaps need to be mindful that whenever we are using AI generated images in our marketing

277
00:20:48,760 --> 00:20:54,260
programs, especially on Instagram in this case, how we as a brand feel about the whole

278
00:20:54,260 --> 00:20:59,400
world knowing that the images that we are using were created by AI versus human graphic

279
00:20:59,400 --> 00:21:00,400
designers.

280
00:21:00,400 --> 00:21:01,680
Could we see backlashes to brands?

281
00:21:01,680 --> 00:21:02,680
Who knows?

282
00:21:02,680 --> 00:21:06,240
So I think there's a bunch of other complex stuff that we just need to keep an eye on

283
00:21:06,240 --> 00:21:08,120
here.

284
00:21:08,120 --> 00:21:13,160
The DM summaries for brands that get a huge amount of views and engagement, I think that's

285
00:21:13,160 --> 00:21:15,800
going to be massive.

286
00:21:15,800 --> 00:21:20,120
That will help you know, just anything that can help manage a busy inbox.

287
00:21:20,120 --> 00:21:25,520
So if you're a social media community manager, that is a tool to keep an eye on.

288
00:21:25,520 --> 00:21:30,800
Yeah, auto flagging the high priority messages based on content and sentiment and all these

289
00:21:30,800 --> 00:21:31,800
other things.

290
00:21:31,800 --> 00:21:32,800
Yeah, completely agree.

291
00:21:32,800 --> 00:21:36,200
What else has been happening this week, Martin?

292
00:21:36,200 --> 00:21:38,120
Google, they're in the game.

293
00:21:38,120 --> 00:21:39,120
Oh, Google.

294
00:21:39,120 --> 00:21:40,520
Thanks for stopping by.

295
00:21:40,520 --> 00:21:42,840
What have they been up to?

296
00:21:42,840 --> 00:21:48,760
They are planning to give Google Assistant a major upgrade by incorporating generative

297
00:21:48,760 --> 00:21:49,760
AI.

298
00:21:49,760 --> 00:21:56,120
So that means that Google Assistant, which, Droid, you probably have this active quite

299
00:21:56,120 --> 00:22:02,360
a bit, it will soon have similar capabilities to ChatGPT and Bard.

300
00:22:02,360 --> 00:22:08,200
This has apparently been revealed in an internal email sent to Google employees.

301
00:22:08,200 --> 00:22:11,680
So why does this matter?

302
00:22:11,680 --> 00:22:16,640
Well, Assistant is a really big interface for consumers.

303
00:22:16,640 --> 00:22:22,040
It's a channel that as marketers, we don't really think about, but it is how lots of

304
00:22:22,040 --> 00:22:28,960
people interface every day with Google, the biggest search engine in the world.

305
00:22:28,960 --> 00:22:36,960
So being aware of these new capabilities and thinking about how we use Google as a general

306
00:22:36,960 --> 00:22:39,840
channel to reach people is going to be quite important.

307
00:22:39,840 --> 00:22:47,880
Now, this revamp is taking place apparently on the mobile version first.

308
00:22:47,880 --> 00:22:51,520
I think that's how most people will use Assistant anyway.

309
00:22:51,520 --> 00:22:56,400
I'm not even sure I've ever used Google Assistant on desktop.

310
00:22:56,400 --> 00:23:01,400
So that's where we expect the revamp to begin.

311
00:23:01,400 --> 00:23:07,000
Google not the only company making moves in this area in terms of grading their smart

312
00:23:07,000 --> 00:23:08,000
Assistant.

313
00:23:08,000 --> 00:23:15,440
Amazon is also believed to be working on a bit of an AI powered reboot for the Assistant

314
00:23:15,440 --> 00:23:22,240
who shall not be named the Fear of Triggering Kit right now.

315
00:23:22,240 --> 00:23:26,320
Every time I come to that word, I think, stop.

316
00:23:26,320 --> 00:23:30,840
And so yeah, if you're listening to this on your speakers at home, be thankful that I

317
00:23:30,840 --> 00:23:32,280
didn't just trigger it.

318
00:23:32,280 --> 00:23:39,960
Okay, so Google and Amazon, they've really put less focus on the smart Assistants in

319
00:23:39,960 --> 00:23:40,960
recent.

320
00:23:40,960 --> 00:23:46,280
They've remained relatively static in terms of their capabilities.

321
00:23:46,280 --> 00:23:56,680
But the resurgence and influx of users into generative AI has them all scrambling to kick

322
00:23:56,680 --> 00:24:03,280
start their digital Assistant programs once again.

323
00:24:03,280 --> 00:24:07,960
It feels to me like we've been talking for years like SEO experts that want to be ahead

324
00:24:07,960 --> 00:24:12,400
of the game and advising us on how we need to pivot and evolve our SEO strategies have

325
00:24:12,400 --> 00:24:15,480
been like, yeah, you're going to be needing to think voice first.

326
00:24:15,480 --> 00:24:16,680
You got to think voice first.

327
00:24:16,680 --> 00:24:21,880
And I'm not really sure that has emerged yet.

328
00:24:21,880 --> 00:24:25,200
This may further change how people interact with computers.

329
00:24:25,200 --> 00:24:32,560
Again, we've got to use, I am on the potential future of chat GPT and the emergence of GPT-5

330
00:24:32,560 --> 00:24:37,160
that we can look at in a moment that's also aligned to this, but it could certainly change

331
00:24:37,160 --> 00:24:38,520
us how we're using them.

332
00:24:38,520 --> 00:24:45,280
And as a little tangent, we've talked a bit on the podcast about building our own tools

333
00:24:45,280 --> 00:24:51,720
with you've done some really awesome stuff taking auto, taking a recording of a piece

334
00:24:51,720 --> 00:24:56,760
of audio like when you do some of your live conference workshops and automatically turning

335
00:24:56,760 --> 00:25:01,280
that into a blog post through a transcription series of process steps, et cetera.

336
00:25:01,280 --> 00:25:05,800
One of the things I've been getting really frustrated with recently is I asked chat GPT

337
00:25:05,800 --> 00:25:07,080
to help me build a Chrome plugin.

338
00:25:07,080 --> 00:25:08,080
It was awesome.

339
00:25:08,080 --> 00:25:09,080
We had a great conversation about it.

340
00:25:09,080 --> 00:25:16,720
One of the things I learned about it, assuming this is true, is you can do a lot of transcription

341
00:25:16,720 --> 00:25:18,240
in the browser.

342
00:25:18,240 --> 00:25:23,800
There is a built-in piece of code in the browser to do transcription, but it's the buggy transcription

343
00:25:23,800 --> 00:25:29,940
that we're all used to using whenever we perhaps even speak to your keyboard, which I often

344
00:25:29,940 --> 00:25:35,840
try and do when I want to send text messages or what have you or write emails, which just

345
00:25:35,840 --> 00:25:45,400
completely lacks the power of what transformer models are brought to transcription.

346
00:25:45,400 --> 00:25:51,240
So we're drifting into that period where it's going to be easier and easier to just speak

347
00:25:51,240 --> 00:25:52,520
to your computer.

348
00:25:52,520 --> 00:25:55,920
I find it much easier to get my ideas down if I speak.

349
00:25:55,920 --> 00:26:00,380
I don't think I've mentioned it on the podcast yet, but as you know, Martin and I abandoned

350
00:26:00,380 --> 00:26:05,960
some of the workflows I was building because I found this awesome tool called audiopen.ai,

351
00:26:05,960 --> 00:26:11,600
absolute sexy transcription tool that I speak to on my phone or in the browser.

352
00:26:11,600 --> 00:26:15,360
And not only does it transcribe it beautifully and keep a record of that, it uses the same

353
00:26:15,360 --> 00:26:20,480
uses GPT-4 to summarize it in a number of any different styles that I want to use.

354
00:26:20,480 --> 00:26:25,440
And you can certainly imagine that you might be able to now be in the kitchen cooking and

355
00:26:25,440 --> 00:26:31,240
dictating emails to Google Assistant in a way where not only does it really understand

356
00:26:31,240 --> 00:26:35,160
you well, but actually it can take the content you give it and transcribe it in a meaningful

357
00:26:35,160 --> 00:26:36,160
way.

358
00:26:36,160 --> 00:26:37,160
For me, that'd be pretty cool.

359
00:26:37,160 --> 00:26:39,680
I think we even talked about, Martin, what are you thinking about doing this in a car

360
00:26:39,680 --> 00:26:40,680
at one point?

361
00:26:40,680 --> 00:26:45,600
Just as you were saying that, I thought, ah yeah, that reminds me of the use case I had

362
00:26:45,600 --> 00:26:53,720
where I use Google Auto or Android Auto, should I say, all the time for sending WhatsApp messages

363
00:26:53,720 --> 00:26:56,640
to my wife or telling it to make phone calls.

364
00:26:56,640 --> 00:27:02,160
And the amount of times I've wanted to be able to say, add something to my calendar

365
00:27:02,160 --> 00:27:07,800
or draft an email to such and such and just give it more complex tasks.

366
00:27:07,800 --> 00:27:10,960
And I know it's not been capable of doing it.

367
00:27:10,960 --> 00:27:14,680
I've just wanted to record voice notes that I wanted it to turn into blog posts.

368
00:27:14,680 --> 00:27:17,920
Again, not being able to do it.

369
00:27:17,920 --> 00:27:27,640
As an aside, the Android app for ChatGPT dropped a couple of weeks ago, I think it was, globally.

370
00:27:27,640 --> 00:27:31,520
So I've been playing with that and that has audio input, doesn't it?

371
00:27:31,520 --> 00:27:34,200
So you can actually voice record.

372
00:27:34,200 --> 00:27:37,320
And that's been a revelation for me.

373
00:27:37,320 --> 00:27:41,440
So as I was just wandering around the house, just capturing a few voice notes and seeing

374
00:27:41,440 --> 00:27:43,160
what I get out of it.

375
00:27:43,160 --> 00:27:47,760
Really useful if people aren't using that, check it out because it uses the Whisper API

376
00:27:47,760 --> 00:27:54,200
and you can just say, right, draft me my email nurturing campaign for this.

377
00:27:54,200 --> 00:27:57,920
In fact, I did it the other day for a client.

378
00:27:57,920 --> 00:28:04,880
I came up with an email nurturing campaign for their upcoming product launch.

379
00:28:04,880 --> 00:28:06,160
Yeah, it worked street.

380
00:28:06,160 --> 00:28:09,320
And you were just dictating it as you were wandering around the house instead of having

381
00:28:09,320 --> 00:28:11,000
to try and type it out, right?

382
00:28:11,000 --> 00:28:15,960
Yeah, just I was literally making a cup of tea.

383
00:28:15,960 --> 00:28:22,360
There's a movie called Her, which is about AI assistants where in the end people end

384
00:28:22,360 --> 00:28:24,440
up falling in love with their AI assistants.

385
00:28:24,440 --> 00:28:29,040
It's a little bit too left field probably for us to get into onto the podcast today.

386
00:28:29,040 --> 00:28:34,040
But what the world ends up looking like in that world is everybody has an earpiece in

387
00:28:34,040 --> 00:28:38,560
and they're not looking at their phone so much as constantly talking to their assistant.

388
00:28:38,560 --> 00:28:42,920
So it would be interesting to think about whether it really does push us to interact

389
00:28:42,920 --> 00:28:51,920
with computers via voice and a little segue into our last story this week, which oh no,

390
00:28:51,920 --> 00:28:57,240
it's our second to last story I should say, which is that OpenAI has filed the trademark

391
00:28:57,240 --> 00:29:01,880
for GPT-5, which is perhaps not super surprising, right?

392
00:29:01,880 --> 00:29:08,160
We're going to need that, but the trademark application has certain aspects in it around

393
00:29:08,160 --> 00:29:11,040
what they expect GPT-5 to include.

394
00:29:11,040 --> 00:29:14,000
And so maybe that's where the more interesting stuff is.

395
00:29:14,000 --> 00:29:18,680
So the application, which was submitted a couple of weeks ago, July 18th, is currently

396
00:29:18,680 --> 00:29:26,320
in process and it covers various aspects of AI tech like artificial speech, audio to text

397
00:29:26,320 --> 00:29:32,720
conversion, voice and speech recognition, and a bunch of other stuff as it relates to speech

398
00:29:32,720 --> 00:29:33,720
processing.

399
00:29:33,720 --> 00:29:41,200
And what it really feels like is trying to turn chat GPT from a text-based tool into

400
00:29:41,200 --> 00:29:44,940
a tool you can speak to and it speaks back to you.

401
00:29:44,940 --> 00:29:47,600
That's kind of, I think, some of the inferences here.

