WEBVTT 00:00:00.000 --> 00:00:03.060 Imagine an AI that doesn't just answer your questions. 00:00:03.660 --> 00:00:06.719 Instead, it proactively browses the entire web, 00:00:07.160 --> 00:00:10.300 digs into complex topics, meticulously analyzes 00:00:10.300 --> 00:00:14.119 data, and then creates completely original content, 00:00:14.539 --> 00:00:16.460 all on its own. Yeah, it's really like having 00:00:16.460 --> 00:00:18.620 a virtual assistant that actually thinks. It 00:00:18.620 --> 00:00:21.039 executes these complex multi -step tasks, freeing 00:00:21.039 --> 00:00:24.519 you up to just observe. or maybe grab another 00:00:24.519 --> 00:00:26.519 coffee. That's a tempting thought. OK, let's 00:00:26.519 --> 00:00:29.519 unpack this a bit. Welcome to the deep dive where 00:00:29.519 --> 00:00:31.179 we try to cut through the noise and bring you 00:00:31.179 --> 00:00:33.439 the insights that really matter. Today, we're 00:00:33.439 --> 00:00:37.280 taking a really deep look at chat GPT's agent 00:00:37.280 --> 00:00:39.429 mode. Right, and here's where it gets super interesting 00:00:39.429 --> 00:00:41.670 for anyone using AI today. This isn't just your 00:00:41.670 --> 00:00:43.710 average chatbot, not even close. We're talking 00:00:43.710 --> 00:00:46.310 about an AI that can handle really intricate 00:00:46.310 --> 00:00:49.590 multi -step workflows with a surprising amount 00:00:49.590 --> 00:00:52.229 of autonomy. We'll explore what agent mode actually 00:00:52.229 --> 00:00:54.049 is, kind of how it works under the hood, and 00:00:54.049 --> 00:00:57.130 crucially, four pretty surprising and actionable 00:00:57.130 --> 00:01:00.469 ways you can use it right now. And it's not all 00:01:00.469 --> 00:01:02.890 perfect. We'll also look at its current limitations, 00:01:03.090 --> 00:01:06.069 the reality of, say, website access, and those 00:01:06.069 --> 00:01:07.950 really key security considerations you need to 00:01:07.950 --> 00:01:10.129 be aware of. This deep dive is essentially about 00:01:10.129 --> 00:01:12.890 giving you a clear roadmap, a way to really leverage 00:01:12.890 --> 00:01:16.650 this new phase of autonomous AI. So let's start 00:01:16.650 --> 00:01:18.829 with the basics. What exactly is agent mode? 00:01:19.030 --> 00:01:20.810 And how is it fundamentally different from the 00:01:20.810 --> 00:01:24.040 chat GPT we might be used to? Okay, so at its 00:01:24.040 --> 00:01:26.519 core, it's Chat GPT acting as what you might 00:01:26.519 --> 00:01:30.000 call a semi -autonomous AI entity. Think of an 00:01:30.000 --> 00:01:33.200 AI entity as basically an AI that can perform 00:01:33.200 --> 00:01:35.000 tasks independently once you give it a clear 00:01:35.000 --> 00:01:37.340 goal. What makes Agent Mode special is how it 00:01:37.340 --> 00:01:39.500 coordinates things. It uses various tools together 00:01:39.500 --> 00:01:41.680 like web browsing to search the internet, its 00:01:41.680 --> 00:01:43.640 own data analysis tools, you know, what used 00:01:43.640 --> 00:01:45.379 to be code interpreter for crunching numbers, 00:01:45.859 --> 00:01:48.599 that image generation with DLE3, and even connections 00:01:48.599 --> 00:01:51.430 to external apps you use. all working together 00:01:51.430 --> 00:01:54.209 to achieve a bigger, often pretty complex objective. 00:01:54.409 --> 00:01:56.030 That's a really key distinction then. The core 00:01:56.030 --> 00:01:59.430 difference is the autonomy. Before, we were almost 00:01:59.430 --> 00:02:01.549 constantly guiding it turn by turn. With Agent 00:02:01.549 --> 00:02:04.469 Mode, you set the overall task like a single 00:02:04.469 --> 00:02:07.530 high -level goal and the AI takes over. It autonomously 00:02:07.530 --> 00:02:10.400 plans. breaks down the work, decides which tool 00:02:10.400 --> 00:02:12.659 to use when, and just keeps executing. It might 00:02:12.659 --> 00:02:15.219 work for 10, 20 minutes, maybe longer, without 00:02:15.219 --> 00:02:17.460 you needing to constantly jump in. That's a big 00:02:17.460 --> 00:02:19.659 shift. It really is. Think about standard chat 00:02:19.659 --> 00:02:22.659 GPT. It's turn -based Q &A. You ask. It answers. 