WEBVTT
1
00:00:00.459 --> 00:00:03.341
So Cloudflare launched something huge around AI agents.
2
00:00:03.421 --> 00:00:05.984
And I think a lot of people haven't really paid attention.
3
00:00:06.184 --> 00:00:13.870
But I think the people that pay attention are going to be able to create businesses that monetize in a completely new way that I think is really interesting.
4
00:00:14.130 --> 00:00:16.732
So I wanted to do an episode breaking down everything.
5
00:00:17.613 --> 00:00:20.454
By the end of this episode, I want you to understand three things.
6
00:00:20.914 --> 00:00:25.915
The first is what is Cloudflare actually doing here in plain English around AI agents?
7
00:00:26.495 --> 00:00:30.275
The second is why does this create a new business model for the Internet?
8
00:00:30.795 --> 00:00:32.576
And the third is startup ideas.
9
00:00:32.776 --> 00:00:35.016
I think you should actually build on top of this.
10
00:00:35.556 --> 00:00:36.857
I want to show you the wedge.
11
00:00:36.957 --> 00:00:43.938
I want to show you who the customer is, what the first version looks like, how I would sell it and how I would become a real company.
12
00:00:44.438 --> 00:00:47.979
The opportunity is way bigger than Cloudflare launched a thing here.
13
00:00:48.419 --> 00:00:53.760
The opportunity is agents are going to need clean, trusted, useful resources to do their jobs.
14
00:00:54.080 --> 00:01:00.142
And the people that figure this out and launch things in the next 18 months are going to do incredible things.
15
00:01:00.402 --> 00:01:04.083
I'm going to show you exactly how everything works in plain English.
16
00:01:04.543 --> 00:01:06.424
I have no affiliation with Cloudflare.
17
00:01:06.444 --> 00:01:07.644
I just want to see you win.
18
00:01:07.944 --> 00:01:08.884
I'm rooting for you.
19
00:01:09.684 --> 00:01:10.605
And I'll see you in there.
20
00:01:10.705 --> 00:01:11.625
Enjoy the episode.
21
00:01:20.106 --> 00:01:21.447
Let's start with the old bargain.
22
00:01:21.487 --> 00:01:25.130
That's the way I think about it, the old bargain of the internet.
23
00:01:25.770 --> 00:01:32.456
For a long time, if you owned a website, you let search engine crawl it because search engines sent you traffic.
24
00:01:32.996 --> 00:01:41.823
And then Google would read your site, Google would index your site, someone would search, your page would hopefully show up, and then a human being would click through.
25
00:01:42.404 --> 00:01:46.127
Once the human being landed on your website, you would monetize that attention.
26
00:01:46.687 --> 00:01:52.531
Maybe you'd show ads, maybe you'd capture their email, maybe you'd sell them a subscription or do some affiliate thing.
27
00:01:53.272 --> 00:01:54.113
That was the trade.
28
00:01:54.233 --> 00:02:02.219
So the crawler got the content, the website got the visitor, the visitor got the answer, and the model funded a massive part of the internet.
29
00:02:02.719 --> 00:02:14.989
Now, AI agents are changing this flow because an AI agent or AI system can read your page, pull the useful answer, give that answer to the user, and the user might not visit your website.
30
00:02:15.349 --> 00:02:19.572
So the content still created value, but the website may lose the visit.
31
00:02:20.052 --> 00:02:21.654
And if the website loses the visit...
32
00:02:22.705 --> 00:02:27.688
it loses the ad impression, the email capture, the affiliate click, all that stuff.
33
00:02:27.808 --> 00:02:28.669
It just loses it.
34
00:02:29.149 --> 00:02:31.411
So that's why a lot of publishers are upset.
35
00:02:31.951 --> 00:02:35.514
But I actually think the publisher conversation is only the first inning.
36
00:02:36.034 --> 00:02:41.578
The bigger idea is that agents are going to use the internet in a way software uses infrastructure.
37
00:02:42.018 --> 00:02:48.042
So they'll request things, they'll call tools, they'll compare products, they'll retrieve data.
38
00:02:48.763 --> 00:02:51.785
By the way, I'm going to get into how Cloudflare plays into all this.
39
00:02:53.235 --> 00:02:57.097
in a second, they'll buy access, they'll take action.
40
00:02:57.757 --> 00:03:02.560
And that basically means that the internet needs a new pricing for machine usage.
41
00:03:03.440 --> 00:03:09.423
Obviously, a human being doesn't want to pay 0.0003 cents to read a page.