402
00:29:47,600 --> 00:29:52,440
And as marketers, if we even just take the first order consequences of that, creating

403
00:29:52,440 --> 00:29:58,080
more interactive and immersive customer experiences that are more like spoken conversations with

404
00:29:58,080 --> 00:29:59,080
real people, right?

405
00:29:59,080 --> 00:30:04,440
If you can skin that tech as part of your customer service offering, either when people

406
00:30:04,440 --> 00:30:10,120
ring you on the phone or when they go on your website to talk to a chat bot, they can choose

407
00:30:10,120 --> 00:30:13,160
to have it as a text interaction or as a spoken interaction.

408
00:30:13,160 --> 00:30:18,120
It's just going to make it more like dealing with real people.

409
00:30:18,120 --> 00:30:21,640
This is something you can already access a bit like this.

410
00:30:21,640 --> 00:30:24,480
So we've talked on the podcast before about Pi.

411
00:30:24,480 --> 00:30:28,860
So if you're using Pi on WhatsApp, you can basically record a voice note in WhatsApp.

412
00:30:28,860 --> 00:30:31,340
It will transcribe it and then respond to you in text.

413
00:30:31,340 --> 00:30:36,200
And if you interact with Pi on the web app, then you have to write what you are trying

414
00:30:36,200 --> 00:30:37,760
to say, but then it speaks back to you.

415
00:30:37,760 --> 00:30:41,520
So it's almost like they've untangled those two things and haven't brought them together

416
00:30:41,520 --> 00:30:43,340
in the same place yet.

417
00:30:43,340 --> 00:30:46,280
But there does seem to be a movement towards that.

418
00:30:46,280 --> 00:30:50,400
And of course, with your news there, Martin, about Google Assistant, I do feel that's where

419
00:30:50,400 --> 00:30:52,480
we're probably moving.

420
00:30:52,480 --> 00:30:59,200
Another key question here is GPT-5 and the trademarking, is it really coming or is this

421
00:30:59,200 --> 00:31:02,000
again a bit of sort of buzz and non-news?

422
00:31:02,000 --> 00:31:04,880
And the senior team at OpenAI were quite lucky.

423
00:31:04,880 --> 00:31:06,320
They're relatively vocal, aren't they?

424
00:31:06,320 --> 00:31:07,720
We'll hear from them on Twitter.

425
00:31:07,720 --> 00:31:10,120
We'll hear from them on a number of podcast interviews.

426
00:31:10,120 --> 00:31:13,240
So we do get a reasonable stream of info coming out of the company.

427
00:31:13,240 --> 00:31:16,720
But the story around GPT-5 is muddled at best as far as I can tell.

428
00:31:16,720 --> 00:31:20,480
And whether we're going to see it in the near future is very open to debate.

429
00:31:20,480 --> 00:31:28,040
The most recent thing I heard or read was they are basically primed to do a GPT-5 training

430
00:31:28,040 --> 00:31:29,640
run at any moment.

431
00:31:29,640 --> 00:31:34,160
But it sounds to me like they're actually using market conditions and the activities

432
00:31:34,160 --> 00:31:40,840
of their competitors to inform that strategy rather than actually having a scheduled plan

433
00:31:40,840 --> 00:31:44,640
for when GPT-5 will be trained and live.

434
00:31:44,640 --> 00:31:49,920
But again, all of that just comes with a fairly large dose of salt.

435
00:31:49,920 --> 00:32:01,360
I am not even thinking about GPT-5 until I've got multimodal image inputs with GPT-4 and

436
00:32:01,360 --> 00:32:03,520
a 32k window.

437
00:32:03,520 --> 00:32:04,760
As promised.

438
00:32:04,760 --> 00:32:05,760
As promised.

439
00:32:05,760 --> 00:32:06,760
Yeah.

440
00:32:06,760 --> 00:32:09,680
Let's get that first and we'll worry about the rest later.

441
00:32:09,680 --> 00:32:13,320
I suspect that that is the truest state of affairs, Martin.

442
00:32:13,320 --> 00:32:14,320
I agree.

443
00:32:14,320 --> 00:32:17,080
Why don't you give us our last story?

444
00:32:17,080 --> 00:32:18,760
Back to Meta for this one.

445
00:32:18,760 --> 00:32:25,360
They've open sourced again because they are champions of open sourcing models, AudioCraft,

446
00:32:25,360 --> 00:32:29,040
which is some kind of AI sound magic.

447
00:32:29,040 --> 00:32:36,000
So it's a text to audio tool called AudioCraft and it can generate any type of sound that

448
00:32:36,000 --> 00:32:37,560
you can imagine.

449
00:32:37,560 --> 00:32:38,560
Apparently.

450
00:32:38,560 --> 00:32:46,000
So we're talking about music, melodies, soundtracks, sound effects, even breaking down sounds into

451
00:32:46,000 --> 00:32:47,000
smaller pieces.

452
00:32:47,000 --> 00:32:54,840
So if you wanted to say a person stood in a large echoey cathedral clicking their fingers,

453
00:32:54,840 --> 00:32:57,760
it would apparently be able to do this.

454
00:32:57,760 --> 00:32:59,960
So what about its music capabilities?

455
00:32:59,960 --> 00:33:06,820
Well, you can pretty much make songs, compositions, melodies, simply by describing it.

456
00:33:06,820 --> 00:33:13,960
So whether you've got the kind of type of music, the genre, instrumentals, temp, all

457
00:33:13,960 --> 00:33:20,080
of this, even maybe the style of a specific musician or a song, you can just say it and

458
00:33:20,080 --> 00:33:22,680
then it will produce it.

459
00:33:22,680 --> 00:33:31,640
Now, again, this isn't just music, as I say, so sound effects, you can use this for films,

460
00:33:31,640 --> 00:33:38,520
video games, podcasts, add jingles, you know, it could be the next radio jingle and the

461
00:33:38,520 --> 00:33:39,520
next earworm.

462
00:33:39,520 --> 00:33:49,040
I love the idea of using AI to create a jingle that you then share using FM radio, old school

463
00:33:49,040 --> 00:33:50,040
tech.

464
00:33:50,040 --> 00:33:55,440
Yeah, that's what we marketers do.

465
00:33:55,440 --> 00:34:03,280
So apparently Meta has gone above and beyond to ensure that every of the 20,000 hours of

466
00:34:03,280 --> 00:34:09,640
music that has been used to train audio craft has been properly licensed.

467
00:34:09,640 --> 00:34:15,760
So the discussion that we've talked about all in today's episode about having content

468
00:34:15,760 --> 00:34:23,520
that is licensed and tagged and marked as training data, they have ticked that bill.

469
00:34:23,520 --> 00:34:28,040
So we don't have to worry about any copyright issues there.

470
00:34:28,040 --> 00:34:32,960
Audio craft is now open source, so I'm sure you're going to be able to see some very innovative

471
00:34:32,960 --> 00:34:40,040
projects in the near future with various products being built on top of it.

472
00:34:40,040 --> 00:34:45,040
So you can get your hands on the code and start building your own audio masterpieces

473
00:34:45,040 --> 00:34:46,040
today.

474
00:34:46,040 --> 00:34:50,920
Yeah, it sounds like from the chat online, I've only seen a bit of it, to be honest,

475
00:34:50,920 --> 00:34:58,440
but this is the new standard in being able to synthesize different types of audio through

476
00:34:58,440 --> 00:35:00,320
text alone.

477
00:35:00,320 --> 00:35:04,200
And I think Google had some tools for this, but apparently these seem much better.

478
00:35:04,200 --> 00:35:11,920
And I think take in in summary, the ongoing explosion in generative AI continues, whether

479
00:35:11,920 --> 00:35:16,500
that's images and text, which we've become used to, but now we're starting to see the

480
00:35:16,500 --> 00:35:22,560
emergence of video and music and other audio really starting to catch up and offer us a

481
00:35:22,560 --> 00:35:26,280
bunch of opportunities to do cool stuff there.

482
00:35:26,280 --> 00:35:33,240
There's one thing on that I just want to touch on, which is the, there's an interesting thought

483
00:35:33,240 --> 00:35:37,920
in this space if you're following prompt engineering communities or anything, and it's basically

484
00:35:37,920 --> 00:35:44,040
this idea that in order to get the best outputs from any generative AI model, so chat GPT

485
00:35:44,040 --> 00:35:49,280
or what have you, the best way to get great outputs is to be a subject matter expert in

486
00:35:49,280 --> 00:35:54,760
the thing that you're talking about, because you know the right questions to ask, the nuances,

487
00:35:54,760 --> 00:35:59,880
the subtleties, you know where it's not quite got it right.

488
00:35:59,880 --> 00:36:02,600
I think this is going to be exactly the same, right?

489
00:36:02,600 --> 00:36:07,720
I do not have a musical bone in my body.

490
00:36:07,720 --> 00:36:13,640
If you give me the audio craft, I will give you crappy audio output, but I would well

491
00:36:13,640 --> 00:36:19,040
imagine that this generative AI model in the hands of a composer is going to be an awesome

492
00:36:19,040 --> 00:36:24,240
tool that's going to be able to expand the capabilities of musicians, composers, SFX

493
00:36:24,240 --> 00:36:26,240
engineers, all the rest of it.

494
00:36:26,240 --> 00:36:29,280
It's going to be a boom for them.

495
00:36:29,280 --> 00:36:30,840
I completely agree.

496
00:36:30,840 --> 00:36:36,760
I guess the other way of looking at that would be also, if I think about image generation

497
00:36:36,760 --> 00:36:44,040
tools, the early versions of this might be frustrating for skilled musicians to use because

498
00:36:44,040 --> 00:36:47,920
they know what they want and they try and explain it, but the tools can't quite give

499
00:36:47,920 --> 00:36:52,700
it and then they can't easily amend it or edit it, which is what mid-journey images

500
00:36:52,700 --> 00:36:55,040
to a certain extent are still like.

501
00:36:55,040 --> 00:36:59,840
You describe it, but if you can't quite get the thing you want, you've got to iterate

502
00:36:59,840 --> 00:37:02,980
through your prompts forever and you think crumbs, maybe it would have been easier to

503
00:37:02,980 --> 00:37:04,720
just do this a different way.

504
00:37:04,720 --> 00:37:10,000
In some cases, maybe we'll see that as well in music and video generation for a while

505
00:37:10,000 --> 00:37:11,000
yet.

506
00:37:11,000 --> 00:37:12,000
Right.

507
00:37:12,000 --> 00:37:17,960
Well, the last part today then is our interview with Brennan Woodruff.

508
00:37:17,960 --> 00:37:21,520
Tell us a bit about this interview as a quick introduction, Martin.

509
00:37:21,520 --> 00:37:27,800
In my conversation with Brennan, he is the co-founder of Go Charlie.

510
00:37:27,800 --> 00:37:33,280
We explore all sorts of interesting themes from the realities of building a startup inside

511
00:37:33,280 --> 00:37:42,240
the rapidly evolving AI space through to how the product itself will empower businesses

512
00:37:42,240 --> 00:37:45,280
and marketers to make content at scale.

513
00:37:45,280 --> 00:37:52,040
But we also have a fascinating discussion around AI agents and how AI agents could automate

514
00:37:52,040 --> 00:37:54,960
passive income streams in the future.

515
00:37:54,960 --> 00:38:01,400
The discussion around AI agents is definitely, or it's a space to watch in the future.

516
00:38:01,400 --> 00:38:07,480
Not just the discussion, but yeah, AI agents, autonomous beings operating 24-7.

517
00:38:07,480 --> 00:38:12,480
Imagine chat GPT just running in the background doing things for you all day, every day.

518
00:38:12,480 --> 00:38:15,240
Yeah, it was definitely thought provoking.

519
00:38:15,240 --> 00:38:20,400
Well, I will look forward to listening to it and hopefully all of you will now as we

520
00:38:20,400 --> 00:38:22,080
segue into it.

521
00:38:22,080 --> 00:38:24,080
Other than that, thank you for your time again, Martin.

522
00:38:24,080 --> 00:38:26,160
I look forward to chatting to you next week.

523
00:38:26,160 --> 00:38:30,800
Ladies and gentlemen, get ready to meet a pioneer in the world of AI.

524
00:38:30,800 --> 00:38:36,360
He's at the cutting edge of generative AI using it to transform marketing and business.

525
00:38:36,360 --> 00:38:41,520
With Go Charlie, he and his team have created the first multimodal generative content engine

526
00:38:41,520 --> 00:38:44,880
purpose built for marketing.

527
00:38:44,880 --> 00:38:50,400
Go Charlie has taken the world by storm, hitting number one on product hunt not just once,

528
00:38:50,400 --> 00:38:52,400
but twice.

529
00:38:52,400 --> 00:38:56,640
Brennan has a relentless spirit needed for leading a startup.