00:02:22.800 --> 00:02:24.680 It waits. Browse mode was better. It could summarize 00:02:24.680 --> 00:02:27.580 a few search results. But agent mode, it's a 00:02:27.580 --> 00:02:29.800 whole different beast. It's self -plans. It can 00:02:29.800 --> 00:02:31.919 browse dozens of websites if needed, analyze 00:02:31.919 --> 00:02:34.280 what it finds, synthesize it all, and then deliver 00:02:34.280 --> 00:02:37.800 a complete result. It's a huge leap in let's 00:02:37.800 --> 00:02:41.879 say, say, cognitive complexity for AI. And to 00:02:41.879 --> 00:02:44.259 get started, you will need a chat GPT Plus account. 00:02:44.539 --> 00:02:46.800 Users usually get a limited number of agent sessions 00:02:46.800 --> 00:02:48.580 per month. You turn it on using the paperclip 00:02:48.580 --> 00:02:51.259 or plus icon you already see in the chat. And 00:02:51.259 --> 00:02:52.860 you're always in control. You can stop at any 00:02:52.860 --> 00:02:54.460 time or even take over its browser if you need 00:02:54.460 --> 00:02:56.840 to nudge it. OK, we've covered what it is, how 00:02:56.840 --> 00:02:59.599 it works. But what does this shift really mean 00:02:59.599 --> 00:03:02.699 for us? How does Agent Mode change our role when 00:03:02.699 --> 00:03:04.879 we're working with AI? You set the goal, the 00:03:04.879 --> 00:03:07.300 AI plans, and executes autonomously. You become 00:03:07.300 --> 00:03:09.900 the strategist. Right, shifting roles. OK, let's 00:03:09.900 --> 00:03:12.759 talk use cases. For complex creators, finding 00:03:12.759 --> 00:03:16.300 fresh, trending ideas is, well, it's a constant 00:03:16.300 --> 00:03:19.020 struggle, isn't it? Huge time sink. Agent Mode 00:03:19.020 --> 00:03:20.939 sounds like it could become your tireless trend 00:03:20.939 --> 00:03:22.759 researcher. Absolutely. That's one of its real 00:03:22.759 --> 00:03:24.879 sweet spots. It's like having a dedicated research 00:03:24.879 --> 00:03:27.990 team ready to go. But here's the key. The quality 00:03:27.990 --> 00:03:30.370 you get out is directly tied to the quality of 00:03:30.370 --> 00:03:33.030 your prompt. You can't just say, find me content 00:03:33.030 --> 00:03:35.150 ideas. That's too generic. You need to be really 00:03:35.150 --> 00:03:38.949 specific. Give the AI a clear role, like act 00:03:38.949 --> 00:03:41.469 as a trend analyst for my personal finance YouTube 00:03:41.469 --> 00:03:44.710 channel targeting Gen Z in Vietnam. That specificity 00:03:44.710 --> 00:03:47.189 is crucial. So it's more like writing a detailed 00:03:47.189 --> 00:03:49.930 brief than just asking a question. You define 00:03:49.930 --> 00:03:52.650 the context, the exact task, maybe data collection 00:03:52.650 --> 00:03:55.449 across Reddit, YouTube, Google Trends, specific 00:03:55.449 --> 00:03:58.590 blogs, then the analysis, and even how you want 00:03:58.590 --> 00:04:01.169 the recommendations formatted. Like a markdown 00:04:01.169 --> 00:04:04.409 table with video topic, keywords, unique angle. 00:04:04.789 --> 00:04:07.050 Precisely. Think about how you'd brief a human 00:04:07.050 --> 00:04:09.770 analyst. You'd give them all that detail, right? 00:04:09.770 --> 00:04:11.949 The audience, the platforms, the desired output 00:04:11.949 --> 00:04:13.830 structure. The more detailed you are, the better 00:04:13.830 --> 00:04:16.509 the agent understands its mission. Then, once 00:04:16.509 --> 00:04:19.110 you hit go, you actually see it working. It systematically 00:04:19.110 --> 00:04:22.350 visits websites, runs searches, reads articles, 00:04:22.529 --> 00:04:24.649 takes notes. It's kind of fascinating to watch, 00:04:24.649 --> 00:04:26.589 like seeing research happen and fast forward. 00:04:26.790 --> 00:04:28.990 That sounds incredibly powerful. But what about 00:04:28.990 --> 00:04:31.259 roadblocks? What happens when it hits a snag, 00:04:31.319 --> 00:04:33.300 like a website blocking its access? That must 00:04:33.300 --> 00:04:35.319 happen quite a bit. Oh yeah, that's a very real 00:04:35.319 --> 00:04:37.680 thing. Lots of sites have... pretty strong anti 00:04:37.680 --> 00:04:40.560 -bot measures. But the agent is designed to adapt. 