WEBVTT

00:00:00.000 --> 00:00:03.100
Imagine waking up and there's a social media

00:00:03.100 --> 00:00:05.900
ad, perfectly crafted, already published, automatically.

00:00:06.320 --> 00:00:09.199
Or maybe getting a brief from an AI research

00:00:09.199 --> 00:00:11.480
intern on the latest industry news, complete

00:00:11.480 --> 00:00:13.800
with all the sources cited. Perhaps even watching

00:00:13.800 --> 00:00:16.539
a bot just tirelessly migrate thousands of old

00:00:16.539 --> 00:00:18.519
customer records right there on your screen.

00:00:18.679 --> 00:00:20.620
These things, they aren't really far off dreams

00:00:20.620 --> 00:00:23.320
anymore. They're actually happening now. Welcome

00:00:23.320 --> 00:00:25.039
to the Deep Dive. Today, we're going to dig into

00:00:25.039 --> 00:00:27.500
this really remarkable guide. It cuts through,

00:00:27.620 --> 00:00:29.960
well... all the noise around AI and automation

00:00:29.960 --> 00:00:32.820
tools. It's basically a definitive list of 12

00:00:32.820 --> 00:00:35.500
game -changing integrations for any ADN. And

00:00:35.500 --> 00:00:37.560
if you don't know ADN, it's this super powerful

00:00:37.560 --> 00:00:40.320
open -source automation platform. These are the

00:00:40.320 --> 00:00:42.320
tools the pros are, let's say, really using.

00:00:42.640 --> 00:00:44.679
Exactly. Yeah, it can feel totally overwhelming

00:00:44.679 --> 00:00:47.159
trying to keep up with everything, right? So

00:00:47.159 --> 00:00:50.159
our mission today, really, is to distill the

00:00:50.159 --> 00:00:53.479
absolute must -knows from this source. We're

00:00:53.479 --> 00:00:55.799
going to look at each tool, what it does, how

00:00:55.799 --> 00:00:58.310
it works out in the real world. and maybe most

00:00:58.310 --> 00:01:00.609
importantly, what kind of pro -level upgrades

00:01:00.609 --> 00:01:03.909
it really enables. Think of it as a shortcut

00:01:03.909 --> 00:01:06.790
to wielding some serious automation power. By

00:01:06.790 --> 00:01:08.650
the end, you'll know exactly how these tools

00:01:08.650 --> 00:01:10.709
can supercharge your own workflow, saving you

00:01:10.709 --> 00:01:13.370
hours, not just minutes. Okay, let's unpack this.

00:01:13.510 --> 00:01:16.049
All right, first up, Google Gemini, it feels

00:01:16.049 --> 00:01:18.180
like. Well, more than just another AI model.

00:01:18.319 --> 00:01:20.239
It's almost like a system upgrade for your whole

00:01:20.239 --> 00:01:22.459
automation setup. I kind of think of it as the

00:01:22.459 --> 00:01:25.280
decathlete of the AI world. What's really interesting

00:01:25.280 --> 00:01:28.379
here is it's direct line into Google's huge multimodal

00:01:28.379 --> 00:01:30.400
infrastructure, which means it's surprisingly

00:01:30.400 --> 00:01:33.939
good at, well, almost everything. analyzing audio,

00:01:34.219 --> 00:01:36.379
transcribing recordings, even examining complex

00:01:36.379 --> 00:01:38.980
documents. Yeah, and for automation specifically,

00:01:39.219 --> 00:01:43.340
the immediate power is video and image generation

00:01:43.340 --> 00:01:46.060
from just simple text prompts. Imagine creating

00:01:46.060 --> 00:01:49.060
really high -quality visual media, like world

00:01:49.060 --> 00:01:51.920
-class stuff, right inside an automation workflow.

00:01:52.079 --> 00:01:55.019
It's pretty wild. Okay, picture this. A local

00:01:55.019 --> 00:01:58.379
coffee shop. An N8N workflow triggers every morning.

00:01:58.640 --> 00:02:01.379
It sends a prompt to Gemini like, create a cinematic

00:02:01.379 --> 00:02:04.280
15 -second video of hot tea being poured in a

00:02:04.280 --> 00:02:06.959
cozy kitchen. And maybe 60 seconds later, boom,

00:02:07.140 --> 00:02:09.219
a ready -to -use video ad for their drink of

00:02:09.219 --> 00:02:11.080
the day pops up on their phone ready for Instagram

00:02:11.080 --> 00:02:13.419
or TikTok. That's basically a whole video production

00:02:13.419 --> 00:02:15.979
studio in one little automation node. OK, but

00:02:15.979 --> 00:02:17.639
here's where it gets really interesting for like

00:02:17.639 --> 00:02:19.919
deeper analysis. You could feed Gemini, say,

00:02:20.060 --> 00:02:23.099
a customer video review. One node transcribes

00:02:23.099 --> 00:02:25.340
the audio. Another analyzes the visual stuff.

00:02:25.419 --> 00:02:27.520
They smiling. How's the product look? A third

00:02:27.520 --> 00:02:29.439
node does sentiment analysis on the text itself.

00:02:29.759 --> 00:02:32.580
Then a fourth Gemini node pulls all three of

00:02:32.580 --> 00:02:35.240
those outputs together into a single really rich

00:02:35.240 --> 00:02:37.759
summary. This gives you insights you just couldn't

00:02:37.759 --> 00:02:39.860
get from text alone. So the fundamental shift

00:02:39.860 --> 00:02:42.439
Gemini offers, it seems, is going beyond just

00:02:42.439 --> 00:02:45.169
text or images separately. Yeah, it's analyzing

00:02:45.169 --> 00:02:47.990
video, audio, and text together, giving you richer,

00:02:48.069 --> 00:02:51.889
unified insights. Okay, next up, Eleven Labs.