530
00:38:56,640 --> 00:39:01,600
He's a supportive mentor and he's a great community manager as demonstrated by the passionate

531
00:39:01,600 --> 00:39:05,160
and engaged Go Charlie fans on the company's Facebook group.

532
00:39:05,160 --> 00:39:08,200
All 1700 of them, I believe.

533
00:39:08,200 --> 00:39:13,600
Brennan has assembled a world-class team of AI experts to bring Go Charlie to life.

534
00:39:13,600 --> 00:39:19,200
He's been an advisor to countless startups, always happy to pay it forward.

535
00:39:19,200 --> 00:39:23,800
At his core, Brennan is an innovator driven by creativity and human connection.

536
00:39:23,800 --> 00:39:27,040
He believes AI should empower people, not replace them.

537
00:39:27,040 --> 00:39:31,200
It's this unique blend of tech expertise and emotional IQ that makes him such an interesting

538
00:39:31,200 --> 00:39:33,040
person to speak to today.

539
00:39:33,040 --> 00:39:39,480
So get ready for an enlightening and entertaining chat with the one and only Brennan Woodrow,

540
00:39:39,480 --> 00:39:44,000
co-founder and chief business officer at Go Charlie.

541
00:39:44,000 --> 00:39:45,920
Wow.

542
00:39:45,920 --> 00:39:47,000
Quite the introduction.

543
00:39:47,000 --> 00:39:50,280
I would love if you were able to introduce me for every event.

544
00:39:50,280 --> 00:39:51,520
That was fantastic, Martin.

545
00:39:51,520 --> 00:39:55,600
Thank you for such an in-depth discussion.

546
00:39:55,600 --> 00:39:59,760
All the things that you hope that people say about you coming to life.

547
00:39:59,760 --> 00:40:00,760
This is awesome.

548
00:40:00,760 --> 00:40:01,760
What can I say?

549
00:40:01,760 --> 00:40:05,240
You've got a good LinkedIn profile.

550
00:40:05,240 --> 00:40:07,520
Charlie may have helped us some of those things.

551
00:40:07,520 --> 00:40:08,520
Yeah.

552
00:40:08,520 --> 00:40:13,480
Well, you know, you've got to practice what you preach on that.

553
00:40:13,480 --> 00:40:17,040
So yeah, I mentioned the Facebook group.

554
00:40:17,040 --> 00:40:19,920
That's one of the places where we've engaged a lot.

555
00:40:19,920 --> 00:40:25,200
That's a great little resource that you've built out there.

556
00:40:25,200 --> 00:40:27,480
Yeah.

557
00:40:27,480 --> 00:40:32,360
I think early days with a startup, the biggest problem you can have is not talking to customers,

558
00:40:32,360 --> 00:40:33,800
not engaging with customers.

559
00:40:33,800 --> 00:40:38,680
And so fortunately for us, Facebook ended up being a great community.

560
00:40:38,680 --> 00:40:42,120
We've seen a lot of AI companies go the route of Discord.

561
00:40:42,120 --> 00:40:45,400
I don't know if there's any gamers listening to this, but Discord kind of reminds me of

562
00:40:45,400 --> 00:40:50,680
Unreal Tournament versus like the Halo that is using Slack and Slack's a little more my

563
00:40:50,680 --> 00:40:53,520
speed, Facebook community a little more my speed.

564
00:40:53,520 --> 00:40:56,920
Unreal Tournament is just way beyond my speed.

565
00:40:56,920 --> 00:41:00,520
So we've been fortunate enough to have customers that want to share feedback, want to share

566
00:41:00,520 --> 00:41:02,480
with other customers.

567
00:41:02,480 --> 00:41:05,280
And it's just been a big part of our ethos since we started.

568
00:41:05,280 --> 00:41:06,280
Great.

569
00:41:06,280 --> 00:41:09,960
Now it does seem like there's some good ideas and inspiration.

570
00:41:09,960 --> 00:41:13,080
I see lots of people giving feedback and you're always asking for it as well.

571
00:41:13,080 --> 00:41:18,160
So I feel like I've jumped ahead there maybe just a little bit.

572
00:41:18,160 --> 00:41:21,840
In the introduction, I mentioned GoCharlie.

573
00:41:21,840 --> 00:41:28,240
So can you give us an overview of what GoCharlie is for anyone that hasn't come across it?

574
00:41:28,240 --> 00:41:29,240
Yeah.

575
00:41:29,240 --> 00:41:35,600
So GoCharlie is a generative AI platform for creating top quality marketing content.

576
00:41:35,600 --> 00:41:38,960
We have sort of a thesis on what makes perfect content.

577
00:41:38,960 --> 00:41:43,840
Just a combination of a language model purpose built for the things that you need it to do

578
00:41:43,840 --> 00:41:48,960
combined with a brand voice engine and knowledge of your products and knowledge of your competitive

579
00:41:48,960 --> 00:41:55,080
landscape and then the target audience platform that you're planning to post to and a couple

580
00:41:55,080 --> 00:41:56,600
other special ingredients.

581
00:41:56,600 --> 00:42:01,000
But for us, we tried to build that into a subscription based product so that people

582
00:42:01,000 --> 00:42:04,240
could ultimately grow their business and pursue their dreams.

583
00:42:04,240 --> 00:42:09,720
Both my parents were entrepreneurs themselves and so I saw firsthand the impact that relationship

584
00:42:09,720 --> 00:42:14,360
driven businesses took when everything moved to digital.

585
00:42:14,360 --> 00:42:18,200
And so I wanted to give tools to my parents that they could use to market their own business

586
00:42:18,200 --> 00:42:23,880
and that's where we really got the idea for GoCharlie.

587
00:42:23,880 --> 00:42:31,240
So it's very much focused then on the marketing and kind of business content production and

588
00:42:31,240 --> 00:42:38,840
this isn't a fluffy, I'm sure I can write poems, but this is focused on the business

589
00:42:38,840 --> 00:42:40,440
user.

590
00:42:40,440 --> 00:42:41,440
For now, yes.

591
00:42:41,440 --> 00:42:48,280
I increasingly think that SMBs, even creators, even the everyday person is going to have

592
00:42:48,280 --> 00:42:52,480
to know how to create content to market themselves on the internet.

593
00:42:52,480 --> 00:42:57,640
So while I think B2B kind of fits the construct of what we sort of view as who our target

594
00:42:57,640 --> 00:43:04,560
customer would be, I think that second B is becoming a lot more blurred with C if you

595
00:43:04,560 --> 00:43:06,960
know that B2C versus B2B comparison.

596
00:43:06,960 --> 00:43:13,120
But it's definitely more tailored towards marketers that need a more advanced tool.

597
00:43:13,120 --> 00:43:17,120
But I do think with some of the AI models coming online that you'll start to see us

598
00:43:17,120 --> 00:43:22,920
being able to do some of those more creative works such as poems or my personal favorite

599
00:43:22,920 --> 00:43:24,720
dad jokes and dog puns.

600
00:43:24,720 --> 00:43:30,960
Yes, you do love a dog pun.

601
00:43:30,960 --> 00:43:35,440
What's the story behind the name Charlie and the mascot?

602
00:43:35,440 --> 00:43:40,520
Yeah, so Charlie is, well, maybe a fun fact for listeners.

603
00:43:40,520 --> 00:43:45,520
So originally the company was called Gaudium.ai and I'm going to be honest with you, most

604
00:43:45,520 --> 00:43:47,880
of you probably couldn't pronounce Gaudium.

605
00:43:47,880 --> 00:43:49,640
I couldn't pronounce Gaudium.

606
00:43:49,640 --> 00:43:53,000
Everyone that we got on the phone call with was like Gaudium, Gaudium.

607
00:43:53,000 --> 00:43:55,680
And then more importantly, no one could spell it.

608
00:43:55,680 --> 00:43:57,920
So that was the other big piece of this.

609
00:43:57,920 --> 00:44:00,400
And we kind of took that away.

610
00:44:00,400 --> 00:44:03,920
We were like, well, we've been calling our AI model Charlie.

611
00:44:03,920 --> 00:44:07,240
And then I was like, should we just call the company Charlie?

612
00:44:07,240 --> 00:44:08,240
Like go Charlie.

613
00:44:08,240 --> 00:44:10,880
Like literally you're telling your dog to go do something is kind of like you're telling

614
00:44:10,880 --> 00:44:12,320
your AI to go do something.

615
00:44:12,320 --> 00:44:13,600
It's fun.

616
00:44:13,600 --> 00:44:15,040
It feels magical.

617
00:44:15,040 --> 00:44:17,920
You're telling your dog to go do something same way as AI model.

618
00:44:17,920 --> 00:44:22,800
And so Charlie was actually the name of Kostas, my co-founder's dog.

619
00:44:22,800 --> 00:44:27,560
So there's a real Charlie, which we sometimes occasionally share content about in the Facebook

620
00:44:27,560 --> 00:44:33,840
group, but that really gave birth to this new age AI puppy that many people think we're

621
00:44:33,840 --> 00:44:36,880
selling instead of actual content generation.

622
00:44:36,880 --> 00:44:37,880
Yeah.

623
00:44:37,880 --> 00:44:44,240
Well, I saw that you were asking for a few months ago on LinkedIn asking if there was

624
00:44:44,240 --> 00:44:51,400
any AI animated tools that you could get to animate the dog.

625
00:44:51,400 --> 00:44:56,440
I did think at the time, that's a nice play, but yeah, people are going to start thinking

626
00:44:56,440 --> 00:44:59,400
you're a character.ai before long.

627
00:44:59,400 --> 00:45:00,400
Yeah.

628
00:45:00,400 --> 00:45:01,400
Yeah.

629
00:45:01,400 --> 00:45:05,800
I've actually went back and forth with our new newly hired CMO about like gifts that

630
00:45:05,800 --> 00:45:08,640
I want to give to early customers and investors.

631
00:45:08,640 --> 00:45:12,840
And I was like, we should do a dog collar so people can talk and show like the dog collar

632
00:45:12,840 --> 00:45:13,840
on their dogs.

633
00:45:13,840 --> 00:45:16,400
And she's like, we're not a dog company.

634
00:45:16,400 --> 00:45:21,840
We are a Martech AI company that just happens to have a dog mascot.

635
00:45:21,840 --> 00:45:24,280
And I was like, okay, fine.

636
00:45:24,280 --> 00:45:28,600
But you know, if we get enough puppy lovers, maybe, maybe the Charlie dog collars will

637
00:45:28,600 --> 00:45:29,600
come.

638
00:45:29,600 --> 00:45:30,600
We'll see.

639
00:45:30,600 --> 00:45:31,600
Let's not write it out.

640
00:45:31,600 --> 00:45:36,200
I expect it's here at Martech conferences around the world.

641
00:45:36,200 --> 00:45:40,600
Who needs a laptop sticker when they can work around with their dog collars?

642
00:45:40,600 --> 00:45:45,720
You mentioned there your co-founders and the team behind it.

643
00:45:45,720 --> 00:45:48,440
So who are the brains behind Go Charlie then?

644
00:45:48,440 --> 00:45:49,440
What are their roles?

645
00:45:49,440 --> 00:45:50,440
Yeah.

646
00:45:50,440 --> 00:45:54,800
So we had a four person bounding team, which I know is a little larger than most these

647
00:45:54,800 --> 00:46:00,080
days, but just some huge technical firepower.

648
00:46:00,080 --> 00:46:07,840
So Kosis Hatalis is our CEO, AI PhD, 10 years of applied AI experience, was one of the fathers

649
00:46:07,840 --> 00:46:12,840
of probabilistic forecasting neural networks, which ultimately was what transformers are

650
00:46:12,840 --> 00:46:14,480
based off of.

651
00:46:14,480 --> 00:46:16,040
So he's wicked.

652
00:46:16,040 --> 00:46:18,800
He built pretty much our AI from the ground up.

653
00:46:18,800 --> 00:46:24,280
And then we paired him with Despina Christu, who's our chief AI scientist.

654
00:46:24,280 --> 00:46:26,880
I regularly refer to her as Wonder Woman.

655
00:46:26,880 --> 00:46:32,400
She's built NLP solutions for Oracle and for 500 companies, and has really been tasked

656
00:46:32,400 --> 00:46:37,240
with bringing Charlie to life in the AWS ecosystem as well.

657
00:46:37,240 --> 00:46:38,800
So 20 years of applied AI experience.

658
00:46:38,800 --> 00:46:43,560
And then we have a four time founder CTO, Ryan Carlton, who basically built the product

659
00:46:43,560 --> 00:46:48,480
experience from the ground up by himself, which we have a very advanced product capability

660
00:46:48,480 --> 00:46:49,480
wise.