00:04:40.699 --> 00:04:43.180 It doesn't just give up. It logs the error, it 00:04:43.180 --> 00:04:45.240 reevaluates its plan, and it tries alternative 00:04:45.240 --> 00:04:47.639 routes. Maybe it looks for other sources citing 00:04:47.639 --> 00:04:50.079 that blocked content or finds public aggregators 00:04:50.079 --> 00:04:52.360 that summarized it. It has these little problem 00:04:52.360 --> 00:04:54.560 -solving loops built in. You know, I still wrestle 00:04:54.560 --> 00:04:57.040 with prompt drift myself sometimes. Trying to 00:04:57.040 --> 00:05:00.139 make a prompt so perfect it covers every possibility. 00:05:00.740 --> 00:05:03.040 It can get really complicated, really fast. How 00:05:03.040 --> 00:05:05.720 do you avoid that? Yeah, that's a common challenge. 00:05:05.899 --> 00:05:07.980 Finding that balance between detail and clarity 00:05:07.980 --> 00:05:10.339 definitely takes some practice. But when it works, 00:05:10.500 --> 00:05:13.019 the output can be incredibly rich. You get a 00:05:13.019 --> 00:05:15.180 detailed report, often with really specific ideas 00:05:15.180 --> 00:05:18.180 like investing with one to five million VND month, 00:05:18.680 --> 00:05:22.100 ETF roadmap for beginners 2025, or BNPL boom. 00:05:22.420 --> 00:05:24.319 How does buy now pay later harm your wallet? 00:05:25.160 --> 00:05:27.220 These aren't just generic suggestions. They're 00:05:27.220 --> 00:05:29.939 tailored data -driven insights based on the criteria 00:05:29.939 --> 00:05:32.259 you set. And a good tip there is probably not 00:05:32.259 --> 00:05:34.740 to just take the first output as final, right? 00:05:35.120 --> 00:05:36.939 Like if the YouTube research seems a bit thin, 00:05:37.180 --> 00:05:39.199 you can follow up. Ask it to dig deeper into 00:05:39.199 --> 00:05:41.420 comments for audience questions. Exactly right. 00:05:41.459 --> 00:05:44.000 It's often an iterative process. Refine, ask 00:05:44.000 --> 00:05:47.699 again, dig deeper. This use case done well can 00:05:47.699 --> 00:05:50.560 genuinely save you, say, three, four hours of 00:05:50.560 --> 00:05:53.120 manual research for each piece of content. That's 00:05:53.120 --> 00:05:55.800 a huge efficiency gain. So thinking about this 00:05:55.800 --> 00:05:58.519 trend finding. What's the key takeaway to make 00:05:58.519 --> 00:06:01.399 it truly effective? Just asking for trends isn't 00:06:01.399 --> 00:06:03.740 enough. Detailed prompts and smart follow -up 00:06:03.740 --> 00:06:06.360 commands are crucial for success. You guide the 00:06:06.360 --> 00:06:09.019 process. Okay, let's shift gears. Conversion 00:06:09.019 --> 00:06:12.699 rate optimization or CRO? For many smaller businesses, 00:06:12.939 --> 00:06:14.939 getting a professional website audit feels out 00:06:14.939 --> 00:06:17.819 of reach. It's expensive. But CRO, just improving 00:06:17.819 --> 00:06:19.680 the percentage of visitors who take an action, 00:06:19.819 --> 00:06:22.120 like buying something, is so important. Can agent 00:06:22.120 --> 00:06:24.860 mode step in here? Act like a virtual CRO expert. 00:06:24.959 --> 00:06:27.439 Yes, this is another really practical application. 00:06:27.699 --> 00:06:29.680 And you're right, professional CRO audits cost 00:06:29.680 --> 00:06:32.899 a lot but offer huge value. Agent mode can give 00:06:32.899 --> 00:06:35.899 you a surprisingly detailed analysis here. The 00:06:35.899 --> 00:06:38.740 key, again, is that detailed prompt. You'd instruct 00:06:38.740 --> 00:06:42.620 it. Act as a UX UI expert with 10 years of experience. 00:06:43.560 --> 00:06:45.779 And UX UI, of course, is user experience and 00:06:45.779 --> 00:06:48.399 user interface, basically. How easy and pleasant 00:06:48.399 --> 00:06:51.439 your site is to use. Giving it that specific 00:06:51.439 --> 00:06:54.420 senior persona changes how it analyzes things. 00:06:54.560 --> 00:06:57.000 And then the mission. Analyze your e -commerce 00:06:57.000 --> 00:06:59.500 site, find the friction points, the bottlenecks, 00:06:59.759 --> 00:07:01.600 and suggest concrete ways to improve click -through 00:07:01.600 --> 00:07:04.500 rates or CTR and overall conversions. Absolutely. 