00:02:52.349 --> 00:02:54.590
This tool takes your automations beyond just

00:02:54.590 --> 00:02:57.610
silent text. It adds personal, human -sounding

00:02:57.610 --> 00:02:59.669
voices. I kind of like to call it the Morgan

00:02:59.669 --> 00:03:01.830
Freeman for your workflows. It really does have

00:03:01.830 --> 00:03:04.370
best -in -class text -to -speech, a huge library

00:03:04.370 --> 00:03:07.250
of amazing voices, and this kind of freaky ability

00:03:07.250 --> 00:03:10.189
to clone your own voice from just a few minutes

00:03:10.189 --> 00:03:11.870
of audio. Right, so instead of that generic,

00:03:12.030 --> 00:03:34.270
your order has shipped, email. Yeah. Oh, totally.

00:03:35.409 --> 00:03:37.289
And you can take it even further. You can use

00:03:37.289 --> 00:03:40.110
multiple cloned voices to create, like, full

00:03:40.110 --> 00:03:42.330
scripted dialogues. Think about employee training.

00:03:42.610 --> 00:03:46.030
An NEN workflow could generate an audio scenario,

00:03:46.349 --> 00:03:48.789
maybe a conversation between a seasoned sales

00:03:48.789 --> 00:03:51.710
manager and a new rep handling a tricky client

00:03:51.710 --> 00:03:54.550
objection. It turns passive learning into this

00:03:54.550 --> 00:03:57.889
engaging little audio play. So adding voice isn't

00:03:57.889 --> 00:04:00.650
just like a gimmick then? No way. It creates

00:04:00.650 --> 00:04:03.090
more personal, more memorable, just more human

00:04:03.090 --> 00:04:06.210
customer experiences. The internet today. Wow.

00:04:07.140 --> 00:04:08.939
It can feel like this chaotic jungle, right?

00:04:08.979 --> 00:04:11.580
All these dynamic apps. Trying to pull clean

00:04:11.580 --> 00:04:13.699
data from sites like Google Maps or Instagram.

00:04:14.000 --> 00:04:15.439
Honestly, it feels like trying to catch a fish

00:04:15.439 --> 00:04:16.980
in the Amazon with your bare hands. You're probably

00:04:16.980 --> 00:04:18.839
just going to get a lot of mud. That chuckles.

00:04:19.120 --> 00:04:20.540
Yeah, that's a good way to put it. And that's

00:04:20.540 --> 00:04:22.600
where Appify comes in. It's kind of like your

00:04:22.600 --> 00:04:25.129
high -tech fishing boat. It takes all that unstructured

00:04:25.129 --> 00:04:26.990
chaos from pretty much any website and turns

00:04:26.990 --> 00:04:29.589
it into clean, structured spreadsheets ready

00:04:29.589 --> 00:04:32.329
for your workflows. Appify's secret weapon is

00:04:32.329 --> 00:04:34.649
something they call actors. These are basically

00:04:34.649 --> 00:04:37.769
pre -built, specialized little robots. Experts

00:04:37.769 --> 00:04:40.350
design them for specific scraping tasks. Need

00:04:40.350 --> 00:04:42.829
data from Instagram profiles? There's an actor.

00:04:43.009 --> 00:04:45.129
Google Maps results. Yep, actor for that too.

00:04:45.269 --> 00:04:47.370
The point is, you don't need to be some master

00:04:47.370 --> 00:04:49.550
coder to get really good data. And what's great

00:04:49.550 --> 00:04:52.079
is the free tier. It's incredibly generous, gives

00:04:52.079 --> 00:04:54.339
you roughly, what, five bucks worth of web scrapes

00:04:54.339 --> 00:04:56.500
a month. For most people, that's actually plenty

00:04:56.500 --> 00:04:59.060
for some pretty ambitious projects. Let's think

00:04:59.060 --> 00:05:02.180
about a local dominator lead list. Imagine wanting

00:05:02.180 --> 00:05:04.560
a complete list of hotels and restaurants in,

00:05:04.600 --> 00:05:07.779
say, Sydney from Google Maps. An anywhere in

00:05:07.779 --> 00:05:10.740
workflow using Appify's Google Maps scrape actor.

00:05:10.939 --> 00:05:13.540
You can pull all that in maybe an hour. Names,

00:05:13.540 --> 00:05:16.060
addresses, phone numbers, ratings, all in a perfect

00:05:16.060 --> 00:05:17.870
spreadsheet. While you're, you know, off doing

00:05:17.870 --> 00:05:20.269
something else. Enjoying the beach, maybe? Or

00:05:20.269 --> 00:05:22.350
how about an e -commerce price watch? You can

00:05:22.350 --> 00:05:25.290
set up a daily NA Warren workflow. Appify scrapes

00:05:25.290 --> 00:05:27.230
the prices of your top five products from Amazon,

00:05:27.449 --> 00:05:29.670
maybe some competitors. If a competitor drops

00:05:29.670 --> 00:05:32.930
their price, bang, instant slack alert. Keeps

00:05:32.930 --> 00:05:34.449
you ahead of the game. Okay, now this is where

00:05:34.449 --> 00:05:37.149
it gets really mind -blowing. Combine Appify

00:05:37.149 --> 00:05:41.089
with something like ChatGPT. Imagine a scheduled

00:05:41.089 --> 00:05:44.709
N8N workflow. Runs daily. Uses an Amplify actor

00:05:44.709 --> 00:05:47.069
to scrape the top 20 trending posts from a hot

00:05:47.069 --> 00:05:50.449
subreddit, or maybe a viral TikTok hashtag. It

00:05:50.449 --> 00:05:53.310
feeds that text straight to ChatGPT, right, but

00:05:53.310 --> 00:05:55.209
prompted to act like a market research analyst.

00:05:56.029 --> 00:05:58.389
Identify recurring products, brands, features.

00:05:58.709 --> 00:06:01.660
List the top three emerging trends. Two sec silence.