661
00:46:49,480 --> 00:46:53,320
And so the fact that he was able to do that on his own front end, back end, everything

662
00:46:53,320 --> 00:46:54,440
is incredible.

663
00:46:54,440 --> 00:46:59,200
So I'm just the lucky guy that gets to talk about all these amazing people, try to sell

664
00:46:59,200 --> 00:47:03,360
the product, try to get partnerships, try to fundraise.

665
00:47:03,360 --> 00:47:08,260
And I was previously at SoftBank and Uber's autonomous vehicle group before that.

666
00:47:08,260 --> 00:47:12,520
So we have a pretty good team of AI expertise, I would say.

667
00:47:12,520 --> 00:47:16,080
And now we're really starting to flex that technical muscle over the next year, I would

668
00:47:16,080 --> 00:47:17,080
say.

669
00:47:17,080 --> 00:47:18,080
Yeah.

670
00:47:18,080 --> 00:47:25,360
And I mean, there are some credentials there worth paying attention to for sure.

671
00:47:25,360 --> 00:47:34,120
On the technical things to come, I'm really interested in what you've done and what you've

672
00:47:34,120 --> 00:47:35,120
been talking about.

673
00:47:35,120 --> 00:47:38,760
So you've got kind of a secret sauce, haven't you, over at Go Charlie?

674
00:47:38,760 --> 00:47:44,280
Your marketing focused LLM that is yours from the ground up.

675
00:47:44,280 --> 00:47:45,280
Is that right?

676
00:47:45,280 --> 00:47:48,920
And have I understood that correctly?

677
00:47:48,920 --> 00:47:49,920
Yeah.

678
00:47:49,920 --> 00:47:51,040
Yeah.

679
00:47:51,040 --> 00:47:53,880
So we have taken a slightly different approach.

680
00:47:53,880 --> 00:47:59,280
There's a lot of people trying to do AI for marketing.

681
00:47:59,280 --> 00:48:03,200
I won't go too far into my thought about app layer until maybe a later question that we'll

682
00:48:03,200 --> 00:48:05,880
probably get into, but app layer versus model layer.

683
00:48:05,880 --> 00:48:13,960
But we felt that at the onset, we always wanted to create AI that was useful and valuable.

684
00:48:13,960 --> 00:48:18,520
And for us, that meant creating an AI that could serve as the foundation for the rest

685
00:48:18,520 --> 00:48:20,320
of our product suite.

686
00:48:20,320 --> 00:48:24,880
And so in the early days, Go Charlie was a collection of models.

687
00:48:24,880 --> 00:48:30,880
So we had some open AI, we had some of our own image recognition, computer vision, image

688
00:48:30,880 --> 00:48:37,320
creation, and now we're consolidating that down into just our own product suite.

689
00:48:37,320 --> 00:48:40,440
So we have Charlie, the large language model.

690
00:48:40,440 --> 00:48:43,100
We're also in active development on a multimodal model.

691
00:48:43,100 --> 00:48:45,560
So that will be text and image in, text and image out.

692
00:48:45,560 --> 00:48:50,080
That's with SRI International, the guys that developed Siri, which is pretty cool.

693
00:48:50,080 --> 00:48:55,320
The fact that those guys are in with us is pretty awesome for this guy from Evansville,

694
00:48:55,320 --> 00:48:56,320
Indiana.

695
00:48:56,320 --> 00:49:01,080
And then we're also releasing a Charlie the agent, which is kind of an AI brain that can

696
00:49:01,080 --> 00:49:06,160
sit on top of the large language model and multimodal capabilities, as well as additional

697
00:49:06,160 --> 00:49:07,240
API tooling.

698
00:49:07,240 --> 00:49:12,920
So giving you the ability to execute across a number of thousands of tools and capabilities

699
00:49:12,920 --> 00:49:16,880
and tasks, all with a few simple words.

700
00:49:16,880 --> 00:49:21,040
So yeah, that kind of what we have on the technical horizon, but the large language

701
00:49:21,040 --> 00:49:23,840
model itself has been pretty amazing.

702
00:49:23,840 --> 00:49:28,320
We have like a blind user test, like kind of like a Coke versus Pepsi test, which I

703
00:49:28,320 --> 00:49:31,840
think you may have even gotten to try, which is like you put a prompt in and you get a

704
00:49:31,840 --> 00:49:36,760
prompt or you get an output from Charlie and another state of the art model.

705
00:49:36,760 --> 00:49:41,240
And we're winning over 85% of those tasks, which is just truly exceptional given the

706
00:49:41,240 --> 00:49:44,480
size of the team we have relative to some of these other state of the art models.

707
00:49:44,480 --> 00:49:47,120
So yeah, we're excited.

708
00:49:47,120 --> 00:49:50,960
That's hopefully going to be live to the public here soon via API.

709
00:49:50,960 --> 00:49:54,040
So even more fun to be had with Charlie.

710
00:49:54,040 --> 00:49:55,040
Yeah.

711
00:49:55,040 --> 00:49:59,320
And that the, the ability to do the test, is that publicly available?

712
00:49:59,320 --> 00:50:01,360
Can listeners go and play with that?

713
00:50:01,360 --> 00:50:04,320
Yeah, we have, we have a playground.

714
00:50:04,320 --> 00:50:08,400
We opened it up to the public, so you don't have to necessarily be a Charlie customer,

715
00:50:08,400 --> 00:50:11,840
although we would love for you to be using the software product as well.

716
00:50:11,840 --> 00:50:19,080
I think for me, one of the fascinating things with the playground was just understanding

717
00:50:19,080 --> 00:50:24,160
the depth up in the value of having the software layer.

718
00:50:24,160 --> 00:50:27,600
And I know you've probably experienced this yourself.

719
00:50:27,600 --> 00:50:33,280
You go to open AI's chat, GPT, like that's pretty close to a raw model part, but then

720
00:50:33,280 --> 00:50:39,400
you have apps like Jasper where they've been very successful building on top of those models.

721
00:50:39,400 --> 00:50:43,300
And so you go into a raw playground and you're like, oh, I want it to do this or why can

722
00:50:43,300 --> 00:50:44,880
I export or why can't I do that?

723
00:50:44,880 --> 00:50:47,840
And it's like, okay, well, that's a software solve.

724
00:50:47,840 --> 00:50:53,040
And so you start to think about like how much value does the AI model in and of itself have?

725
00:50:53,040 --> 00:50:57,440
And then more interestingly, I think to both of us is like how that evolves.

726
00:50:57,440 --> 00:51:00,400
Can the AI model be truly the entire product?

727
00:51:00,400 --> 00:51:04,760
Whereas do you need a combination of software plus AI?

728
00:51:04,760 --> 00:51:08,720
And I think the answer is the latter, but interesting thought exercise.

729
00:51:08,720 --> 00:51:11,080
Yeah, I would agree.

730
00:51:11,080 --> 00:51:15,880
And I think we actually see that in the way that chat GPT in and of itself came about,

731
00:51:15,880 --> 00:51:16,880
didn't we?

732
00:51:16,880 --> 00:51:22,360
The large language models are the underlying model behind chat GPT back in November.

733
00:51:22,360 --> 00:51:26,120
November 1st, you could use GPT-3.

734
00:51:26,120 --> 00:51:28,840
You could use it and you could build on it and you could access the model and play with

735
00:51:28,840 --> 00:51:29,840
it.

736
00:51:29,840 --> 00:51:30,840
But nobody did.

737
00:51:30,840 --> 00:51:33,840
I say quote unquote, nobody did.

738
00:51:33,840 --> 00:51:42,000
November 30th or end of December, everybody did.

739
00:51:42,000 --> 00:51:49,400
And it was because it was suddenly as easy as using WhatsApp or using whatever Facebook

740
00:51:49,400 --> 00:51:50,400
messenger.

741
00:51:50,400 --> 00:51:56,360
But like you say, it still is very much the closest thing you can get to having that raw

742
00:51:56,360 --> 00:52:01,080
model interaction that most people would want to use.

743
00:52:01,080 --> 00:52:08,080
Send people to playground.openai.com and show somebody a top P setting and then watch them

744
00:52:08,080 --> 00:52:10,960
panic slightly as they go, I don't know what it does.

745
00:52:10,960 --> 00:52:13,680
And other people say to them, no, to be honest, I'm not entirely sure.

746
00:52:13,680 --> 00:52:15,520
I just move the temperature one.

747
00:52:15,520 --> 00:52:16,520
Yeah.

748
00:52:16,520 --> 00:52:22,160
And you find that some of these early adopters, they're like hard code or hardcore coders.

749
00:52:22,160 --> 00:52:24,760
They may want to get into the technical weeds.

750
00:52:24,760 --> 00:52:29,520
I think mass adoption for these tools, it has to be super simple.

751
00:52:29,520 --> 00:52:33,760
Prompting I think is it's going to have to go the way of obsolete and it's just going

752
00:52:33,760 --> 00:52:36,160
to have to feel like you're talking to a friend.

753
00:52:36,160 --> 00:52:39,880
But there's some people that really want to get into the technical weeds and the early

754
00:52:39,880 --> 00:52:41,080
adoption side of things.

755
00:52:41,080 --> 00:52:42,600
They're like, how do I change this scale?

756
00:52:42,600 --> 00:52:44,320
And how do I do this?

757
00:52:44,320 --> 00:52:49,160
That was something that is us building towards mass adoption was an interesting part.

758
00:52:49,160 --> 00:52:53,200
You might be building bells and whistles for these early adopters that the masses don't

759
00:52:53,200 --> 00:52:54,200
care about.

760
00:52:54,200 --> 00:52:55,200
Yeah.

761
00:52:55,200 --> 00:52:56,200
And very much so.

762
00:52:56,200 --> 00:52:59,280
And actually would find actively off putting.

763
00:52:59,280 --> 00:53:00,280
Yeah.

764
00:53:00,280 --> 00:53:01,280
Yeah.

765
00:53:01,280 --> 00:53:09,680
I found when you see tools like Microsoft Copilot and you see the video and how easy

766
00:53:09,680 --> 00:53:15,280
that promises is going to make a classic office productivity and how it promises to do all

767
00:53:15,280 --> 00:53:17,680
of that.

768
00:53:17,680 --> 00:53:21,320
Even that is going to have to be incredibly simple.

769
00:53:21,320 --> 00:53:25,440
You took there about prompting.

770
00:53:25,440 --> 00:53:30,640
Users like my mother who has had an office job all of her life, she knows how to use

771
00:53:30,640 --> 00:53:31,640
Excel.

772
00:53:31,640 --> 00:53:34,480
She knows how to use Microsoft Word and Outlook and all of that.

773
00:53:34,480 --> 00:53:40,320
But does she want to learn how to craft a prompt in order to get her spreadsheet to

774
00:53:40,320 --> 00:53:44,640
do the thing that she's been doing by clicking here, here, here, here and here?

775
00:53:44,640 --> 00:53:47,360
People might say, well, it will save you 20 clicks.

776
00:53:47,360 --> 00:53:51,840
20 clicks is fine if you know the 20 clicks.

777
00:53:51,840 --> 00:53:56,600
Having to sit there and be like, oh God, it's not quite doing it.

778
00:53:56,600 --> 00:54:02,840
Maybe if I do chain of thought prompting, I can get it to make me a pivot table.

779
00:54:02,840 --> 00:54:06,760
No, I can just click on it and it will make me a bloody pivot table.

780
00:54:06,760 --> 00:54:07,760
Yeah.

781
00:54:07,760 --> 00:54:14,240
It's like, I asked my team this question quite a bit, which is like, are we literally spending

782
00:54:14,240 --> 00:54:20,440
a bunch of money and time and resources to get a 5% incremental improvement on this experience?

783
00:54:20,440 --> 00:54:24,680
It would be great if this AI model could make a perfect image on the first try, but if it

784
00:54:24,680 --> 00:54:28,320
doesn't, I'm throwing that image out or I'm giving a different prompt.

785
00:54:28,320 --> 00:54:34,400
If the AI screws up my data table, I'm definitely not coming back to that.

786
00:54:34,400 --> 00:54:39,600
I think we're just as iteration or two away from the AI actually being able to do it with

787
00:54:39,600 --> 00:54:44,820
any sort of reliability, but it always sucks when you see the really plot demo video and

788
00:54:44,820 --> 00:54:46,600
then you get into the product experience.

789
00:54:46,600 --> 00:54:49,960
It is just nothing like it whatsoever.

790
00:54:49,960 --> 00:54:53,360
It's basically like the internet games you see advertised all the time where they look

791
00:54:53,360 --> 00:54:57,880
really cool in the video and then you start playing, you're like, this is awful.

792
00:54:57,880 --> 00:54:58,880
Yeah.