00:07:04.800 --> 00:07:06.860 Your prompt should detail the outlaces process. 00:07:06.939 --> 00:07:09.120 Maybe ask it to start with foundational research 00:07:09.120 --> 00:07:11.399 on industry best practices for your specific 00:07:11.399 --> 00:07:14.540 niche. Then move to specific page analysis, really 00:07:14.540 --> 00:07:16.259 looking at things like your product thumbnails, 00:07:16.740 --> 00:07:18.800 the clarity and placement of your call -to -action 00:07:18.800 --> 00:07:20.600 buttons, you know, those buy now or learn more 00:07:20.600 --> 00:07:23.160 buttons, and how well you communicate product 00:07:23.160 --> 00:07:26.439 benefits. You're guiding its focus. And crucially, 00:07:26.860 --> 00:07:29.199 specifying the output format helps a lot, like 00:07:29.199 --> 00:07:32.110 a structured Markdown table. product area, specific 00:07:32.110 --> 00:07:34.689 issue, negative impact, and a suggested fix. 00:07:34.910 --> 00:07:37.029 Exactly. And this is where it shines. Imagine 00:07:37.029 --> 00:07:39.610 getting specific, actionable recommendations 00:07:39.610 --> 00:07:42.350 like replace the rotating hero carousel with 00:07:42.350 --> 00:07:44.589 a static image. They usually perform better. 00:07:45.269 --> 00:07:48.290 Or fix cropped product thumbnails so people can 00:07:48.290 --> 00:07:51.209 actually see what they're buying. Or always show 00:07:51.209 --> 00:07:54.470 prices clearly on product cards. Maybe even move 00:07:54.470 --> 00:07:57.470 ad to cart to be the main CTA on product cards. 00:07:57.930 --> 00:08:00.209 These aren't vague ideas. They're concrete suggestions. 00:08:00.170 --> 00:08:03.529 So, after maybe 10 -15 minutes, you get this 00:08:03.529 --> 00:08:05.930 detailed audit. You might point out inconsistent 00:08:05.930 --> 00:08:08.329 product images or benefit statements that are 00:08:08.329 --> 00:08:10.750 too generic, or even a button color that just 00:08:10.750 --> 00:08:12.470 blends into the background. These are the kinds 00:08:12.470 --> 00:08:14.850 of small changes that can directly impact revenue. 00:08:15.110 --> 00:08:16.550 What's the biggest risk if you don't put the 00:08:16.550 --> 00:08:18.490 effort into a detailed prompt for this kind of 00:08:18.490 --> 00:08:21.170 task? You'll just get vague, generic advice, 00:08:21.350 --> 00:08:23.970 not specific, actionable solutions tailored to 00:08:23.970 --> 00:08:27.269 your site. Okay, market research. Before launching 00:08:27.269 --> 00:08:29.970 anything new, understanding customer pain points 00:08:29.970 --> 00:08:33.629 is everything, isn't it? Agent mode sounds like 00:08:33.629 --> 00:08:36.090 it could automate reading hundreds, maybe thousands 00:08:36.090 --> 00:08:38.750 of customer reviews to find those hidden frustrations 00:08:38.750 --> 00:08:42.049 and desires. Oh, this is a huge time saver. Huge. 00:08:42.350 --> 00:08:44.190 And it helps take some of the human bias out 00:08:44.190 --> 00:08:46.730 of the research process, too. You just need to 00:08:46.730 --> 00:08:49.389 clearly define the scope. Tell it your product, 00:08:49.730 --> 00:08:51.970 like a new smart thermos, and task it to find 00:08:51.970 --> 00:08:54.429 common problems and wishes for similar products 00:08:54.429 --> 00:08:57.110 already out there. That competitive angle is 00:08:57.110 --> 00:08:59.009 vital. And you can point it towards specific 00:08:59.009 --> 00:09:01.769 data sources, like tech blogs, YouTube reviews, 00:09:02.090 --> 00:09:03.669 forums where people actually talk about these 00:09:03.669 --> 00:09:05.690 things. You could even build in a backup plan. 00:09:05.909 --> 00:09:07.970 Right? Like if we can't access Amazon reviews 00:09:07.970 --> 00:09:09.950 directly, tell it to look for review roundup 00:09:09.950 --> 00:09:12.690 articles instead. Exactly. You're building resilience 00:09:12.690 --> 00:09:15.090 into its research plan. Then it gets to work. 00:09:15.309 --> 00:09:18.110 It collects and pulls out the common complaints, 00:09:18.169 --> 00:09:20.149 