00:06:02.019 --> 00:06:04.980
Whoa. I mean, just imagine scaling that, spotting

00:06:04.980 --> 00:06:06.899
the next big thing before anyone else even knows

00:06:06.899 --> 00:06:09.240
it exists. So, Appify's real trick is making

00:06:09.240 --> 00:06:11.500
that complex web scraping seem simple. Exactly.

00:06:11.699 --> 00:06:14.000
It uses those pre -built actors to just pull

00:06:14.000 --> 00:06:16.620
clean, structured data nice and easy. Okay, so

00:06:16.620 --> 00:06:18.839
most browser automation, it runs headless. You

00:06:18.839 --> 00:06:20.439
don't see it. It's invisible in the background.

00:06:20.920 --> 00:06:22.800
Airtop is different. It's more like screen sharing

00:06:22.800 --> 00:06:24.519
with a ghost. You can actually watch it work

00:06:24.519 --> 00:06:26.699
in real time. Yeah, and that visual feedback,

00:06:26.899 --> 00:06:28.980
that's actually huge. You give it a command,

00:06:29.100 --> 00:06:31.319
and you literally see the... ai's little cursor

00:06:31.319 --> 00:06:34.079
moving typing clicking through the task it's

00:06:34.079 --> 00:06:37.449
invaluable for debugging Especially complex stuff.

00:06:37.550 --> 00:06:40.629
And honestly, it builds a ton of trust. You see

00:06:40.629 --> 00:06:42.810
it doing the work. Think about like a legacy

00:06:42.810 --> 00:06:45.550
data migration. Some business needs to move thousands

00:06:45.550 --> 00:06:48.189
of customer records from some ancient web -based

00:06:48.189 --> 00:06:52.209
CRM. No API, nothing. Doing that manually. Soul

00:06:52.209 --> 00:06:55.050
-crushing work. With Airtop, an N8N workflow

00:06:55.050 --> 00:06:56.990
reads a spreadsheet, tells the Airtop agent,

00:06:57.129 --> 00:06:59.410
log in, go here, type this, type that, click

00:06:59.410 --> 00:07:01.709
save. You can just sit back, watch it, chew through

00:07:01.709 --> 00:07:03.790
maybe a week's worth of manual data entry in

00:07:03.790 --> 00:07:05.649
an afternoon. And here's a really cool angle.

00:07:06.040 --> 00:07:09.100
Human -AI collaboration. You can build a system

00:07:09.100 --> 00:07:12.160
for it. A human starts a task, right? Does the

00:07:12.160 --> 00:07:13.860
critical thinking parts. But when they hit a

00:07:13.860 --> 00:07:15.959
really tedious bit, like filling in 20 lines

00:07:15.959 --> 00:07:18.100
of company history or something, they click a

00:07:18.100 --> 00:07:20.399
little bookmarklet. The Airtop agent jumps in,

00:07:20.459 --> 00:07:22.000
takes over the current browser session, does

00:07:22.000 --> 00:07:24.199
the boring part, then seamlessly hands control

00:07:24.199 --> 00:07:26.980
back. It's like a true co -pilot system. So the

00:07:26.980 --> 00:07:29.339
biggest win with Airtop is really seeing it work

00:07:29.339 --> 00:07:32.060
visually. Yeah, it helps debug complicated flows

00:07:32.060 --> 00:07:34.100
and just builds that crucial trust in the automation.

00:07:34.759 --> 00:07:37.120
Standard AI models, you know, they're powerful,

00:07:37.360 --> 00:07:40.379
but they often have this limitation. Their knowledge

00:07:40.379 --> 00:07:43.220
is kind of frozen in time. Asking them about

00:07:43.220 --> 00:07:45.220
yesterday's news, it's a bit like asking a history

00:07:45.220 --> 00:07:47.620
professor to predict the future. It doesn't really

00:07:47.620 --> 00:07:50.779
work. Right, and perplexity is the, well, the

00:07:50.779 --> 00:07:52.759
elegant solution here. It's not just another

00:07:52.759 --> 00:07:55.480
language model. They call it a conversational

00:07:55.480 --> 00:07:58.300
answer engine. Think of it like the perfect research

00:07:58.300 --> 00:08:01.800
intern. You get top -tier AI reasoning. plus

00:08:01.800 --> 00:08:04.439
live internet access, plus this almost obsessive

00:08:04.439 --> 00:08:07.560
need to cite its sources. Okay, imagine an instant

00:08:07.560 --> 00:08:10.040
market report. You could have an NEN workflow,

00:08:10.379 --> 00:08:13.519
runs every morning at 8 a .m. It sends perplexity

00:08:13.519 --> 00:08:16.259
a prompt. Act like a market analyst. What were

00:08:16.259 --> 00:08:18.819
the top three biggest funding rounds in AI automation

00:08:18.819 --> 00:08:21.180
in the last week? Summarize them, include the

00:08:21.180 --> 00:08:24.100
source links. Moments later, you get this perfectly

00:08:24.100 --> 00:08:26.120
formatted Slack message, details the funding,

00:08:26.220 --> 00:08:28.459
gives concise summaries, and crucially, verifiable

00:08:28.459 --> 00:08:31.139
citations and links for every single fact. Yeah,

00:08:31.220 --> 00:08:33.600
or you could automate due diligence on potential

00:08:33.600 --> 00:08:36.620
partners. Feed it a company website. The workflow

00:08:36.620 --> 00:08:39.759
sends targeted questions to Perplexity. Summarize

00:08:39.759 --> 00:08:42.340
their products. Find the three latest news articles.

00:08:42.460 --> 00:08:45.379
Who are the key executives? It pulls all those

00:08:45.379 --> 00:08:48.259
summaries together into one clean document. You

00:08:48.259 --> 00:08:50.659
get a comprehensive brief in minutes, not hours.