793
00:54:58,880 --> 00:55:02,800
It's the promised land versus the actual reality.

794
00:55:02,800 --> 00:55:07,320
We had a similar thing, actually me and Paul were talking on the show recently about code

795
00:55:07,320 --> 00:55:11,920
interpreter and how he stuck a bunch of data into code interpreter.

796
00:55:11,920 --> 00:55:18,920
It was some data about a client project and he was trying to get it to analyze.

797
00:55:18,920 --> 00:55:19,920
Oh no, it wasn't that.

798
00:55:19,920 --> 00:55:27,720
It was time sheets within his agency and tried to analyze profitable jobs versus non-profitable

799
00:55:27,720 --> 00:55:30,920
jobs and utilization rates and all of that.

800
00:55:30,920 --> 00:55:34,440
He has this already, so he knows the data and that's why he was using it on a data set

801
00:55:34,440 --> 00:55:39,200
that he was familiar with to see if it matched what it told him.

802
00:55:39,200 --> 00:55:40,840
On several occasions it was just wrong.

803
00:55:40,840 --> 00:55:46,800
It's like straight up hallucinated made up stuff.

804
00:55:46,800 --> 00:55:50,040
You don't have to do that many times before you're just like, well, I'm just not going

805
00:55:50,040 --> 00:55:55,480
to bother.

806
00:55:55,480 --> 00:56:01,960
If autonomous self-driving vehicles had the level of reliability that people are flocking

807
00:56:01,960 --> 00:56:12,720
generative AI software at, we would never have autonomous driving vehicles.

808
00:56:12,720 --> 00:56:17,280
We were talking about this before the podcast, that influences where you can actually put

809
00:56:17,280 --> 00:56:20,920
AI into service in your organization.

810
00:56:20,920 --> 00:56:25,320
I even get there with many customers where it's like, maybe a rules-based solution is

811
00:56:25,320 --> 00:56:26,620
what you need.

812
00:56:26,620 --> 00:56:31,740
Maybe you don't need a massive generative AI model to create these personalized emails

813
00:56:31,740 --> 00:56:34,600
where only three sentences are personalized.

814
00:56:34,600 --> 00:56:36,880
Maybe that's just not the thing to do.

815
00:56:36,880 --> 00:56:39,200
I think it's going to be really interesting to see play out.

816
00:56:39,200 --> 00:56:44,280
I think some of that hype is already going away, but even on a personal basis, code interpreter

817
00:56:44,280 --> 00:56:48,760
scared the hell out of me from its functionality suite because it is very impressive.

818
00:56:48,760 --> 00:56:54,560
Some of the things I've seen it do, but reliability is just such a problem in this space, even

819
00:56:54,560 --> 00:56:57,120
from the people that are the best at it.

820
00:56:57,120 --> 00:56:59,040
Yeah, very much so.

821
00:56:59,040 --> 00:57:03,680
It's interesting that you talk there about a rules-based approach.

822
00:57:03,680 --> 00:57:09,400
I had a client approach me recently, wanted to bring me in on an AI project.

823
00:57:09,400 --> 00:57:13,800
They were halfway through this kind of change management piece and wanted to introduce some

824
00:57:13,800 --> 00:57:14,800
AI.

825
00:57:14,800 --> 00:57:19,000
They were very keen to introduce some AI, I should reiterate this, they were really

826
00:57:19,000 --> 00:57:21,120
keen to introduce some AI.

827
00:57:21,120 --> 00:57:24,000
When I sat down with them, they didn't need AI.

828
00:57:24,000 --> 00:57:29,840
They needed some fairly simple marketing automation.

829
00:57:29,840 --> 00:57:35,520
Standard workflow stuff that you would build in a HubSpot campaign, just some rules-based

830
00:57:35,520 --> 00:57:37,920
systems.

831
00:57:37,920 --> 00:57:42,720
Because AI is everywhere, I think it's blinded people.

832
00:57:42,720 --> 00:57:46,920
The solutions are really well established and work really quite well.

833
00:57:46,920 --> 00:57:50,120
Yeah, I think it all gets back to that perception thing.

834
00:57:50,120 --> 00:57:56,600
Most people's perception, even in the business world of what AI is, is all based off of Hollywood.

835
00:57:56,600 --> 00:57:58,400
That's the narrative that the media has picked up too.

836
00:57:58,400 --> 00:57:59,960
It's like mass labor replacement.

837
00:57:59,960 --> 00:58:03,120
It's like, no, no, it's not there yet.

838
00:58:03,120 --> 00:58:06,480
It could potentially get there.

839
00:58:06,480 --> 00:58:10,240
I think that we're going to see a cooling off of the hype cycle as people start to see

840
00:58:10,240 --> 00:58:12,240
that a little bit.

841
00:58:12,240 --> 00:58:14,280
Yeah, I tend to agree.

842
00:58:14,280 --> 00:58:17,480
Now, I said in the intro, you've just said something that's quite interesting about the

843
00:58:17,480 --> 00:58:20,000
Hollywood jobs replacement.

844
00:58:20,000 --> 00:58:26,240
I said in the intro that you believe AI should empower people and not replace them.

845
00:58:26,240 --> 00:58:27,240
Give me that use case then.

846
00:58:27,240 --> 00:58:29,560
I'm a business owner.

847
00:58:29,560 --> 00:58:31,920
I'm looking for an AI solution.

848
00:58:31,920 --> 00:58:35,160
How am I going to turn to GoCharlie in its present form?

849
00:58:35,160 --> 00:58:36,240
What's it going to do for me?

850
00:58:36,240 --> 00:58:40,960
Give me those use cases that your customers are finding really powerful.

851
00:58:40,960 --> 00:58:47,480
Yeah, so I think our platform has been built flexible enough to where different life cycles

852
00:58:47,480 --> 00:58:51,480
of the company, it can have different value levers.

853
00:58:51,480 --> 00:58:54,360
We actually did a session for SF Tech Week.

854
00:58:54,360 --> 00:58:57,480
It was AI driven content creation.

855
00:58:57,480 --> 00:59:00,200
So we literally went through the life cycle of a company.

856
00:59:00,200 --> 00:59:03,800
So I asked like, who in the audience wants to start a company, hasn't really put pen

857
00:59:03,800 --> 00:59:04,800
to paper on it?

858
00:59:04,800 --> 00:59:08,800
The person raised their hand, they came up, we created the name for the company.

859
00:59:08,800 --> 00:59:10,720
We created a brand profile.

860
00:59:10,720 --> 00:59:15,480
We created the channels that they should address first and then the first couple pieces of

861
00:59:15,480 --> 00:59:17,040
content for that.

862
00:59:17,040 --> 00:59:22,720
And then we kind of moved up the maturity cycle into company that already had a website

863
00:59:22,720 --> 00:59:25,080
and was already starting to sell some product.

864
00:59:25,080 --> 00:59:28,500
And for them, it's more just about growth and continuing to tell consistent messaging

865
00:59:28,500 --> 00:59:29,960
across platform.

866
00:59:29,960 --> 00:59:32,680
So we moved from foundational to growth stage.

867
00:59:32,680 --> 00:59:36,360
And from there, we were able to create a bunch of content purely based off their website,

868
00:59:36,360 --> 00:59:38,200
work they had already done.

869
00:59:38,200 --> 00:59:42,360
Then they had already put the paper and turn that into several social media campaigns.

870
00:59:42,360 --> 00:59:45,960
So they can now become omnichannel with their growth strategy.

871
00:59:45,960 --> 00:59:50,640
And then I think for the more later stage companies where they have a new product launch,

872
00:59:50,640 --> 00:59:54,480
we see it being more that assistant of telling that message across a number of different

873
00:59:54,480 --> 01:00:02,000
channels, helping you with your SEO, helping you really just 10x, 100x the content that

874
01:00:02,000 --> 01:00:05,940
you might have been putting out there with the same team using GoCharlie.

875
01:00:05,940 --> 01:00:11,520
Our primary focus has mostly been on SEO, as well as social media content creation.

876
01:00:11,520 --> 01:00:15,680
But I think we have a very interesting suite called content repurposing, which is a bit

877
01:00:15,680 --> 01:00:16,680
of a mouthful.

878
01:00:16,680 --> 01:00:21,600
But my thinking there is like, it allows you to create with anything in source material,

879
01:00:21,600 --> 01:00:26,560
allows you to use videos, talks that you've given, blogs, maybe even co-write with people

880
01:00:26,560 --> 01:00:30,240
that you respect their own thought leadership on.

881
01:00:30,240 --> 01:00:31,640
And so we've given you those tools.

882
01:00:31,640 --> 01:00:35,040
So it no longer becomes just a, hey, we're helping you market.

883
01:00:35,040 --> 01:00:37,800
It also helps you establish authority.

884
01:00:37,800 --> 01:00:41,200
And so for those businesses just starting out, I think that's the way we think about

885
01:00:41,200 --> 01:00:42,200
it.

886
01:00:42,200 --> 01:00:50,400
Content repurposer or whatever the product is called.

887
01:00:50,400 --> 01:00:57,000
Every time I've shown that to people and I've shown that to dozens of businesses now on

888
01:00:57,000 --> 01:01:01,480
various workshops, specifically the video.

889
01:01:01,480 --> 01:01:08,200
So they're just rolling in a YouTube link and going, write me a blog.

890
01:01:08,200 --> 01:01:12,720
And I'm watching people's response because they are so skeptical and the quality of it

891
01:01:12,720 --> 01:01:13,720
is good.

892
01:01:13,720 --> 01:01:17,880
As with any AI, I would always encourage people to make edits, make it your own.

893
01:01:17,880 --> 01:01:25,000
But for a first draft, what it gives you is like, wow, that was pretty impressive.

894
01:01:25,000 --> 01:01:32,240
I think that product itself is testament to when you put some of these models together,

895
01:01:32,240 --> 01:01:37,440
you start to get some really insane workflow productivity benefits.

896
01:01:37,440 --> 01:01:41,680
Like content repurposing, even just the workflow you were talking about, going from a YouTube

897
01:01:41,680 --> 01:01:42,680
video to a blog.

898
01:01:42,680 --> 01:01:46,720
In the past, you would have to use a transcription service to transcribe it.

899
01:01:46,720 --> 01:01:50,640
Then you would have to go review that transcription and make sure that it was accurate.

900
01:01:50,640 --> 01:01:55,120
Then you would have to take that and summarize it into the main points.

901
01:01:55,120 --> 01:01:58,760
Then you would have to take those main points and determine what content you wanted to make

902
01:01:58,760 --> 01:02:00,200
from that.

903
01:02:00,200 --> 01:02:01,920
We're just doing that all in one step.

904
01:02:01,920 --> 01:02:05,800
And then even in a couple other clicks, you can go from having a blog to having a blog

905
01:02:05,800 --> 01:02:09,520
and social media campaign, having some leadership posts for social.

906
01:02:09,520 --> 01:02:13,720
So now you're actually extending the life of your content over a much longer period

907
01:02:13,720 --> 01:02:18,800
of time by just backlinking through additional pieces that all draw off the same source material.

908
01:02:18,800 --> 01:02:22,960
So I think models working together is going to be a trend that we start to see a lot more

909
01:02:22,960 --> 01:02:26,720
in the future.

910
01:02:26,720 --> 01:02:31,800
I certainly see that in the way that I'm playing with certain tools as well.

911
01:02:31,800 --> 01:02:38,040
Using Zapier to build workflows, using things like transcription tools.

912
01:02:38,040 --> 01:02:45,280
So the Whisper API and sticking real-time content into there and having it grow into

913
01:02:45,280 --> 01:02:52,280
my Google Drive, a summarization of the meeting that I was just at or what have you.

914
01:02:52,280 --> 01:02:54,120
In fact, I went to a talk the other day.

915
01:02:54,120 --> 01:02:58,280
It was back here and I just got back home from the Mekong conference.

916
01:02:58,280 --> 01:03:00,080
There was a local talk.

917
01:03:00,080 --> 01:03:04,280
There's some project going on in the city that I live in in Derby.

918
01:03:04,280 --> 01:03:09,760
And I recorded that and just took it straight up to Whisper and got it to spit out a nice

919
01:03:09,760 --> 01:03:14,840
little summary of the whole session with the key points and next steps and ideas for how

920
01:03:14,840 --> 01:03:15,840
I could get involved.

921
01:03:15,840 --> 01:03:22,400
And I thought that's just a really neat way of bringing these different models together.

922
01:03:22,400 --> 01:03:28,480
I got to ask since you mentioned, who's your football club?

923
01:03:28,480 --> 01:03:29,920
The mighty Derby County.

924
01:03:29,920 --> 01:03:30,920
Oh, wow.