42
00:03:09.443 --> 00:03:10.544
That would feel ridiculous.
43
00:03:11.104 --> 00:03:16.287
If I clicked on a recipe and it said, please pay one third of a cent to see the sauce,
44
00:03:16.847 --> 00:03:20.708
I would close a laptop and probably order tacos out of spite.
45
00:03:21.209 --> 00:03:25.090
But a machine doesn't care if the payment is tiny and automatic.
46
00:03:25.730 --> 00:03:30.532
If the data helps the agent complete the job, the agent can pay.
47
00:03:31.192 --> 00:03:32.833
And that's really the mental model.
48
00:03:33.513 --> 00:03:40.576
The human web monetized attention and the agent web monetizes useful resources.
49
00:03:41.750 --> 00:03:43.992
So what is Cloudflare actually doing here?
50
00:03:44.493 --> 00:03:47.536
Well, there's a few pieces I want to go over.
51
00:03:47.656 --> 00:03:54.983
The first is AI crawl control, which gives site owners visibility and control over AI crawlers.
52
00:03:55.463 --> 00:03:57.645
You can see a crawler activity.
53
00:03:58.146 --> 00:04:00.568
You can allow certain crawlers to do things.
54
00:04:00.668 --> 00:04:02.290
And you can block certain crawlers.
55
00:04:02.570 --> 00:04:04.452
You can understand who's accessing your content.
56
00:04:04.852 --> 00:04:09.015
They also announced pay per crawl, and that's the direct monetization piece.
57
00:04:09.035 --> 00:04:11.216
That's the piece that's going to get people paid.
58
00:04:11.837 --> 00:04:16.880
A site owner now can charge AI crawlers when they access content.
59
00:04:17.321 --> 00:04:26.327
The crawler could either present payment intent in the request or receive a 402 payment required response with the pricing.
60
00:04:26.527 --> 00:04:30.588
Then monetization gateway is even a bigger version of that.
61
00:04:30.788 --> 00:04:35.790
So Cloudflare is basically saying this should not just apply to crawlers reading pages.
62
00:04:36.390 --> 00:04:46.874
Any resource behind Cloudflare could have payment rules, a web page, a data set, an API, an MCP tool call, basically a premium endpoint.
63
00:04:47.414 --> 00:04:48.835
a file, a search index.
64
00:04:49.316 --> 00:04:58.162
And the payment rail they're talking about is X402, which uses the HTTP 402 payment required status code.
65
00:04:59.083 --> 00:05:00.764
The agent requests the resource.
66
00:05:01.184 --> 00:05:05.848
The server says, this costs this amount, usually a fraction of a penny.
67
00:05:06.408 --> 00:05:09.631
And the agent pays and retries with proof of payment.
68
00:05:10.131 --> 00:05:14.575
Cloudflare can verify at the edge before the request hits the origin.
69
00:05:15.195 --> 00:05:20.881
Basically, in plain English, CloudFair is trying to make paid access feel like part of the internet itself.
70
00:05:21.282 --> 00:05:23.123
So you don't have this giant checkout flow.
71
00:05:23.143 --> 00:05:25.146
You don't have to create an account first.
72
00:05:25.206 --> 00:05:26.347
There's no sales call.
73
00:05:26.627 --> 00:05:28.709
There's no enterprise procurement stuff.
74
00:05:29.170 --> 00:05:30.571
Basically, there's a request.
75
00:05:31.232 --> 00:05:33.453
that becomes the transaction.
76
00:05:33.533 --> 00:05:35.535
And that's a really, really big idea.
77
00:05:36.675 --> 00:05:45.581
And it's a big idea because the request becomes the transaction and then there's these tiny resources on the internet that could become businesses.
78
00:05:45.881 --> 00:05:46.922
And that's what we're going to get into.
79
00:05:46.962 --> 00:05:48.323
That's why I'm doing this episode.
80
00:05:48.343 --> 00:05:51.505
I'm doing this episode not just because Cloud4 launches thing.
81
00:05:53.046 --> 00:05:56.288
It's interesting because of all the businesses that could be built on top of it.
82
00:05:57.448 --> 00:05:59.649
A dataset could charge per lookup.
83
00:05:59.709 --> 00:06:02.190
An API can charge per successful call.
84
00:06:02.370 --> 00:06:05.871
A research archive could charge per answer.
85
00:06:06.252 --> 00:06:09.053
A product catalog can charge per comparison.
86
00:06:09.533 --> 00:06:10.273
You get the idea.