00:08:50.960 --> 00:08:54.279
So Perplexity basically solves that frozen knowledge.

00:08:55.209 --> 00:08:58.389
issue we see in other AIs. Exactly. It accesses

00:08:58.389 --> 00:09:01.090
and cites live internet info for current verifiable

00:09:01.090 --> 00:09:03.610
answers. All right. Now, if you've ever built

00:09:03.610 --> 00:09:05.429
a RAG system, and that's retrieval augmented

00:09:05.429 --> 00:09:08.809
generation, where an AI uses external knowledge

00:09:08.809 --> 00:09:11.070
bases to answer questions, you've probably hit

00:09:11.070 --> 00:09:13.450
its big weakness, irrelevant search results.

00:09:13.570 --> 00:09:15.690
Oh, yeah. That's a real headache. Your support

00:09:15.690 --> 00:09:18.529
bot gets asked, what's your refund policy? The

00:09:18.529 --> 00:09:20.250
first search might pull up 10 documents that

00:09:20.250 --> 00:09:23.090
just have the word policy, privacy policy, shipping

00:09:23.090 --> 00:09:25.909
policy, maybe even the office's open door policy.

00:09:26.149 --> 00:09:28.350
The AI gets confused. It gives a vague answer.

00:09:28.470 --> 00:09:31.190
Or worse. It confidently gives the wrong answer.

00:09:31.330 --> 00:09:33.950
So re -ranker acts like an intelligent editor

00:09:33.950 --> 00:09:36.309
for that messy list. It takes those initial search

00:09:36.309 --> 00:09:39.049
results and it analyzes them for semantic relevance.

00:09:39.129 --> 00:09:41.289
It looks for the actual meaning, the intent,

00:09:41.429 --> 00:09:44.470
not just keywords. Then it re -ranks them, pushing

00:09:44.470 --> 00:09:46.889
the actual refund policy document right to the

00:09:46.889 --> 00:09:49.809
top and kind of tossing out the irrelevant noise.

00:09:50.090 --> 00:09:51.850
And that's not just technical, right? It's about

00:09:51.850 --> 00:09:54.950
building trust. Users will forgive and I don't

00:09:54.950 --> 00:09:57.240
know. They rarely forgive a confidently wrong

00:09:57.240 --> 00:10:00.000
answer. ReRanker seems like it dramatically boosts

00:10:00.000 --> 00:10:02.740
that RJ accuracy and therefore user confidence.

00:10:03.100 --> 00:10:05.539
You know, I still wrestle with prompt drift myself

00:10:05.539 --> 00:10:08.360
sometimes, especially trying to get really precise

00:10:08.360 --> 00:10:10.799
answers from an AI when there's just a sea of

00:10:10.799 --> 00:10:13.399
data. That's honestly why ReRanker is so appealing.

00:10:13.519 --> 00:10:15.559
It's like having this super smart librarian sorting

00:10:15.559 --> 00:10:18.179
things out first. And for a pro level upgrade,

00:10:18.340 --> 00:10:21.340
think about a massive like 300 page employee

00:10:21.340 --> 00:10:24.320
handbook. An employee asks something specific.

00:10:24.730 --> 00:10:27.049
What's our policy on international business travel

00:10:27.049 --> 00:10:29.269
for engineering projects and what are the per

00:10:29.269 --> 00:10:33.230
diem limits? ReRanker could surgically find the

00:10:33.230 --> 00:10:36.070
one single paragraph in that huge document that

00:10:36.070 --> 00:10:38.730
perfectly answers that question. It turns a document

00:10:38.730 --> 00:10:41.450
graveyard into an instant precise answer engine.

00:10:42.059 --> 00:10:44.580
Pretty cool. So ReRanker's core job is boosting

00:10:44.580 --> 00:10:48.159
AI answer accuracy through relevance. By intelligently

00:10:48.159 --> 00:10:50.139
filtering search results for semantic relevance,

00:10:50.299 --> 00:10:53.679
making sure the context is precise. PDS. For

00:10:53.679 --> 00:10:55.639
years, they've been the absolute bane of automation,

00:10:55.820 --> 00:10:57.340
haven't they? It's like data trapped in amber,

00:10:57.440 --> 00:10:58.899
the information's in there, but it's frozen,

00:10:59.100 --> 00:11:01.820
hard to get at, super messy to copy -paste reliably.

00:11:02.279 --> 00:11:04.779
Yeah, but the Mistral integration. It finally

00:11:04.779 --> 00:11:07.559
seems to crack that code. Mistral's AI models

00:11:07.559 --> 00:11:09.960
are exceptionally good at accurately parsing

00:11:09.960 --> 00:11:13.080
and pulling out clean, usable text, even from

00:11:13.080 --> 00:11:15.860
really complex PDFs. Okay, think about vendor

00:11:15.860 --> 00:11:19.159
invoices. Usually a human opens the PDF, finds

00:11:19.159 --> 00:11:21.220
the vendor name, invoice number, amount, due

00:11:21.220 --> 00:11:23.639
date, manually types it all into QuickBooks or

00:11:23.639 --> 00:11:25.860
Xero. With Mistral, you could have an automated

00:11:25.860 --> 00:11:28.460
N8N workflow watch and inbox for PDF attachments.

00:11:29.080 --> 00:11:31.460
Sends the PDF to Mistral, pulls out clean text.