925
01:03:30,920 --> 01:03:36,360
We're sitting all the way in League One at the moment, having been relegated from the

926
01:03:36,360 --> 01:03:40,360
Premier League in 2008 with the lowest points total in history.

927
01:03:40,360 --> 01:03:43,560
We scored 11 points and won one game all season.

928
01:03:43,560 --> 01:03:47,840
Hey, you know, being memorable is always something worthwhile, right?

929
01:03:47,840 --> 01:03:53,720
I'm personally a Spurs fan, but they seem poised to never return to Champions League

930
01:03:53,720 --> 01:03:54,720
glory.

931
01:03:54,720 --> 01:03:55,720
Yeah.

932
01:03:55,720 --> 01:04:03,000
And you look like you're about to lose Harry Kane as well to the Munich.

933
01:04:03,000 --> 01:04:06,000
I used to slick my hair back like his and people would call me Harry Kane.

934
01:04:06,000 --> 01:04:10,240
I like to think I'm a little more distinguished now, but yeah, that's just me.

935
01:04:10,240 --> 01:04:16,640
If you haven't seen it in the past week or two, he just did an episode on Hot Ones on

936
01:04:16,640 --> 01:04:18,360
the first week.

937
01:04:18,360 --> 01:04:20,360
Oh yeah.

938
01:04:20,360 --> 01:04:23,520
It's worth a watch.

939
01:04:23,520 --> 01:04:28,480
Hopefully my AI ambitions take me somewhere near sports at some point, but we'll see.

940
01:04:28,480 --> 01:04:32,240
Are you a football fan?

941
01:04:32,240 --> 01:04:34,400
No, I'm a basketball fan.

942
01:04:34,400 --> 01:04:37,480
If there's any Hoosiers listening, I'm an Indiana Hoosier.

943
01:04:37,480 --> 01:04:40,720
So basketball is my jam.

944
01:04:40,720 --> 01:04:45,160
Baseball I enjoy going to a game every now and again, but basketball is definitely me.

945
01:04:45,160 --> 01:04:51,360
I always try and catch a major league game when I'm over in the States, but fortunately

946
01:04:51,360 --> 01:04:56,720
light cancellations this year scuppered me from seeing the Guardians, which was a shame.

947
01:04:56,720 --> 01:04:57,720
Oh no.

948
01:04:57,720 --> 01:04:58,720
All right.

949
01:04:58,720 --> 01:05:03,000
We'll have to have you back over when I'm back in San Francisco and we can do a Giants

950
01:05:03,000 --> 01:05:04,000
game.

951
01:05:04,000 --> 01:05:08,440
Yeah, I'd be definitely up for that.

952
01:05:08,440 --> 01:05:09,800
So let's get back on topic.

953
01:05:09,800 --> 01:05:15,160
I'm interested in your building an AI startup.

954
01:05:15,160 --> 01:05:18,680
I'm guessing that it's not all roses all of the time.

955
01:05:18,680 --> 01:05:23,240
Talk me through some of the challenges of this business.

956
01:05:23,240 --> 01:05:29,840
Yeah, I think there's some that are unique to us and there's some that are unique to

957
01:05:29,840 --> 01:05:31,240
the space.

958
01:05:31,240 --> 01:05:36,280
And so I won't spend as much time on the unique to us because maybe they won't be as relevant

959
01:05:36,280 --> 01:05:37,280
for the listeners.

960
01:05:37,280 --> 01:05:42,320
But I think just being first time founders, regardless of our pedigree, I think works

961
01:05:42,320 --> 01:05:48,040
against you in a space that many people expected rightfully that incumbents would start to

962
01:05:48,040 --> 01:05:52,600
move into once the dam sort of broke of getting these products out there.

963
01:05:52,600 --> 01:05:59,680
I think another is just when you're starting out, like usually it takes a couple years

964
01:05:59,680 --> 01:06:02,640
before you're kind of like, alright, this is the idea.

965
01:06:02,640 --> 01:06:07,180
This is the identity we're going to be and we kind of shot out of the gate with one idea

966
01:06:07,180 --> 01:06:08,820
of what we were going to be.

967
01:06:08,820 --> 01:06:14,640
But then as the space just took off, you might find that that initial focal area is not the

968
01:06:14,640 --> 01:06:17,040
area you want to spend the long term time in.

969
01:06:17,040 --> 01:06:22,320
So that was that's definitely been a difficult spot for us to just how do you steer the

970
01:06:22,320 --> 01:06:27,240
ship and have some sort of clear idea of where you're going as a space is constantly evolving

971
01:06:27,240 --> 01:06:29,800
and more players are coming in.

972
01:06:29,800 --> 01:06:33,900
Now being in the AI space itself is it's fascinating.

973
01:06:33,900 --> 01:06:34,900
It's frustrating.

974
01:06:34,900 --> 01:06:37,180
It's definitely not all roses, as you mentioned.

975
01:06:37,180 --> 01:06:42,240
You have big players releasing models and spending money in a way that you will never

976
01:06:42,240 --> 01:06:43,620
be able to as a startup.

977
01:06:43,620 --> 01:06:48,360
You have startups that had a couple years head start and a much better go to market

978
01:06:48,360 --> 01:06:53,100
motion like I respect a lot of Jasper for what they were able to accomplish so quickly

979
01:06:53,100 --> 01:06:56,120
before many weren't even knowledgeable about the space.

980
01:06:56,120 --> 01:06:57,980
And I think they will continue to be a player.

981
01:06:57,980 --> 01:07:02,660
But much like us, we're still figuring out like how do you play alongside these large

982
01:07:02,660 --> 01:07:06,280
incumbents and get them champion to your cause.

983
01:07:06,280 --> 01:07:10,480
I think the other thing is just like we're a remote based team.

984
01:07:10,480 --> 01:07:11,480
That's always fun.

985
01:07:11,480 --> 01:07:15,700
And then I think you're kind of going through this pandemic transition into more people

986
01:07:15,700 --> 01:07:17,760
wanting to be back in person.

987
01:07:17,760 --> 01:07:19,720
And so a lot more business is being done in person.

988
01:07:19,720 --> 01:07:24,940
And how do you change your startups capital spend when you need to incorporate travel

989
01:07:24,940 --> 01:07:25,940
a little bit more?

990
01:07:25,940 --> 01:07:28,500
That comes off the books versus that.

991
01:07:28,500 --> 01:07:33,160
But the AI space is it's it's troublesome.

992
01:07:33,160 --> 01:07:38,080
I think it's got a lot of the same risk though that you might have with being a retailer

993
01:07:38,080 --> 01:07:39,620
that's selling on Amazon, right?

994
01:07:39,620 --> 01:07:46,120
Like a retailer selling on Amazon always runs the inherent risk that you might get copied

995
01:07:46,120 --> 01:07:50,120
and become a commodity good for Amazon and they might start selling your product at a

996
01:07:50,120 --> 01:07:51,520
lower price.

997
01:07:51,520 --> 01:07:55,380
Same way with AWS's cloud, same way with Google's cloud, same way with Microsoft's cloud.

998
01:07:55,380 --> 01:07:58,000
They all have their own models that they're hawking.

999
01:07:58,000 --> 01:08:04,000
So like how do you get them bought into your cause so that you can build on top of these

1000
01:08:04,000 --> 01:08:06,520
behemoths worth of infrastructure?

1001
01:08:06,520 --> 01:08:09,960
I actually think that's largely why inflection and open AI are starting to build their own

1002
01:08:09,960 --> 01:08:14,200
infrastructure because they realize in the power of these cloud providers and they don't

1003
01:08:14,200 --> 01:08:15,200
want to be a part of that.

1004
01:08:15,200 --> 01:08:18,320
It remains to be seen how that will work out long term.

1005
01:08:18,320 --> 01:08:24,600
But I think the other piece is just like there's so much noise.

1006
01:08:24,600 --> 01:08:29,200
There's so many celebrity founders that are getting funded by their larger VC friends

1007
01:08:29,200 --> 01:08:32,880
that they've cultivated over years, cultivated this image over a year.

1008
01:08:32,880 --> 01:08:34,520
We're a relatively new kid on the block.

1009
01:08:34,520 --> 01:08:39,880
And so in that regard, you have to like make a lot of noise in a new way.

1010
01:08:39,880 --> 01:08:41,380
And there's a lot more expectations.

1011
01:08:41,380 --> 01:08:43,480
You have to like pull off the 100X.

1012
01:08:43,480 --> 01:08:44,960
You can't get away with the 10X.

1013
01:08:44,960 --> 01:08:49,440
Like the celebrity founder doing 10X is going to be able to pull off a huge round.

1014
01:08:49,440 --> 01:08:54,440
The relatively unknown founder is going to have to pull off 100X to really make noise

1015
01:08:54,440 --> 01:08:56,360
and get the funding.

1016
01:08:56,360 --> 01:09:02,860
So I think it's just like keeping yourself mentally stable in an area of so much noise

1017
01:09:02,860 --> 01:09:06,840
and just keeping that determination going.

1018
01:09:06,840 --> 01:09:11,280
I know that's a lot more psychologically driven than like how hard it is to play in the AI

1019
01:09:11,280 --> 01:09:12,280
space.

1020
01:09:12,280 --> 01:09:18,560
But I think that's for me been more part of it than just like being in the AI space.

1021
01:09:18,560 --> 01:09:26,280
I think there is something to start up life which requires that resilience of mind.

1022
01:09:26,280 --> 01:09:30,200
And you know, that is, it's not a space I want to play in.

1023
01:09:30,200 --> 01:09:35,880
So, you know, hats off to all of you that are living in it day to day.

1024
01:09:35,880 --> 01:09:36,880
Yeah.

1025
01:09:36,880 --> 01:09:42,880
So it's all fun and games until you get a nasty gram from a customer at 3am and you're

1026
01:09:42,880 --> 01:09:46,340
like, I just want to respond really meanly to this.

1027
01:09:46,340 --> 01:09:48,640
Like, that's just not business.

1028
01:09:48,640 --> 01:09:51,840
You have to take it as like they cared enough to send this to me.

1029
01:09:51,840 --> 01:09:52,840
You know, I don't like what it says.

1030
01:09:52,840 --> 01:09:55,280
They cared enough to send it to me.

1031
01:09:55,280 --> 01:09:56,280
Yeah.

1032
01:09:56,280 --> 01:09:59,560
Customers, hey, who'd have them?

1033
01:09:59,560 --> 01:10:04,880
Looking for that immortal revenue model with zero humans, but still revenue.

1034
01:10:04,880 --> 01:10:07,720
Oh no, I'm a people person.

1035
01:10:07,720 --> 01:10:11,400
I'm a people person, Brennan, honestly.

1036
01:10:11,400 --> 01:10:14,680
So you spoke there about customers.

1037
01:10:14,680 --> 01:10:22,200
Are there any particular standouts where you think, or studies that you can share of people

1038
01:10:22,200 --> 01:10:26,920
that have used the product and come to you with some really amazing eye opening use cases

1039
01:10:26,920 --> 01:10:30,520
and they've killed it?

1040
01:10:30,520 --> 01:10:33,520
Yeah.

1041
01:10:33,520 --> 01:10:35,180
There's a bunch of these.

1042
01:10:35,180 --> 01:10:41,960
There's one moment that I always distinctly remember that like really, I don't know that

1043
01:10:41,960 --> 01:10:43,560
I could say that it's cemented.

1044
01:10:43,560 --> 01:10:48,640
My decision is being right to jump from SoftBank into a pre-revenue startup, but it certainly

1045
01:10:48,640 --> 01:10:52,400
made me feel like it was the right decision.

1046
01:10:52,400 --> 01:10:57,200
And that was the first time that we released the first iteration of content repurposing.

1047
01:10:57,200 --> 01:11:01,640
We were the first ones that have like YouTube to anything.

1048
01:11:01,640 --> 01:11:07,600
And I distinctly remember there was a customer in our chat or in Facebook community and they

1049
01:11:07,600 --> 01:11:12,880
simply responded, I was paying an agency $1,400 a month to do this.

1050
01:11:12,880 --> 01:11:17,760
And you guys are now doing this for a simple monthly price and I can do this for all the

1051
01:11:17,760 --> 01:11:19,080
content.

1052
01:11:19,080 --> 01:11:24,520
And right there, I was like, one, that's the firepower of AI, but two, like they can now

1053
01:11:24,520 --> 01:11:29,040
invest that $1,400 a month into their business, into making their dreams come true.

1054
01:11:29,040 --> 01:11:33,240
Like that for me was like, this is the thing.

1055
01:11:33,240 --> 01:11:36,360
This is a bigger dopamine rush than anything else.

1056
01:11:36,360 --> 01:11:41,960
I think I saw a lot of people, so we even still have this.