87
00:06:11.053 --> 00:06:17.036
And someone in the comment section is basically going to say something like, yeah, but agents don't have wallets.
88
00:06:17.536 --> 00:06:21.057
Well, we're moving into a future where agents are going to have wallets.
89
00:06:21.077 --> 00:06:22.418
They're going to have email addresses.
90
00:06:23.625 --> 00:06:32.641
And assuming that's true, and I think there's a good chance that that happens, because it's already starting to happen, they're going to make transactions.
91
00:06:34.575 --> 00:06:38.476
And Cloudflare's AI index points in the same direction.
92
00:06:38.556 --> 00:06:50.020
So they've talked about making websites easier for AI systems to use things like MCPs, LLM.txt, search APIs, and bulk data APIs.
93
00:06:50.380 --> 00:06:54.141
That matters because agents need access, but they also need structure.
94
00:06:54.781 --> 00:06:57.682
So humans can tolerate messy websites.
95
00:06:57.722 --> 00:07:00.663
It might not be a good experience, but they can tolerate it.
96
00:07:02.590 --> 00:07:07.891
We click around, we zoom in, we read the FAQ from 2019, we open a PDF.
97
00:07:07.911 --> 00:07:09.451
We're good at suffering human beings.
98
00:07:09.851 --> 00:07:12.292
But agents need really clean doors.
99
00:07:12.412 --> 00:07:15.952
So they need information in a format they could trust and use.
100
00:07:16.613 --> 00:07:29.295
So the shift is websites become resource layers, content becomes indexes, expertise becomes scalable, data becomes metered, and tools become agent accessible.
101
00:07:29.315 --> 00:07:29.935
Okay.
102
00:07:30.455 --> 00:07:34.158
And that's what I'm paying attention to as a founder because I can build on top of it.
103
00:07:34.979 --> 00:07:40.323
So there's this stack that's forming around this whole new internet and I want to go through it.
104
00:07:40.463 --> 00:07:43.085
It's actually simpler than we think.
105
00:07:43.626 --> 00:07:45.888
So first you have a messy internet.
106
00:07:46.368 --> 00:07:52.072
That is PDFs, pricing pages, old blog posts, YouTube videos, support docs,
107
00:07:53.674 --> 00:07:59.959
comparison sites, websites that look like they were designed when people still said Web 2.0 without any irony.
108
00:08:00.839 --> 00:08:04.422
Then someone cleans that into structured data.
109
00:08:05.123 --> 00:08:16.532
Then someone makes it agent readable through an API, MCP tool, search index, feed, llms.txt, or some other clean access point, but it needs to be cleaned.
110
00:08:17.232 --> 00:08:18.933
Then someone adds payment rules.
111
00:08:19.353 --> 00:08:24.634
Some things are going to remain free because free creates a lot of distribution.
112
00:08:24.654 --> 00:08:28.296
But some things are going to be paid because they're extremely valuable.
113
00:08:28.656 --> 00:08:31.877
And some things are going to be blocked because they should stay private.
114
00:08:32.737 --> 00:08:35.078
And then someone adds trust in analytics.
115
00:08:35.578 --> 00:08:37.059
Like is the data fresh?
116
00:08:37.559 --> 00:08:38.839
Is the source reliable?
117
00:08:39.259 --> 00:08:40.600
Which agents are using it?
118
00:08:41.360 --> 00:08:45.088
which requests are worth money, which resources drive outcomes.
119
00:08:45.829 --> 00:08:51.160
And that stack is what's going to create this whole generation of thousands of companies.
120
00:08:53.200 --> 00:09:04.745
When you're looking for ideas here, the question really is, what resource does an agent need badly enough and often enough and reliably enough to pay for?
121
00:09:05.065 --> 00:09:07.586
That's basically the biggest question of all of this.
122
00:09:07.626 --> 00:09:08.887
That's what I'm asking myself.
123
00:09:09.787 --> 00:09:16.770
And let's get into the three ideas that'll get your creative juices flowing around building businesses on top of that.
124
00:09:16.870 --> 00:09:20.191
But before I get into it, I saw a lot of chatter on X,
125
00:09:22.092 --> 00:09:25.644
Levels.io, who's an incredible indie hacker, was talking about how
126
00:09:26.565 --> 00:09:32.707
you know, how hard it is now to build a business and how a lot of people's traffic and revenue has gone down.
127
00:09:33.088 --> 00:09:45.352
I think in large part, you know, because of AI overviews and how basically a lot of software is just not, it's not useful anymore because people are vibe coding it themselves and stuff like that.