00:11:31.620 --> 00:11:34.240
That text goes to maybe ChatGP. which is prompted

00:11:34.240 --> 00:11:36.480
to extract those specific fields, vendor name,

00:11:36.659 --> 00:11:39.139
invoice number, etc. Then the workflow automatically

00:11:39.139 --> 00:11:40.840
creates a new bill in your accounting software

00:11:40.840 --> 00:11:43.340
and schedules a payment reminder. All that manual

00:11:43.340 --> 00:11:46.039
error prone work, just gone. And this allows

00:11:46.039 --> 00:11:48.139
for a really powerful bridge between physical

00:11:48.139 --> 00:11:51.159
and digital docs too. Imagine an employee just

00:11:51.159 --> 00:11:53.399
takes a picture of a multi -page document with

00:11:53.399 --> 00:11:56.240
their phone. Those images trigger an end -make

00:11:56.240 --> 00:11:59.200
-end workflow. An OCR node that's optical character

00:11:59.200 --> 00:12:02.019
recognition converts the images to rough text.

00:12:02.240 --> 00:12:05.860
Then that messy OCR text gets sent to Mistral

00:12:05.860 --> 00:12:08.980
for a second pass. Mistral's advanced model cleans

00:12:08.980 --> 00:12:11.139
up the OCR errors, understands the structure,

00:12:11.340 --> 00:12:14.259
and spits out much, much more accurate text.

00:12:14.379 --> 00:12:16.860
Suddenly, stacks of paper become clean, structured

00:12:16.860 --> 00:12:21.269
data you can actually use. lies in finally unlocking

00:12:21.269 --> 00:12:24.450
that data stuck inside PDFs. Exactly. It accurately

00:12:24.450 --> 00:12:28.090
extracts clean, usable text from previously inaccessible

00:12:28.090 --> 00:12:32.470
PDF data. This next tool, the ThinkTool, it fundamentally

00:12:32.470 --> 00:12:35.429
upgrades how AI agents solve problems. You know,

00:12:35.490 --> 00:12:37.669
standard AI can sometimes be a bit too eager.

00:12:37.730 --> 00:12:39.350
It rushes to an answer without really considering

00:12:39.350 --> 00:12:41.429
the consequences. The ThinkTool is basically

00:12:41.429 --> 00:12:43.970
the AI's pause button. It forces the agent to

00:12:43.970 --> 00:12:46.649
stop, reflect, analyze before it acts. It's kind

00:12:46.649 --> 00:12:48.269
of the digital version of measure twice, cut

00:12:48.269 --> 00:12:50.620
once. Right. So let's say you ask your AI agent,

00:12:50.759 --> 00:12:53.519
hey, please move my 2 p .m. meeting to 3 p .m.

00:12:53.879 --> 00:12:56.340
Without think, I might just blindly do it, right?

00:12:56.360 --> 00:12:58.779
Maybe double booking you. But with think enabled,

00:12:59.039 --> 00:13:02.340
the AI's internal process might look more like,

00:13:02.440 --> 00:13:05.220
OK, user wants meeting A moved to 3 p .m. Step

00:13:05.220 --> 00:13:08.120
one, check calendar at 3 p .m. Meeting B is already

00:13:08.120 --> 00:13:11.440
there. Goal, avoid conflicts. Action, inform

00:13:11.440 --> 00:13:14.750
user of conflict. Ask for decision. That little

00:13:14.750 --> 00:13:16.870
bit of reflection avoids potentially big mistakes.

00:13:17.129 --> 00:13:18.870
Yeah, and you can use this to get much higher

00:13:18.870 --> 00:13:20.970
quality creative work, too, using what you could

00:13:20.970 --> 00:13:23.149
call an iterative thinking loop. Say the task

00:13:23.149 --> 00:13:26.049
is draft a marketing plan. Step one, AI creates

00:13:26.049 --> 00:13:28.490
a basic outline. Step two, that outline goes

00:13:28.490 --> 00:13:30.769
to a second AI node, maybe prompted like, you're

00:13:30.769 --> 00:13:32.509
a top marketing critic. Find the three biggest

00:13:32.509 --> 00:13:35.450
weaknesses here. Step three, the original outline

00:13:35.450 --> 00:13:38.370
and that critique go to a third AI node. Refine

00:13:38.370 --> 00:13:39.950
the original outline based on this critique.

00:13:40.429 --> 00:13:42.049
You can run that loop a few times before the

00:13:42.049 --> 00:13:44.330
AI even starts writing the full plan. It mimics

00:13:44.330 --> 00:13:46.590
how humans draft and refine, leading to a way

00:13:46.590 --> 00:13:49.110
more thoughtful final output. So Think basically

00:13:49.110 --> 00:13:52.110
makes the AI more reliable by forcing that pause

00:13:52.110 --> 00:13:55.990
for reflection. It forces the AI to pause, reflect,

00:13:56.309 --> 00:13:59.529
analyze. That prevents errors and improves the

00:13:59.529 --> 00:14:02.840
quality. You know, the AI world used to feel

00:14:02.840 --> 00:14:05.820
dominated by just a few really big names, but

00:14:05.820 --> 00:14:08.960
now it's definitely diversifying. These next

00:14:08.960 --> 00:14:11.059
two integrations, they give advanced automators

00:14:11.059 --> 00:14:13.960
a much wider, more strategic set of tools. First,

00:14:14.159 --> 00:14:17.179
DeepSeq. This is a powerful but also really efficient

00:14:17.179 --> 00:14:20.139
large language model in LLM. It offers pretty

00:14:20.139 --> 00:14:22.820
solid performance, but at a significantly lower

00:14:22.820 --> 00:14:25.820
price. Think about high -volume, repetitive tasks,

00:14:26.120 --> 00:14:28.379
basic text classification, simple summaries.

00:14:28.600 --> 00:14:30.860
Using a premium model for that is often total

00:14:30.860 --> 00:14:33.879
overkill, cost -wise. Switching to DeepSeek for

00:14:33.879 --> 00:14:36.799
those tasks can genuinely cut your AI API costs

00:14:36.799 --> 00:14:39.600
by like 30%, maybe even 50%, with hardly any

00:14:39.600 --> 00:14:41.679
noticeable drop in quality for those specific

00:14:41.679 --> 00:14:44.399
uses. It's just the smart, strategic budget choice

00:14:44.399 --> 00:14:47.179
for certain jobs. Okay, and this next one, OpenRouter.