1057
01:11:41,960 --> 01:11:46,060
We have like a chat-based experience, but we also have a template-based experience.

1058
01:11:46,060 --> 01:11:49,080
And then we have an image generation experience.

1059
01:11:49,080 --> 01:11:52,740
And so I started to see people doing some really interesting stuff where they would

1060
01:11:52,740 --> 01:11:58,520
prompt the chat to describe a visual for an idea they were working on and then feed that

1061
01:11:58,520 --> 01:12:00,680
into the image generation model.

1062
01:12:00,680 --> 01:12:05,000
And so we started to think, how can we just bring the image creation part right into that

1063
01:12:05,000 --> 01:12:06,240
experience natively?

1064
01:12:06,240 --> 01:12:08,640
So you see that with the next iteration of the product.

1065
01:12:08,640 --> 01:12:13,760
But I think using the models in collaboration is where I started to see that really mind-blowing

1066
01:12:13,760 --> 01:12:14,760
stuff.

1067
01:12:14,760 --> 01:12:15,760
Yeah.

1068
01:12:15,760 --> 01:12:23,040
Again, I think in fact, actually, Pasi Kozakov, is that how you pronounce her name?

1069
01:12:23,040 --> 01:12:28,560
The chief decision scientist over at Google, she was keynoting at the Mekong conference

1070
01:12:28,560 --> 01:12:29,760
last week.

1071
01:12:29,760 --> 01:12:34,160
And her whole talk was that this isn't an AI revolution that we're going through now.

1072
01:12:34,160 --> 01:12:35,880
This is a design revolution.

1073
01:12:35,880 --> 01:12:43,360
This is all about design and designing workflows and designing systems and processes and taking

1074
01:12:43,360 --> 01:12:50,960
things from A to B to C. Effectively, recipes was the kind of analogy that she talked about.

1075
01:12:50,960 --> 01:12:52,680
And I think that's true.

1076
01:12:52,680 --> 01:12:54,680
Jasper on that, recipes.

1077
01:12:54,680 --> 01:13:06,280
Hers was actually more of a kitchen-based full work analogy.

1078
01:13:06,280 --> 01:13:12,560
But no, I didn't speak to the Jasper team when I was there, actually, because they were

1079
01:13:12,560 --> 01:13:13,560
exhibiting.

1080
01:13:13,560 --> 01:13:16,880
But yeah, I think workflows is the thing, isn't it?

1081
01:13:16,880 --> 01:13:19,680
How do you, that kind of multimodality.

1082
01:13:19,680 --> 01:13:21,760
But also, I want to get, I've got this idea.

1083
01:13:21,760 --> 01:13:23,520
I want to turn that into a thing.

1084
01:13:23,520 --> 01:13:24,520
Give me a blog outline.

1085
01:13:24,520 --> 01:13:30,720
Give me the blog outline into a blog with the social media post.

1086
01:13:30,720 --> 01:13:32,800
That kind of one-click campaign.

1087
01:13:32,800 --> 01:13:35,680
When people see that idea, that's great.

1088
01:13:35,680 --> 01:13:39,000
You've spoken about the content repurposing.

1089
01:13:39,000 --> 01:13:44,240
The steps involved in that are not as simple as watch YouTube video, make notes.

1090
01:13:44,240 --> 01:13:47,240
If you're doing it manually, that is not the workflow.

1091
01:13:47,240 --> 01:13:51,560
There are lots of steps to go through to get a blog post.

1092
01:13:51,560 --> 01:13:58,160
What you've done is you've managed to use AI and build a sequence of steps that does

1093
01:13:58,160 --> 01:14:01,080
it all for you in one go.

1094
01:14:01,080 --> 01:14:02,080
Yeah.

1095
01:14:02,080 --> 01:14:09,840
And I think this is interesting and I'm curious to explore this with you as well.

1096
01:14:09,840 --> 01:14:13,240
We talk about redesigning workflows.

1097
01:14:13,240 --> 01:14:18,080
Funny enough, I still get people asking for the intermediate steps.

1098
01:14:18,080 --> 01:14:25,840
I have still never been able to get someone to give me a solid intellectual reason for

1099
01:14:25,840 --> 01:14:30,040
what they use raw transcripts for.

1100
01:14:30,040 --> 01:14:34,640
Without fail, every time someone mentions, yeah, I can create transcripts with this.

1101
01:14:34,640 --> 01:14:35,640
I'm like, cool.

1102
01:14:35,640 --> 01:14:38,040
What do you do with those transcripts?

1103
01:14:38,040 --> 01:14:42,080
And then they tell me, oh, we create summaries or we create a blog post or we create social

1104
01:14:42,080 --> 01:14:43,080
media.

1105
01:14:43,080 --> 01:14:47,960
And I'm like, so then what would be the point of having a transcript versus just those outputs?

1106
01:14:47,960 --> 01:14:49,960
Like why would you want it?

1107
01:14:49,960 --> 01:14:51,880
It boggles people mind.

1108
01:14:51,880 --> 01:14:52,880
They can't come up with it.

1109
01:14:52,880 --> 01:14:57,280
But then like you extrapolate that idea out with this designing workflows situation we're

1110
01:14:57,280 --> 01:14:59,240
in.

1111
01:14:59,240 --> 01:15:03,000
What level of autonomy do you think people are willing to give to AI?

1112
01:15:03,000 --> 01:15:06,920
Do we have to get to this like versioning setup with AI where you could see all those

1113
01:15:06,920 --> 01:15:11,400
intermediate steps that are happening almost like you would version changes in a Google

1114
01:15:11,400 --> 01:15:16,720
Doc for people to allow AI to have any sort of autonomy, much less full autonomy.

1115
01:15:16,720 --> 01:15:21,880
And I think that's going to be the larger question we have to answer as we exit this

1116
01:15:21,880 --> 01:15:27,840
design phase is like, what are we willing to give up in this workflow?

1117
01:15:27,840 --> 01:15:28,840
Yeah.

1118
01:15:28,840 --> 01:15:37,720
And that reminds me of the Wired magazine after ChatGPT came out and made all the headlines.

1119
01:15:37,720 --> 01:15:39,320
They wrote an article.

1120
01:15:39,320 --> 01:15:43,840
It was an editorial piece saying how they as a publisher are going to use generative

1121
01:15:43,840 --> 01:15:47,720
AI or how they will and won't use generative AI.

1122
01:15:47,720 --> 01:15:52,960
And one of the things that they said in there was that for editing, they will not use generative

1123
01:15:52,960 --> 01:15:57,080
AI because editing requires judgment.

1124
01:15:57,080 --> 01:16:04,040
It requires human judgment to determine what is important and what is the, what requires

1125
01:16:04,040 --> 01:16:06,520
the kind of weight and emphasis within a story.

1126
01:16:06,520 --> 01:16:13,200
And that's not something that you can hand over to a bot or to an AI.

1127
01:16:13,200 --> 01:16:14,200
And I think that's right.

1128
01:16:14,200 --> 01:16:17,800
You know, when we, again, it's, I think it's one of the things going back to your point

1129
01:16:17,800 --> 01:16:19,240
about autonomous vehicles.

1130
01:16:19,240 --> 01:16:22,480
Ultimately, that is the big question about autonomous vehicles, isn't it?

1131
01:16:22,480 --> 01:16:26,960
Like that's how much are you prepared to relinquish?

1132
01:16:26,960 --> 01:16:35,760
The moment a bot driven car kills a person, even though it might do it, you know, a hundred

1133
01:16:35,760 --> 01:16:41,680
times less frequent than people on mass, people go, well, it made a decision and it ended

1134
01:16:41,680 --> 01:16:43,720
up killing a person, so we can't trust it.

1135
01:16:43,720 --> 01:16:46,040
Even though it might fundamentally be safer.

1136
01:16:46,040 --> 01:16:50,280
I was at, wittily my first week at Uber was when the autonomous vehicle hit that woman

1137
01:16:50,280 --> 01:16:52,200
and killed her.

1138
01:16:52,200 --> 01:16:57,000
And the ripples that sent through the company, as well as our long-term investment decisions

1139
01:16:57,000 --> 01:17:01,360
in autonomous vehicles, like was rather massive.

1140
01:17:01,360 --> 01:17:04,520
And I tried to think about this as well because I'm like people die in car accidents all the

1141
01:17:04,520 --> 01:17:08,800
time, not to get too morbid, but like every day it happens.

1142
01:17:08,800 --> 01:17:13,800
And the only reason I could get to is like, it's not news if someone dies in a normal

1143
01:17:13,800 --> 01:17:14,800
car accident.

1144
01:17:14,800 --> 01:17:19,360
It is news if someone dies in an autonomous vehicle car accident and we can blame big

1145
01:17:19,360 --> 01:17:20,720
tech for it.

1146
01:17:20,720 --> 01:17:25,660
And I worry that we're getting into that same cycle with a lot of these AI systems where

1147
01:17:25,660 --> 01:17:31,100
we may end up limiting the utopia that could be delivered by some of these systems purely

1148
01:17:31,100 --> 01:17:35,800
because we're just going to bury them with bad news stories.

1149
01:17:35,800 --> 01:17:39,840
Yeah, and this is where we need that balance, don't we?

1150
01:17:39,840 --> 01:17:45,800
You see actually in the news cycle recently, in fact, I've argued recently that for the

1151
01:17:45,800 --> 01:17:55,120
likes of Sam Altman talking about the capabilities of AGI and how it could be a kind of destroyer

1152
01:17:55,120 --> 01:17:58,680
of jobs and it's okay, we've got a kill switch.

1153
01:17:58,680 --> 01:18:00,660
We can turn it off at the wall if we need to.

1154
01:18:00,660 --> 01:18:01,660
We can definitely do that.

1155
01:18:01,660 --> 01:18:05,460
I think that's a really interesting piece of marketing hype.

1156
01:18:05,460 --> 01:18:09,720
People want to buy your product if you're telling people that it's this close to becoming

1157
01:18:09,720 --> 01:18:15,120
like, it's going to take over the world and we're having to really rein it in and restrain

1158
01:18:15,120 --> 01:18:16,120
it.

1159
01:18:16,120 --> 01:18:19,960
Whereas you listen to other people in the field that are at the leading edge and they're

1160
01:18:19,960 --> 01:18:25,720
just like, no, that's not it.

1161
01:18:25,720 --> 01:18:26,720
Yeah.

1162
01:18:26,720 --> 01:18:27,720
Yeah.

1163
01:18:27,720 --> 01:18:29,480
And then like what even is the definition of AGI?

1164
01:18:29,480 --> 01:18:34,040
I mean, we have a heavy discussion of marketing already.

1165
01:18:34,040 --> 01:18:38,920
Like I kind of think of AGI these days is like brand voice is for marketing.

1166
01:18:38,920 --> 01:18:44,880
Like every expert seems to have a different idea of what that is and what it constitutes.

1167
01:18:44,880 --> 01:18:50,560
And anytime that you can't agree on what the definition of something is, then typically

1168
01:18:50,560 --> 01:18:53,560
you're going to have people saying that they're already there.

1169
01:18:53,560 --> 01:19:00,120
So I think we should, my hope is we adopt a uniform way of thinking about to the valuation.

1170
01:19:00,120 --> 01:19:05,360
Yeah, let's get to consensus on that one point alone, just that everybody recognizes it when

1171
01:19:05,360 --> 01:19:06,360
they see it.

1172
01:19:06,360 --> 01:19:07,360
You're right.

1173
01:19:07,360 --> 01:19:11,400
So this has been quite a broad ranging discussion.

1174
01:19:11,400 --> 01:19:15,920
We should probably start wrapping it up soon, but I'm interested to know, you've talked

1175
01:19:15,920 --> 01:19:24,200
about AI agents and that being something that you're developing internally.

1176
01:19:24,200 --> 01:19:34,360
That excites me, when we can start to build pools or AIs that can use tools.

1177
01:19:34,360 --> 01:19:37,160
We see glimmers of this with the likes of Code Interpreter.

1178
01:19:37,160 --> 01:19:43,600
We see glimmers of this with plugins and various other auto GPT.

1179
01:19:43,600 --> 01:19:46,160
Although I think that's probably wildly hyped.

1180
01:19:46,160 --> 01:19:50,000
I don't think anybody's actually done anything useful with auto GPT from what I can see so

1181
01:19:50,000 --> 01:19:51,000
far.

1182
01:19:51,000 --> 01:19:56,640
But going forward when this can happen, what does the future look like?

1183
01:19:56,640 --> 01:20:04,280
And on from that, what do you hope the future looks like for the customers of GoCharlie

1184
01:20:04,280 --> 01:20:10,480
in say, let's say 12 months time when this landscape will be very different?

1185
01:20:10,480 --> 01:20:14,120
It's a fascinating thing to think about.