the common praises, and starts grouping them 00:09:20.149 --> 00:09:23.289 into themes, things like battery life, durability, 00:09:23.610 --> 00:09:26.990 connectivity issues. It also identifies sentiment 00:09:26.990 --> 00:09:29.289 keywords of those words that show positive or 00:09:29.289 --> 00:09:32.789 negative feelings, like frustrating or unreliable 00:09:32.789 --> 00:09:36.490 versus delightful or robust. It understands the 00:09:36.490 --> 00:09:38.370 emotion behind the words. And the output could 00:09:38.370 --> 00:09:40.830 be something like a table showing the complaint 00:09:40.830 --> 00:09:43.570 theme, maybe an example quote from real review, 00:09:44.009 --> 00:09:45.889 the mentioned frequency, how often it comes up, 00:09:45.909 --> 00:09:48.210 and importantly, a proposed feature solution. 00:09:48.629 --> 00:09:50.610 Could it even generate data for a pie chart, 00:09:50.629 --> 00:09:53.409 maybe, to visualize the biggest issues? Yeah, 00:09:53.429 --> 00:09:56.389 absolutely. So for that smart thermos example, 00:09:56.629 --> 00:09:59.330 it might find... Battery life shorter than expected 00:09:59.330 --> 00:10:02.629 is, say, a 22 % complaint. And right next to 00:10:02.629 --> 00:10:05.129 that, it suggests a solution. Increase battery 00:10:05.129 --> 00:10:08.090 to three, four hours minimum, add USB -C fast 00:10:08.090 --> 00:10:10.049 charging, maybe a wireless charging coaster. 00:10:10.669 --> 00:10:13.350 Or it finds app Bluetooth sync is unreliable, 00:10:13.490 --> 00:10:17.309 comes up 20 % of the time, and suggests use BLE 00:10:17.309 --> 00:10:20.990 5 .3, add an offline mode, build robust reconnection 00:10:20.990 --> 00:10:23.570 logic. It connects the problem directly to a 00:10:23.570 --> 00:10:25.750 potential solution. Manually reading all those 00:10:25.750 --> 00:10:28.809 reviews takes days. And honestly, it's easy to 00:10:28.809 --> 00:10:30.750 develop confirmation bias, right? You start noticing 00:10:30.750 --> 00:10:32.769 the complaints that match your own ideas. Absolutely. 00:10:33.029 --> 00:10:35.389 Human bias is a killer here. Agent mode just 00:10:35.389 --> 00:10:37.250 looks at the patterns, the frequency. It doesn't 00:10:37.250 --> 00:10:38.909 have preconceived notions. It just says, battery 00:10:38.909 --> 00:10:40.889 life is the biggest issue, mentioned 22 % of 00:10:40.889 --> 00:10:43.049 the time. It gives you that objective view. How 00:10:43.049 --> 00:10:45.029 does this directly shape product development 00:10:45.029 --> 00:10:46.649 then? It seems like more than just a feature 00:10:46.649 --> 00:10:49.389 list. It pinpoints real user problems, which 00:10:49.389 --> 00:10:51.750 directly guides creating features people actually 00:10:51.750 --> 00:10:54.720 want and need. Let's talk creative work. If you're 00:10:54.720 --> 00:10:57.139 designing products or marketing materials, knowing 00:10:57.139 --> 00:11:00.080 what designs are actually selling, not just what 00:11:00.080 --> 00:11:02.399 looks cool or what you personally like, is super 00:11:02.399 --> 00:11:05.440 important. Can agent mode help research successful 00:11:05.440 --> 00:11:08.539 designs and even create visual inspiration, like 00:11:08.539 --> 00:11:11.139 mood boards, those collections of images that 00:11:11.139 --> 00:11:14.000 set a style? Exactly. This is about moving from 00:11:14.000 --> 00:11:17.720 just subjective taste to... data -informed creativity. 00:11:18.320 --> 00:11:20.740 But again, the prompt needs specific success 00:11:20.740 --> 00:11:23.019 criteria. Don't just ask for popular designs. 