00:14:47.450 --> 00:14:49.429
This sounds genuinely exciting. It tackles what

00:14:49.429 --> 00:14:51.809
I sometimes call subscription creep. If you're

00:14:51.809 --> 00:14:54.029
using OpenAI and Anthropic and Google Gemini,

00:14:54.149 --> 00:14:56.090
suddenly your credit card statement looks like

00:14:56.090 --> 00:14:58.909
the guest list for an AI conference. Yeah, exactly.

00:14:59.169 --> 00:15:01.870
So OpenRouter acts like a universal remote for

00:15:01.870 --> 00:15:04.629
AI models. It gives you access to this massive

00:15:04.629 --> 00:15:07.090
and always growing library of different AI models,

00:15:07.330 --> 00:15:09.850
but all through a single interface and, crucially,

00:15:09.970 --> 00:15:12.789
a single billing account. And this makes OpenRouter

00:15:12.789 --> 00:15:15.169
a really strategic weapon, optimizing for both

00:15:15.169 --> 00:15:18.049
cost and quality. You could build an NAN workflow

00:15:18.049 --> 00:15:20.830
that acts as an intelligent router. A user's

00:15:20.830 --> 00:15:23.269
prompt comes in. First, maybe send it to a cheap,

00:15:23.289 --> 00:15:25.710
fast AI model via open router just to classify

00:15:25.710 --> 00:15:28.129
its intent. Is it simple classification, complex

00:15:28.129 --> 00:15:30.830
reasoning, creative writing, code generation?

00:15:31.899 --> 00:15:34.659
Based on that classification, an N8n switch node

00:15:34.659 --> 00:15:36.940
dynamically routes the original prompt to the

00:15:36.940 --> 00:15:39.059
best possible model for that specific task, again

00:15:39.059 --> 00:15:41.500
via OpenRouter. Simple task. Use the cheapest

00:15:41.500 --> 00:15:44.019
model that works. Complex reasoning. Route it

00:15:44.019 --> 00:15:46.500
to a top -tier model like Cloud4 Opus. Creative

00:15:46.500 --> 00:15:49.360
writing. Pick a model known for flair. It automatically

00:15:49.360 --> 00:15:51.419
picks the right tool for the job. So DeepSeq

00:15:51.419 --> 00:15:53.879
and OpenRouter really allow for a more strategic,

00:15:54.039 --> 00:15:56.700
nuanced use of AI. Yeah, they help optimize cost

00:15:56.700 --> 00:15:59.240
and quality by routing tasks to the most appropriate

00:15:59.240 --> 00:16:02.350
AI model available. Okay, this next one. MCP,

00:16:02.470 --> 00:16:05.269
that's multi -agent communication protocol. It

00:16:05.269 --> 00:16:07.409
sounds a bit less flashy, maybe more foundational,

00:16:07.610 --> 00:16:10.809
but it seems really crucial for where AI is heading.

00:16:11.190 --> 00:16:13.289
The basic problem it's trying to solve is that

00:16:13.289 --> 00:16:16.049
most AI tools kind of live in their own walled

00:16:16.049 --> 00:16:19.090
gardens, right? Claude's over here, ChatGPT's

00:16:19.090 --> 00:16:21.649
over there. MCP is one of the first big attempts

00:16:21.649 --> 00:16:23.909
to build standardized bridges between them, letting

00:16:23.909 --> 00:16:25.750
them share information and commands seamlessly.

00:16:26.049 --> 00:16:28.570
Yeah, think of it like this. You know how massive

00:16:28.570 --> 00:16:31.509
bridges connect big cities across wide rivers?

00:16:31.990 --> 00:16:34.929
Protocols like MCP, they're the essential infrastructure

00:16:34.929 --> 00:16:36.649
that's going to eventually connect all these

00:16:36.649 --> 00:16:39.789
isolated islands in the AI world into like one

00:16:39.789 --> 00:16:42.309
single cohesive continent. It's the underlying

00:16:42.309 --> 00:16:44.450
plumbing that makes everything flow together.

00:16:44.750 --> 00:16:47.529
So the NEN integration with MCP creates this

00:16:47.529 --> 00:16:50.389
actual two -way channel between NEN and an external

00:16:50.389 --> 00:16:53.000
AI. Maybe, like, flawed. You could literally

00:16:53.000 --> 00:16:54.840
chat with Claude and say something like, hey,

00:16:54.919 --> 00:16:57.220
check my NAN calendar and find me a 30 -minute

00:16:57.220 --> 00:16:59.340
slot for a meeting with Sarah sometime next Tuesday

00:16:59.340 --> 00:17:01.600
afternoon. Right. And here's the cool back and

00:17:01.600 --> 00:17:05.380
forth, the dance. Claude, using MCP, sends a

00:17:05.380 --> 00:17:08.519
structured request over to NAN. Your NAN workflow

00:17:08.519 --> 00:17:11.220
gets it, checks your Google Calendar, finds an

00:17:11.220 --> 00:17:13.400
open slot, and sends a response back to Claude

00:17:13.400 --> 00:17:16.539
via MCP. Claude then presents it to you. Okay,

00:17:16.619 --> 00:17:19.400
I found an open slot at 3 .00 p .m. Want me to

00:17:19.400 --> 00:17:22.859
book it? You say yes, Claude sends one final

00:17:22.859 --> 00:17:25.559
command back to Anan, and Anan creates the calendar

00:17:25.559 --> 00:17:28.779
event. That's real collaboration. You know, we

00:17:28.779 --> 00:17:31.160
often get really excited about the new AI products,

00:17:31.380 --> 00:17:33.619
the flashy interfaces, but the long -term future,

00:17:33.720 --> 00:17:36.059
the real revolution, I think is defined by these

00:17:36.059 --> 00:17:38.099
underlying protocols. Think about the early internet.