1186
01:20:14,120 --> 01:20:20,400
I'm in this AI agents DM group on Twitter that has been eye opening.

1187
01:20:20,400 --> 01:20:25,000
And one of our board members actually, his name is Dustin Dannenhauer.

1188
01:20:25,000 --> 01:20:30,600
He was one of the foremost thinkers on AI agents before this whole like LLM based agent

1189
01:20:30,600 --> 01:20:34,000
thing became a thing.

1190
01:20:34,000 --> 01:20:39,760
I think agents is such an interesting topic because for those that don't know, agents

1191
01:20:39,760 --> 01:20:44,080
is basically like you can execute a more complex task for you.

1192
01:20:44,080 --> 01:20:48,340
But at its simplest, it's in an environment and is able to execute a task.

1193
01:20:48,340 --> 01:20:52,120
So you might be able to think about a large language model and some of the things that

1194
01:20:52,120 --> 01:20:56,640
does is as an agent, a very simple agent, but an agent nonetheless, it was an environment

1195
01:20:56,640 --> 01:20:58,360
it does task for you.

1196
01:20:58,360 --> 01:21:05,560
Now if you think about agents is sort of we're thinking about what they constitute.

1197
01:21:05,560 --> 01:21:09,680
We would theoretically want an agent where you say, hey, I would like you to research

1198
01:21:09,680 --> 01:21:16,000
this topic, establish a stance on it, make a blog post about it, make a series of tweets

1199
01:21:16,000 --> 01:21:20,880
that kind of carry that messaging through, add some images and cite your sources.

1200
01:21:20,880 --> 01:21:22,440
So that's like very complex.

1201
01:21:22,440 --> 01:21:26,040
That would normally take one person maybe a week before AI.

1202
01:21:26,040 --> 01:21:28,720
Now it might take a day or two with AI.

1203
01:21:28,720 --> 01:21:31,680
We want it to do that whole thing all at once.

1204
01:21:31,680 --> 01:21:35,800
So that's what we're kind of envisioning our agent product is doing is like at your behest

1205
01:21:35,800 --> 01:21:37,040
doing thing.

1206
01:21:37,040 --> 01:21:41,060
But then where it gets super interesting is over time it warns you.

1207
01:21:41,060 --> 01:21:43,200
And so then the agent starts to become proactive.

1208
01:21:43,200 --> 01:21:47,440
So then instead of saying, or instead of you coming into the product and prompting, it's

1209
01:21:47,440 --> 01:21:51,160
coming to you and saying, hey, based upon your business, the competitive landscape,

1210
01:21:51,160 --> 01:21:55,580
the industry that you're in, here's what we recommend you create content about this

1211
01:21:55,580 --> 01:21:57,680
week and we've already created some graphs.

1212
01:21:57,680 --> 01:22:01,740
You want us to post this for you or would you prefer to come in and edit more?

1213
01:22:01,740 --> 01:22:08,400
So to me, agents are the bedrock of a more proactive AI system, more proactive assistant,

1214
01:22:08,400 --> 01:22:12,000
the more real collaborator with your business.

1215
01:22:12,000 --> 01:22:15,280
Now if I think about like, where does that go in the future?

1216
01:22:15,280 --> 01:22:18,840
First and foremost, I think that agents need to turn doorknobs.

1217
01:22:18,840 --> 01:22:23,560
They don't need to be talking to in special languages to one another.

1218
01:22:23,560 --> 01:22:27,800
I think that will probably come most likely, but I think the most successful agents will

1219
01:22:27,800 --> 01:22:34,040
learn how to use existing tools, APIs, because that will garner mass adoption.

1220
01:22:34,040 --> 01:22:38,400
If I could tell something from Go Charlie, create an email campaign and post it to my

1221
01:22:38,400 --> 01:22:42,960
Clavio account, I'm going to use that tool versus one that has to have some special sort

1222
01:22:42,960 --> 01:22:44,960
of coding to go into it.

1223
01:22:44,960 --> 01:22:51,040
So you need an agent to be able to turn doorknobs, not create code to make something happen.

1224
01:22:51,040 --> 01:22:55,400
Longer term, I think the agents just become part of your company.

1225
01:22:55,400 --> 01:23:01,000
I do think that because of AI, companies that are just starting will be smaller.

1226
01:23:01,000 --> 01:23:06,080
I think the companies that are larger will get smaller or will do more.

1227
01:23:06,080 --> 01:23:10,320
But I think also because of AI, those companies, when they make layoffs, you're going to have

1228
01:23:10,320 --> 01:23:13,920
a lot more companies started because you won't need as much.

1229
01:23:13,920 --> 01:23:18,400
My hope for the future is that I could have an AI agent that basically runs a passive

1230
01:23:18,400 --> 01:23:20,400
income side hustle for me.

1231
01:23:20,400 --> 01:23:23,180
So I can go wig on a beach in Mexico and paint at night.

1232
01:23:23,180 --> 01:23:25,720
But like, I think we're a ways off from there.

1233
01:23:25,720 --> 01:23:28,820
I think there's some regulatory concerns with that.

1234
01:23:28,820 --> 01:23:34,920
But ultimately, I kind of view AI as like, at its best, it's the digital twin of your

1235
01:23:34,920 --> 01:23:37,080
desires.

1236
01:23:37,080 --> 01:23:40,960
You tell it what you want to be done and it can go do it for you.

1237
01:23:40,960 --> 01:23:45,400
That's kind of the future world that I think we're moving towards.

1238
01:23:45,400 --> 01:23:49,040
But it all boils down to doing more with less.

1239
01:23:49,040 --> 01:23:55,360
So fast forward 18 months, go Charlie agent is flying, you're sat on a beach living off

1240
01:23:55,360 --> 01:24:01,560
the income from drop shipping dot agent AI.

1241
01:24:01,560 --> 01:24:03,040
That is that what I can expect to see.

1242
01:24:03,040 --> 01:24:08,480
I mean, it's funny that that's is a brilliant picture of agents.

1243
01:24:08,480 --> 01:24:14,160
And I think that what you've touched on is that the bit that's easy to overlook with

1244
01:24:14,160 --> 01:24:19,840
agents, because like I mentioned them, AI is being able to use calls and think about

1245
01:24:19,840 --> 01:24:24,160
auto GPT being able to plug into an API and do something.

1246
01:24:24,160 --> 01:24:30,000
But it's the proactive piece that I think is so often overlooked.

1247
01:24:30,000 --> 01:24:33,440
The idea that this is just running in the background, right?

1248
01:24:33,440 --> 01:24:38,360
The thing that lots of people are probably thinking when they use the likes of chat GPT

1249
01:24:38,360 --> 01:24:41,400
for any amount of time, you know, you use it for a month or so.

1250
01:24:41,400 --> 01:24:44,760
And then you're thinking, I wish it could just go and find this out for me now when

1251
01:24:44,760 --> 01:24:50,240
it's not available to you on your, you know, you haven't got your phone to hand or whatever.

1252
01:24:50,240 --> 01:24:54,600
If you've got an AI agent that's always there, always there, just kind of dialed in doing

1253
01:24:54,600 --> 01:24:58,280
little jobs for you, doing research, finding out what the market conditions are like, saying

1254
01:24:58,280 --> 01:25:02,840
here's a blog post, here's a trending topic that you want to jump on the bandwagon of.

1255
01:25:02,840 --> 01:25:07,200
If you've got that, I mean, that is an absolute game changer as if you're a marketing manager,

1256
01:25:07,200 --> 01:25:08,640
marketing assistant, whatever.

1257
01:25:08,640 --> 01:25:11,880
So you painted an exciting picture there Brennan.

1258
01:25:11,880 --> 01:25:12,880
Yeah.

1259
01:25:12,880 --> 01:25:15,460
I mean, we were talking about workflow design.

1260
01:25:15,460 --> 01:25:18,320
What if the best workflow is no workflow at all?

1261
01:25:18,320 --> 01:25:25,000
Oh, that is, if we can plug that in with the customerless revenue generation model, we're

1262
01:25:25,000 --> 01:25:28,000
in the dream ticket.

1263
01:25:28,000 --> 01:25:34,400
Then we'll both go retire and watch Giants baseball games and sit on the beach in Mexico.

1264
01:25:34,400 --> 01:25:35,400
How about that?

1265
01:25:35,400 --> 01:25:36,400
I'm into it.

1266
01:25:36,400 --> 01:25:38,160
Yeah, most definitely.

1267
01:25:38,160 --> 01:25:43,000
So where can people find out more about Go Charlie?

1268
01:25:43,000 --> 01:25:44,000
Yeah.

1269
01:25:44,000 --> 01:25:49,560
So our website, gocharlie.ai, we are actually doing a phenomenal event.

1270
01:25:49,560 --> 01:25:55,200
Not sure when this is going to air, but September 6th, we're throwing the AI party to celebrate

1271
01:25:55,200 --> 01:25:59,840
the close of our $2 million pre-seed round, the launch of our large language model publicly,

1272
01:25:59,840 --> 01:26:01,880
the launch of our agent product.

1273
01:26:01,880 --> 01:26:07,320
We're going to sit down with some people from SRI and NVIDIA, as well as the CEO of Hawk

1274
01:26:07,320 --> 01:26:10,600
Media and some other people on the future of AI marketing as well.

1275
01:26:10,600 --> 01:26:16,160
So definitely check us out there, but you can find us on Instagram, Twitter, LinkedIn,

1276
01:26:16,160 --> 01:26:17,760
gocharlie.ai is the handle.

1277
01:26:17,760 --> 01:26:19,960
We didn't do anything too, too fun with it.

1278
01:26:19,960 --> 01:26:22,820
We make the content fun, not our handles.

1279
01:26:22,820 --> 01:26:25,320
But then me personally, just Brennan Woodruff on LinkedIn.

1280
01:26:25,320 --> 01:26:30,080
I love hearing from customers, love hearing from AI enthusiasts, pretty active in the

1281
01:26:30,080 --> 01:26:31,080
site community.

1282
01:26:31,080 --> 01:26:33,560
So just shoot me a note.

1283
01:26:33,560 --> 01:26:36,760
And in terms of signing up, what are the plans to start from?

1284
01:26:36,760 --> 01:26:39,480
Have you got a free plan, free trial, teams plan?

1285
01:26:39,480 --> 01:26:40,680
What does that look like?

1286
01:26:40,680 --> 01:26:41,680
Yeah.

1287
01:26:41,680 --> 01:26:44,440
So we have three tiers, but we also have a free tier.

1288
01:26:44,440 --> 01:26:46,640
So we just recently made a free tier available.

1289
01:26:46,640 --> 01:26:48,800
We want everybody to benefit from AI.

1290
01:26:48,800 --> 01:26:50,720
It helps us improve the product.

1291
01:26:50,720 --> 01:26:53,360
And so there's some limited usage on a monthly basis.

1292
01:26:53,360 --> 01:26:57,240
And then as your business scales and you need more advanced content throughput, we have

1293
01:26:57,240 --> 01:27:00,260
a freelancer tier, then we have a professional tier.

1294
01:27:00,260 --> 01:27:02,000
Those are both single seat options.

1295
01:27:02,000 --> 01:27:03,700
One is limited use of the platform.

1296
01:27:03,700 --> 01:27:04,700
One is full use.

1297
01:27:04,700 --> 01:27:09,160
And then we have a team's tier for agencies, enterprises and beyond.

1298
01:27:09,160 --> 01:27:12,120
That's a five seat plan.

1299
01:27:12,120 --> 01:27:17,400
So those are evolved over time pricing wise, but yeah, we have flavors for everybody in

1300
01:27:17,400 --> 01:27:19,200
every walk of your business.

1301
01:27:19,200 --> 01:27:23,840
Great, well, as a user of GoCharlie, I would encourage people to go and check it out.

1302
01:27:23,840 --> 01:27:27,080
It was the tool of the week very early on in the podcast.

1303
01:27:27,080 --> 01:27:28,640
So people want to hear my thoughts on it.

1304
01:27:28,640 --> 01:27:30,280
They can go check that out.

1305
01:27:30,280 --> 01:27:33,480
Well, thank you very much for joining us today.

1306
01:27:33,480 --> 01:27:35,760
Of course, Martin, it's always a pleasure, my man.

1307
01:27:35,760 --> 01:27:40,480
We got to do this more often.

1308
01:27:40,480 --> 01:27:43,940
Thank you for listening to Artificially Intelligent Marketing.

1309
01:27:43,940 --> 01:27:49,980
To stay on top of the latest trends, tips and tools in the world of marketing AI, be

1310
01:27:49,980 --> 01:27:51,740
sure to subscribe.

1311
01:27:51,740 --> 01:28:14,800
We look forward to seeing you again next week.