128
00:09:45.813 --> 00:09:50.154
My take is it's an incredible time to be building.
129
00:09:50.194 --> 00:09:51.655
In fact, it's the best time ever.
130
00:09:52.295 --> 00:09:56.757
Now, I think the types of businesses to create are businesses like this.
131
00:09:56.857 --> 00:09:59.578
They're not like little tools that could be vibe-coded.
132
00:09:59.758 --> 00:10:08.042
These are startup ideas that could be profitable on day one, that have some moat, that are...
133
00:10:09.102 --> 00:10:13.247
are reinvented for an internet that's AI native and agent ready.
134
00:10:13.307 --> 00:10:22.217
So I don't want you to go on X and just be devastated that there's no entrepreneurship is done, RIP startups.
135
00:10:22.337 --> 00:10:26.101
No, the world is moving, the internet is changing, you have to move with it.
136
00:10:26.181 --> 00:10:28.824
And here's three startup ideas that move with it.
137
00:10:29.084 --> 00:10:29.425
Let's go.
138
00:10:30.102 --> 00:10:33.945
So the first startup idea is a niche data refinery.
139
00:10:34.325 --> 00:10:38.408
This is probably the one I would start with because it's the most practical.
140
00:10:38.688 --> 00:10:39.668
The idea is really simple.
141
00:10:39.728 --> 00:10:49.095
So pick one niche where valuable information is messy, fragmented, changing, annoying to collect,
142
00:10:49.935 --> 00:10:53.938
then turn that information into clean fuel for agents.
143
00:10:54.638 --> 00:10:59.201
I'm calling it a data refinery because the raw material already exists.
144
00:10:59.441 --> 00:11:01.362
So the internet already has that data.
145
00:11:02.083 --> 00:11:07.226
It's sitting in Google Maps, it's sitting in job posts, it's sitting in reviews and local directories.
146
00:11:08.046 --> 00:11:10.289
PDFs, it sits in pricing pages.
147
00:11:10.569 --> 00:11:14.733
It's in all these places and your job is going to be refining it.
148
00:11:15.314 --> 00:11:17.757
So let's make it concrete.
149
00:11:17.797 --> 00:11:20.279
I'll give you an example so it drives the point home.
150
00:11:21.253 --> 00:11:30.180
So imagine you pick med spas because, you know, they're growing a lot in the U.S. And that's the niche you want to pick.
151
00:11:30.240 --> 00:11:30.501
Great.
152
00:11:30.801 --> 00:11:47.875
A med spa owner wants to know what competitors are charging, what treatments they offer, what reviews complain about, which clinics are hiring, which services are trending, what offers are working, and how the local market is changing.
153
00:11:48.355 --> 00:11:51.296
So that information today lives everywhere, really.
154
00:11:51.916 --> 00:12:02.360
It lives on Google reviews, on competitors' websites, on Instagram, on job posts, meta ad libraries, in the owner's head, in employee's heads.
155
00:12:03.620 --> 00:12:09.362
An agent can do incredible work for a med spa owner if it had that information cleanly.
156
00:12:09.482 --> 00:12:10.723
It would be huge.
157
00:12:11.263 --> 00:12:19.092
It could say something like your Botox pricing is above the local median, but your reviews do not support premium positioning yet.
158
00:12:19.332 --> 00:12:20.513
That would be super valuable.
159
00:12:21.174 --> 00:12:28.062
It could say three competitors near you started promoting exosome treatments in the last 60 days.
160
00:12:28.923 --> 00:12:30.324
Super valuable information to know.
161
00:12:31.125 --> 00:12:36.370
It could say the most common complaint in local reviews is you're confusing pricing.
162
00:12:36.750 --> 00:12:38.552
So your offer should lead with simplicity.
163
00:12:39.093 --> 00:12:39.733
Great to know.
164
00:12:40.594 --> 00:12:47.421
It could say too fast growing competitors are hiring injectors, which probably means they're expanding capacity.
165
00:12:47.621 --> 00:12:48.322
Really good to know.
166
00:12:48.362 --> 00:12:49.783
So all useful information.
167
00:12:50.323 --> 00:12:53.747
And more importantly, the usefulness comes from the data.
168
00:12:54.187 --> 00:12:56.108
not from a generic AI wrapper.
169
00:12:56.668 --> 00:13:01.569
So here's how I would build a wedge into this space, into this business.
170
00:13:01.649 --> 00:13:03.910
I would pick one niche and I would pick one city.