00:17:38.259 --> 00:17:40.400
We love the website, sure, but the magic was

00:17:40.400 --> 00:17:42.940
really in those quiet protocols like HTTP for

00:17:42.940 --> 00:17:46.440
web pages, SMTP for email. MCP feels like the

00:17:46.440 --> 00:17:48.880
next generation of that foundational layer. It's

00:17:48.880 --> 00:17:50.920
building the rules of the road that will eventually

00:17:50.920 --> 00:17:53.319
let thousands of specialized AI agents work together,

00:17:53.500 --> 00:17:56.059
solving problems way too complex for any single

00:17:56.059 --> 00:17:59.009
AI today. These foundational protocols like MCP

00:17:59.009 --> 00:18:01.410
are critical because they let different AIs actually

00:18:01.410 --> 00:18:04.089
talk and work together. Exactly. They enable

00:18:04.089 --> 00:18:06.369
that sophisticated, seamless communication and

00:18:06.369 --> 00:18:09.450
collaboration between diverse AI systems. OK,

00:18:09.490 --> 00:18:11.390
for a lot of automators, let's be honest, the

00:18:11.390 --> 00:18:13.890
database has often just been a glorified Google

00:18:13.890 --> 00:18:16.670
Sheet or maybe Airtable. And look, they're great

00:18:16.670 --> 00:18:18.609
for getting started, but you hit a wall pretty

00:18:18.609 --> 00:18:20.869
quickly. They get slow. They get messy. They

00:18:20.869 --> 00:18:23.329
just lack the real power and scalability you

00:18:23.329 --> 00:18:25.589
need for serious automation. Right. And your

00:18:25.589 --> 00:18:27.559
super base. Think of it like a full -powered

00:18:27.559 --> 00:18:29.880
PostgreSQL database, which is serious stuff,

00:18:30.099 --> 00:18:33.259
but with a surprisingly friendly interface. And

00:18:33.259 --> 00:18:35.400
these extra superpowers built specifically for

00:18:35.400 --> 00:18:38.099
automation. If Google Sheets is like a notepad,

00:18:38.220 --> 00:18:40.519
Supabase is more like a searchable, indexed,

00:18:40.519 --> 00:18:43.420
infinitely large library for all your data. Remember

00:18:43.420 --> 00:18:45.500
that e -commerce price watch example we talked

00:18:45.500 --> 00:18:47.799
about with Appify? With a spreadsheet, you'd

00:18:47.799 --> 00:18:50.380
only see the current prices. With Supabase, you

00:18:50.380 --> 00:18:53.079
build a history. A scheduled NAN workflow runs

00:18:53.079 --> 00:18:56.140
daily, uses Appify to scrape prices, then uses

00:18:56.140 --> 00:18:58.259
the Supabase node to save that data along with

00:18:58.259 --> 00:19:00.440
a timestamp. After a few months, you have this

00:19:00.440 --> 00:19:02.660
incredibly valuable historical data. Then you

00:19:02.660 --> 00:19:05.160
can use another NAN workflow or even an AI to

00:19:05.160 --> 00:19:08.099
ask Supabase. Analyze the price history for product

00:19:08.099 --> 00:19:10.700
X. What are the trends? Does competitor Y always

00:19:10.700 --> 00:19:12.799
run sales on holidays? You're moving from just

00:19:12.799 --> 00:19:15.099
reacting day to day to actually predicting future

00:19:15.099 --> 00:19:17.440
strategy based on real data. And this part is

00:19:17.440 --> 00:19:19.880
truly revolutionary, especially for AI builders.

00:19:20.519 --> 00:19:23.299
Supabase has built -in support for vector embeddings.

00:19:23.460 --> 00:19:26.000
This basically lets you turn your private, secure

00:19:26.000 --> 00:19:28.859
Supabase database into a super -powered knowledge

00:19:28.859 --> 00:19:31.299
base for your RD systems, like the ones we discussed

00:19:31.299 --> 00:19:34.720
with Reranger. So instead of uploading your sensitive

00:19:34.720 --> 00:19:37.099
company documents to some third -party service,

00:19:37.559 --> 00:19:39.980
you create a workflow that reads your documents,

00:19:40.259 --> 00:19:42.700
uses an AI to turn them into vector embeddings,

00:19:42.779 --> 00:19:44.539
those are just numeric representations of the

00:19:44.539 --> 00:19:47.420
data's meaning, them securely in your own Supabase

00:19:47.420 --> 00:19:50.799
instance. Then your REG agent queries your private

00:19:50.799 --> 00:19:53.259
database for context. You get complete ownership

00:19:53.259 --> 00:19:55.599
control and privacy over your company's brain.

00:19:55.960 --> 00:19:58.579
That's huge. So Supabase isn't just a better

00:19:58.579 --> 00:20:01.220
database. It fundamentally changes how you manage

00:20:01.220 --> 00:20:03.440
data for automation. It provides that scalable,

00:20:03.480 --> 00:20:06.619
secure, private, real database power. It totally

00:20:06.619 --> 00:20:09.319
transforms data management. Forms. They're often

00:20:09.319 --> 00:20:11.259
the very front door for business data, right?

00:20:11.319 --> 00:20:13.619
Yeah. But most forms are, let's face it, kind

00:20:13.619 --> 00:20:16.819
of dumb. Yeah. especially when you connect it

00:20:16.819 --> 00:20:20.700
to N8n, transforms a simple form into this intelligent,

00:20:20.759 --> 00:20:23.500
powerful starting point for really complex automations.

00:20:23.680 --> 00:20:26.359
Yeah, Tally's known for two things mainly. It's

00:20:26.359 --> 00:20:29.259
super clean, almost notion -like building experience,

00:20:29.460 --> 00:20:31.819
which people love, and it's incredibly generous,

00:20:32.039 --> 00:20:34.559
free -tier, unlimited forms, unlimited submissions.