00:11:23.519 --> 00:11:25.779 Define popular. Maybe it's designs with over 00:11:25.779 --> 00:11:28.879 100 sales or 50 positive reviews uploaded in 00:11:28.879 --> 00:11:30.980 the last year. You need those metrics. So you 00:11:30.980 --> 00:11:33.320 give it your niche, say, programming t -shirt 00:11:33.320 --> 00:11:35.879 designs, tell the success metrics, and ask it 00:11:35.879 --> 00:11:38.120 to collect images or detailed descriptions, and 00:11:38.120 --> 00:11:40.019 then analyze each one. Look at the style, colors, 00:11:40.220 --> 00:11:42.600 fonts, graphics, like a design critic crossed 00:11:42.600 --> 00:11:46.129 with a data analyst. Precisely. The agent then 00:11:46.129 --> 00:11:49.570 browses platforms, maybe Etsy, Redbubble, filters 00:11:49.570 --> 00:11:52.129 based on your rules, analyzes the sales data, 00:11:52.269 --> 00:11:54.870 the reviews, and identifies common threads among 00:11:54.870 --> 00:11:56.990 the winners. Now initially, it might just give 00:11:56.990 --> 00:11:58.970 you back a data table describing the designs, 00:11:59.570 --> 00:12:03.090 like design A, works on my machine, meme, simple 00:12:03.090 --> 00:12:06.960 text, dark shirt. Useful, but not visual. Ah, 00:12:07.039 --> 00:12:09.279 okay. So getting the actual visual mood board 00:12:09.279 --> 00:12:11.679 requires an extra step, doesn't it? Yes. This 00:12:11.679 --> 00:12:14.120 is where you smartly switch tools within the 00:12:14.120 --> 00:12:16.320 agent mode workflow. You tell it, okay, based 00:12:16.320 --> 00:12:18.960 on that analysis table you just made, now use 00:12:18.960 --> 00:12:22.340 daily3 to create a visual mood board. And daily 00:12:22.340 --> 00:12:25.100 three, for listeners, is that AI model that generates 00:12:25.100 --> 00:12:27.379 images from text descriptions. You could ask 00:12:27.379 --> 00:12:30.039 for, say, six to eight t -shirt mock -ups showing 00:12:30.039 --> 00:12:32.360 the top design styles using diverse colors and 00:12:32.360 --> 00:12:34.720 fonts, all based on the sales data it just crunched. 00:12:34.779 --> 00:12:36.899 So the result isn't just a pretty picture. It's 00:12:36.899 --> 00:12:38.820 a mood board grounded in what's actually proven 00:12:38.820 --> 00:12:41.620 to sell. That helps make much safer, more effective 00:12:41.620 --> 00:12:44.679 creative choices. Less guesswork. And this whole 00:12:44.679 --> 00:12:46.659 concept gets even more powerful when you bring 00:12:46.659 --> 00:12:50.179 in connectors. Ah, yes. Connectors. For anyone 00:12:50.179 --> 00:12:52.600 unfamiliar, connectors are basically chat GPT's 00:12:52.600 --> 00:12:55.399 way of linking up and talking to other apps you 00:12:55.399 --> 00:12:57.519 use. Right. You can connect it to things like 00:12:57.519 --> 00:13:02.360 Google Drive, your calendar, Gmail, Slack, Microsoft 00:13:02.360 --> 00:13:04.960 Teams, a whole range of business tools. Setting 00:13:04.960 --> 00:13:06.779 it up is usually pretty straightforward. Go into 00:13:06.779 --> 00:13:09.139 settings, find connectors, pick your app, and 00:13:09.139 --> 00:13:12.500 grant the permissions it needs. Whoa. Okay, imagine 00:13:12.500 --> 00:13:15.139 scaling this up. Once it's connected, you could 00:13:15.139 --> 00:13:18.200 automate some really complex workflows, like 00:13:18.200 --> 00:13:20.799 picture telling it. Every Monday morning, research 00:13:20.799 --> 00:13:23.299 the blogs of my top three competitors, summarize 00:13:23.299 --> 00:13:25.840 any new articles they published, analyze their 00:13:25.840 --> 00:13:28.419 content strategy shifts, save that full report 00:13:28.419 --> 00:13:30.600 to a Google Doc, and then email me notification 00:13:30.600 --> 00:13:34.200 with the link. Exactly. Or, based on that trend 00:13:34.200 --> 00:13:36.279 research you did earlier, generate 10 content 00:13:36.279 --> 00:13:39.159 ideas, and for each one, create an event in my 00:13:39.159 --> 00:13:41.200 Google Calendar, scheduling it for next week. 00:13:41.500 --> 00:13:44.120 Hmm. The possibilities for automating routine 00:13:44.120 --> 00:13:46.899 tasks are genuinely transformative. It really 00:13:46.899 --> 00:13:49.610 opens things up. That level of integration sounds 00:13:49.610 --> 00:13:52.110 amazing, but it also brings us back to the security 00:13:52.110 --> 00:13:53.909 point, doesn't it? You're giving it keys to your 00:13:53.909 --> 00:13:56.029 kingdom, essentially. Yeah, absolutely. That 00:13:56.029 --> 00:13:58.570 power comes with responsibility, which makes 00:13:58.570 --> 00:14:01.070 our earlier caution about data access even more 00:14:01.070 --> 00:14:03.649 important when you start using connectors. So 00:14:03.649 --> 00:14:05.750 what's the SQL sauce for getting that visual 00:14:05.750 --> 00:14:08.269 mood board, not just the data points? It's that 00:14:08.269 --> 00:14:12.659 two -step process. Analyze first. Then use daily 00:14:12.659 --> 00:14:16.000 E3 for image generation based on the Ollis's. 00:14:16.139 --> 00:14:18.240 OK, we've seen some incredible potential here, 00:14:18.659 --> 00:14:21.600 but let's be honest. We need a clear eyed look 00:14:21.600 --> 00:14:24.440 at agent modes performance right now. Its strengths. 