171
00:13:04.410 --> 00:13:06.911
I would track 100 businesses, not more.
172
00:13:06.931 --> 00:13:09.372
I would do it manually at first.
173
00:13:10.252 --> 00:13:27.657
I would create a spreadsheet with your business name, website, services, prices, review count, review rating, top review complaints, Instagram links, recent posts, visible ad changes, hiring signals, and booking flow.
174
00:13:28.137 --> 00:13:30.358
Then I would create 10 outputs from that data.
175
00:13:30.718 --> 00:13:36.360
It could be like a local pricing map, a competitor gap report, a list of offer ideas,
176
00:13:37.380 --> 00:13:45.349
A services to ad recommendation, a review complaint summary, a hiring signal report, a monthly market movement report.
177
00:13:46.150 --> 00:13:48.312
Now you have something to sell here.
178
00:13:49.113 --> 00:13:52.597
And the first customer actually might not be a med spa owner.
179
00:13:53.197 --> 00:13:56.220
And that's a part I think people are missing a little bit.
180
00:13:56.881 --> 00:14:01.805
In the early days, your first customer is often the person already selling into the niche.
181
00:14:02.005 --> 00:14:17.476
So instead of trying to sell agent readable competitive analysis to a med spa owner, which is a phrase that is just confusing to the average person, sell it to med spa marketing agencies, consultants, freelancers, clients.
182
00:14:18.056 --> 00:14:20.777
software companies, and even AI implementation people.
183
00:14:21.237 --> 00:14:31.861
You could say, I built local market intelligence for med spas, and you can use it to create better audits, better offers, better landing pages, and better campaigns for your clients.
184
00:14:31.921 --> 00:14:33.802
They're going to be able to charge more.
185
00:14:34.962 --> 00:14:39.164
So what you've built is super valuable.
186
00:14:39.524 --> 00:14:41.005
And that's just a way easier sale.
187
00:14:42.385 --> 00:14:48.670
A MedSpa marketing agency might sell a client for like $5K a month as a growth package.
188
00:14:49.170 --> 00:14:57.216
And if your data helps them close one more client or improve their work, they can pay you $300, $500, $800 a month.
189
00:14:57.576 --> 00:15:01.178
And over time, you turn your spreadsheet into a real product.
190
00:15:01.238 --> 00:15:07.523
So first it's a report, then it's a dashboard, then it becomes an API, then maybe an MCP tool.
191
00:15:08.023 --> 00:15:12.006
then agents can pay per lookup or per report when the rails are ready.
192
00:15:12.046 --> 00:15:15.589
So you're doing basically a crawl, walk, run strategy.
193
00:15:16.130 --> 00:15:19.352
I used med spas as an example, but it doesn't need to be med spas.
194
00:15:19.893 --> 00:15:23.376
I live in Miami and there just so happens to be a lot of med spas here.
195
00:15:24.136 --> 00:15:26.098
But you can do it for roofing.
196
00:15:26.318 --> 00:15:30.922
You can do it for, let's say you were to do it for roofing.
197
00:15:31.562 --> 00:15:34.204
You would track storm events maybe.
198
00:15:34.224 --> 00:15:41.549
You would track permit data, insurance signals, local reviews, competitor offers, and ad angles.
199
00:15:41.569 --> 00:15:43.370
You can do it for real estate investing.
200
00:15:43.730 --> 00:15:45.011
You track zoning changes.
201
00:15:45.031 --> 00:15:45.951
You track permits.
202
00:15:45.971 --> 00:15:47.332
You track ownership records.
203
00:15:47.352 --> 00:15:48.473
You track rent comps.
204
00:15:49.013 --> 00:15:50.654
You track tax delinquencies.
205
00:15:50.715 --> 00:15:52.256
And you track insurance shifts.
206
00:15:53.096 --> 00:15:54.557
You can do it for e-commerce.
207
00:15:54.637 --> 00:16:03.924
You can track competitor SKUs, pricing changes, review complaints, influencer rates, UGC hooks, Shopify apps, shipping promises.
208
00:16:04.484 --> 00:16:05.485
You can do law firms.
209
00:16:05.945 --> 00:16:10.869
You can track local competitors, practice area positioning, ad copy, reviews, intake.
210
00:16:10.889 --> 00:16:11.669
You get the idea.
211
00:16:12.310 --> 00:16:14.712
The filter is pretty simple.
212
00:16:15.152 --> 00:16:16.513
The data should be valuable.