00:20:34.779 --> 00:20:37.700
It really is the perfect modern front door for

00:20:37.700 --> 00:20:40.740
kicking off your N8n workflows. Okay, imagine

00:20:40.740 --> 00:20:43.369
this. Waking up Monday morning to find three

00:20:43.369 --> 00:20:45.369
new client projects are completely set up in

00:20:45.369 --> 00:20:47.990
all your systems. Paperwork sent, team notified,

00:20:48.269 --> 00:20:51.289
project spaces created, everything. This is totally

00:20:51.289 --> 00:20:53.950
possible with a Tally powered workflow. A new

00:20:53.950 --> 00:20:56.519
client. fills out a tally .so onboarding form.

00:20:56.660 --> 00:20:59.339
When they hit submit, it triggers an NAN webhook.

00:20:59.460 --> 00:21:02.420
The workflow grabs that data and instantly creates

00:21:02.420 --> 00:21:04.480
a customer profile and maybe a subscription in

00:21:04.480 --> 00:21:06.640
Stripe, creates a new shared client folder in

00:21:06.640 --> 00:21:08.599
Google Drive, invites the client to a private

00:21:08.599 --> 00:21:10.640
Slack channel, sends a personalized welcome aboard

00:21:10.640 --> 00:21:13.299
email, adds a new project and tasks in Asana

00:21:13.299 --> 00:21:15.660
or Trello. That whole process, which might take

00:21:15.660 --> 00:21:18.700
30, even 60 minutes manually, is prone to errors,

00:21:18.839 --> 00:21:21.519
done flawlessly in seconds. And for a really

00:21:21.519 --> 00:21:24.099
pro -level setup, you can use Tally's conditional

00:21:24.099 --> 00:21:26.660
logic to turn your form into an intelligent business

00:21:26.660 --> 00:21:29.380
router. Tally lets you have hidden fields and

00:21:29.380 --> 00:21:32.480
logic based on answers. So build a contact us

00:21:32.480 --> 00:21:35.339
form with, say, a budget dropdown. If a user

00:21:35.339 --> 00:21:38.700
selects over $10 ,000, the form submission triggers

00:21:38.700 --> 00:21:42.279
your high priority lead workflow in N8N. That

00:21:42.279 --> 00:21:44.799
immediately slacks your top salesperson and maybe

00:21:44.799 --> 00:21:46.900
creates a detailed briefing doc. But if they

00:21:46.900 --> 00:21:49.019
select under $1 ,000, it triggers completely

00:21:49.019 --> 00:22:00.240
different low So Taliesin Adams really goes way

00:22:00.240 --> 00:22:02.259
beyond just being a basic form builder. Oh, yeah.

00:22:02.319 --> 00:22:04.579
It becomes an intelligent, dynamic trigger for

00:22:04.579 --> 00:22:07.599
complex, multi -step business automations. Wow.

00:22:08.400 --> 00:22:11.099
We've just walked through 12 pretty incredible

00:22:11.099 --> 00:22:13.160
tools, haven't we? It feels like a complete...

00:22:13.470 --> 00:22:15.410
professional -grade orchestra for automation.

00:22:15.710 --> 00:22:18.529
You've got these creative powerhouses like Gemini

00:22:18.529 --> 00:22:21.369
and Eleven Labs, tireless data wranglers like

00:22:21.369 --> 00:22:24.109
Apify and Mistraw. You have foundational databases

00:22:24.109 --> 00:22:27.829
like Supabase, smart filters like ReRanker, and

00:22:27.829 --> 00:22:30.650
this whole suite of different AI brains, plus

00:22:30.650 --> 00:22:32.609
the tools needed to actually bridge them together.

00:22:32.809 --> 00:22:35.910
Yeah, and in this new era, I really think your

00:22:35.910 --> 00:22:38.390
value isn't defined by trying to play every single

00:22:38.390 --> 00:22:40.630
instrument yourself. You know, being the marketer

00:22:40.630 --> 00:22:42.150
and the analyst and the administrator and the

00:22:42.150 --> 00:22:45.079
researcher all... at once your role more and

00:22:45.079 --> 00:22:48.079
more is becoming the conductor right and it ain't

00:22:48.079 --> 00:22:50.240
as your stage these 12 tools we talk about they're

00:22:50.240 --> 00:22:52.099
your orchestra sections and your creative vision

00:22:52.099 --> 00:22:55.079
that's the music your job is to write the score

00:22:55.079 --> 00:22:57.160
that's the workflow the thing that brings all

00:22:57.160 --> 00:22:58.920
these specialized powerful instruments together

00:22:58.920 --> 00:23:01.059
in perfect harmony to create something really

00:23:01.059 --> 00:23:03.960
amazing You know, for most people, Monday morning

00:23:03.960 --> 00:23:06.819
probably means another week of those same repetitive

00:23:06.819 --> 00:23:09.660
tasks, kind of playing the same single note over

00:23:09.660 --> 00:23:11.779
and over again. But for those who understand

00:23:11.779 --> 00:23:14.799
what's actually possible now, the week ahead

00:23:14.799 --> 00:23:16.700
looks fundamentally different. It's like a whole

00:23:16.700 --> 00:23:19.559
new symphony just waiting to be composed. You

00:23:19.559 --> 00:23:22.339
are now the conductor standing before this orchestra

00:23:22.339 --> 00:23:25.099
of truly immense power. So the only question

00:23:25.099 --> 00:23:27.299
left really is what symphony will you create?

00:23:27.680 --> 00:23:29.819
We really hope this deep dive has sparked some

00:23:29.819 --> 00:23:32.140
new ideas for how you can transform the way you

00:23:32.140 --> 00:23:35.079
work. Until next time, keep exploring, keep learning,

00:23:35.220 --> 00:23:37.400
and keep creating. Outiero Music