00:14:24.970 --> 00:14:28.169 but also its weaknesses. What does it really 00:14:28.169 --> 00:14:31.289 nail? Definitely. It truly excels at those massive 00:14:31.289 --> 00:14:33.210 research projects, the kind that would take a 00:14:33.210 --> 00:14:36.730 human hours, maybe days. It's fantastic for repetitive 00:14:36.730 --> 00:14:39.490 tasks, for spotting patterns across lots of different 00:14:39.490 --> 00:14:42.269 data sources, and for structured analysis. And 00:14:42.269 --> 00:14:44.330 a key thing is, it follows complex instructions 00:14:44.330 --> 00:14:46.850 step by step without skipping things, which humans, 00:14:47.029 --> 00:14:48.730 well, we sometimes do, especially when tasks 00:14:48.730 --> 00:14:51.169 get tedious. That consistency is a big plus. 00:14:51.370 --> 00:14:53.450 But it's not flawless. What are the current limitations? 00:14:53.549 --> 00:14:55.509 Where does it still kind of struggle? Well, as 00:14:55.509 --> 00:14:57.909 we mentioned, website blocking is a significant 00:14:57.909 --> 00:15:00.990 one. Many major sites, especially e -commerce 00:15:00.990 --> 00:15:04.009 and big media outlets, have strong anti -bot 00:15:04.009 --> 00:15:06.710 shields, and the agent can't always get through. 00:15:06.970 --> 00:15:09.090 Sometimes the results might be incomplete or 00:15:09.090 --> 00:15:12.629 it might pull in irrelevant info. So human verification 00:15:12.629 --> 00:15:15.570 and editing are absolutely essential. It's not 00:15:15.570 --> 00:15:18.049 quite set it and forget it yet. And I've heard 00:15:18.049 --> 00:15:20.769 the timing can be a bit unpredictable, too. Like 00:15:20.769 --> 00:15:23.090 sometimes a task takes way longer than you'd 00:15:23.090 --> 00:15:25.639 expect, and occasionally it might just Stop, 00:15:25.899 --> 00:15:28.399 mid -task, without a clear warning, without a 00:15:28.399 --> 00:15:30.100 frustrating. Yeah, that can happen. It's still 00:15:30.100 --> 00:15:32.100 evolving tech. And that brings us to the critical 00:15:32.100 --> 00:15:35.279 safety and security points. Remember, the agent 00:15:35.279 --> 00:15:37.200 sees whatever you give it permission to see. 00:15:37.259 --> 00:15:38.940 If you connect your Gmail, it can read your email. 00:15:39.000 --> 00:15:41.500 If you connect Drive, it sees those files. It 00:15:41.500 --> 00:15:43.299 can also view websites you happen to be logged 00:15:43.299 --> 00:15:45.360 into within its own browser session. So if you're 00:15:45.360 --> 00:15:46.919 logged into sensitive accounts, technically, 00:15:47.100 --> 00:15:49.639 it could see that page content. OpenAI says they 00:15:49.639 --> 00:15:51.639 don't store this interaction data long -term, 00:15:51.720 --> 00:15:54.559 which provides some comfort. But caution is definitely 00:15:54.559 --> 00:15:57.000 the best approach here. We'd strongly recommend 00:15:57.000 --> 00:15:59.240 not connecting super sensitive business accounts, 00:15:59.620 --> 00:16:01.860 things with private customer data or financial 00:16:01.860 --> 00:16:04.379 info, at least until the technology matures more 00:16:04.379 --> 00:16:07.340 and the security aspects are even clearer. Agreed. 00:16:07.600 --> 00:16:10.379 Our honest take right now. Agent mode is excellent 00:16:10.379 --> 00:16:14.019 for research, analysis, ideation tasks, where 00:16:14.019 --> 00:16:16.500 you can easily check, refine, and ultimately 00:16:16.500 --> 00:16:19.240 control the output. We wouldn't yet fully trust 00:16:19.240 --> 00:16:22.039 it for tasks that could directly harm your business 00:16:22.039 --> 00:16:24.039 if something went wrong, like, say, having it 00:16:24.039 --> 00:16:25.700 communicate directly with clients or execute 00:16:25.700 --> 00:16:28.200 financial transactions. Maybe down the road, 00:16:28.240 --> 00:16:31.269 but not quite yet. So boiling it down. What's 00:16:31.269 --> 00:16:33.490 the single most critical thing to keep in mind 00:16:33.490 --> 00:16:36.269 when using agent mode? Always verify the results 00:16:36.269 --> 00:16:38.929 and be extremely cautious with sensitive data 00:16:38.929 --> 00:16:42.830 access. Trust, but verify. Sponsor. So wrapping 00:16:42.830 --> 00:16:45.070 this up, what's the big picture here? What does 00:16:45.070 --> 00:16:47.789 this all really mean? OK. This deep dive really 00:16:47.789 --> 00:16:52.090 suggests that the era of, well, maybe AI employees 00:16:52.090 --> 00:16:54.250 isn't quite the right term, but AI assistants 00:16:54.250 --> 00:16:58.070 that can act autonomously, that era has truly 00:16:58.070 --> 00:17:00.480 arrived. Agent mode feels like a fundamental 00:17:00.480 --> 00:17:03.039 shift in how we approach work, moving beyond 00:17:03.039 --> 00:17:05.299 just using tools to having genuine automated 00:17:05.299 --> 00:17:07.380 help. Yeah, and it's crucial to frame it correctly. 00:17:07.420 --> 00:17:10.019 It's not about AI replacing human creativity 00:17:10.019 --> 00:17:12.579 or strategic thinking. Not at all. It's about 00:17:12.579 --> 00:17:14.759 freeing you up from the time -consuming, often 00:17:14.759 --> 00:17:17.019 tedious, research and execution parts of the 00:17:17.019 --> 00:17:19.160 job so you can focus your energy on the higher 00:17:19.160 --> 00:17:21.319 -level strategic decisions, the insights, the 00:17:21.319 --> 00:17:23.259 things that really drive growth and innovation. 00:17:23.599 --> 00:17:26.180 It seems clear that businesses and even individuals 00:17:26.180 --> 00:17:28.140 who start experimenting with these autonomous 00:17:28.140 --> 00:17:31.009 AI AI agents now are likely going to gain a significant 00:17:31.009 --> 00:17:33.970 edge, the potential to save time, and just the 00:17:33.970 --> 00:17:36.269 sheer scale at which it can process information. 00:17:37.049 --> 00:17:39.329 It's undeniable. It really is about working smarter. 00:17:39.829 --> 00:17:41.730 Definitely. So your roadmap to becoming an agent 00:17:41.730 --> 00:17:44.190 mode master really starts right now. Try one 00:17:44.190 --> 00:17:46.450 of the use cases we talked about. Maybe start 00:17:46.450 --> 00:17:48.609 with the website analysis on your own site, something 00:17:48.609 --> 00:17:51.509 you know well. Begin with relatively simple tasks, 00:17:51.829 --> 00:17:55.210 build complexity gradually, and critically, always. 00:17:55.450 --> 00:17:58.269 Always verify the results. Don't just trust them 00:17:58.269 --> 00:18:00.609 blindly, especially at first. And experiment. 00:18:00.750 --> 00:18:02.650 Play around with different prompts. Change the 00:18:02.650 --> 00:18:05.269 role you sign the AI. Tweak the context. Adjust 00:18:05.269 --> 00:18:07.289 the output structure. See what works best for 00:18:07.289 --> 00:18:10.130 your specific needs and goals. And, as we keep 00:18:10.130 --> 00:18:12.650 saying, always prioritize security when you're 00:18:12.650 --> 00:18:15.150 thinking about connecting accounts. Be mindful 00:18:15.150 --> 00:18:17.690 of what you're granting access to. The technology 00:18:17.690 --> 00:18:19.930 isn't perfect, no doubt. It's still evolving 00:18:19.930 --> 00:18:24.000 fast. But it's powerful enough. right now, today, 00:18:24.240 --> 00:18:26.559 to be a genuine game changer in how you work, 00:18:26.960 --> 00:18:28.700 and maybe even how you think about problems. 00:18:28.960 --> 00:18:31.640 Just remember this framing. AI handles the heavy 00:18:31.640 --> 00:18:33.880 lifting of research and execution. You handle 00:18:33.880 --> 00:18:36.599 the vision, the strategy, the critical thinking, 00:18:36.680 --> 00:18:39.240 and the human touch. That's a really powerful 00:18:39.240 --> 00:18:41.299 combination for anyone looking to innovate and 00:18:41.299 --> 00:18:41.759 work smarter.