213
00:16:17.295 --> 00:16:22.819
The data should be repeatable, changing, fragmented, and annoying.
214
00:16:23.100 --> 00:16:28.824
And valuable data basically means better decisions make or save money.
215
00:16:29.645 --> 00:16:30.746
What does repeated mean?
216
00:16:30.786 --> 00:16:33.388
Well, repeated means the customer needs it again and again.
217
00:16:33.448 --> 00:16:34.669
It's not like a one-time thing.
218
00:16:35.129 --> 00:16:36.069
What does changing mean?
219
00:16:36.109 --> 00:16:39.311
It means that freshness matters in the data.
220
00:16:39.871 --> 00:16:42.032
What does fragmented mean?
221
00:16:42.132 --> 00:16:45.073
It means that one person can easily collect it.
222
00:16:45.993 --> 00:16:46.933
What does annoying mean?
223
00:16:47.173 --> 00:16:48.994
That basically means there's some margin there.
224
00:16:49.794 --> 00:16:51.635
And that's my first startup idea.
225
00:16:52.455 --> 00:16:57.837
It's basically one idea that can give you a lot of ideas, depending on what niche that you have some unfair advantage in.
226
00:16:59.918 --> 00:17:04.685
Take one niche's messy internet, turn it into clean fuel for agents.
227
00:17:04.945 --> 00:17:07.308
I don't want you to forget that phrase, clean fuel for agents.
228
00:17:08.149 --> 00:17:09.731
Let's get on to the next startup idea.
229
00:17:10.597 --> 00:17:14.779
So startup idea number two is agent readiness for businesses.
230
00:17:15.359 --> 00:17:25.783
This is like SEO for the agent internet, but I wanted to find it very specifically because AI SEO is already becoming a buzzy, fuzzy phrase.
231
00:17:26.303 --> 00:17:32.866
The real business is helping companies become easy for agents to understand, trust, compare, and recommend.
232
00:17:33.446 --> 00:17:35.107
Think about a B2B SaaS company.
233
00:17:35.547 --> 00:17:37.648
A human buyer lands on the homepage,
234
00:17:38.663 --> 00:17:50.232
reads the hero image, clicks around, looks at the pricing page, reads some docs, maybe books a demo, watches a case study, and maybe asks a friend about this particular product.
235
00:17:51.152 --> 00:17:54.055
Agents are compressing that whole process.
236
00:17:54.555 --> 00:17:58.358
If someone asks their AI assistant, find me the best payroll provider,
237
00:17:59.376 --> 00:18:04.260
For a 15-person company in California, the agent has to understand the market.
238
00:18:04.680 --> 00:18:07.962
So the agent is basically asking, who's this product for?
239
00:18:08.042 --> 00:18:09.003
What does it cost?
240
00:18:09.383 --> 00:18:10.364
What does it replace?
241
00:18:10.404 --> 00:18:11.605
What sort of integrations?
242
00:18:11.645 --> 00:18:12.666
What are the risks?
243
00:18:13.346 --> 00:18:15.068
What does implementation look like?
244
00:18:15.148 --> 00:18:16.209
What do customers say?
245
00:18:16.549 --> 00:18:18.070
How does it compare to alternatives?
246
00:18:18.430 --> 00:18:21.833
Most websites actually make this harder than it needs to be.
247
00:18:22.113 --> 00:18:23.194
They hide pricing.
248
00:18:23.234 --> 00:18:24.435
They bury docs.
249
00:18:25.475 --> 00:18:27.576
There's just not a lot of information there.
250
00:18:27.896 --> 00:18:33.278
They're not constantly updating their website, so they end up having stale websites, stale comparison pages.
251
00:18:34.379 --> 00:18:38.480
Some of the most important information is actually in PDFs.
252
00:18:38.680 --> 00:18:42.502
And finding policies is extremely hard to find in general.
253
00:18:43.624 --> 00:18:45.825
And there's just a lot of marketing speak.
254
00:18:45.845 --> 00:18:49.467
So you'll see things like unlocking operational excellence.
255
00:18:50.147 --> 00:18:52.789
So there's a lot of just like foggy copywriting.
256
00:18:53.589 --> 00:18:58.392
So the startup idea is basically to make these websites agent readable.
257
00:18:58.712 --> 00:19:00.393
Here's the wedge I would use.
258
00:19:01.113 --> 00:19:02.534
Start with a paid audit.
259
00:19:02.634 --> 00:19:07.135
And I've done episodes on the podcast with Corey Gannam around some of these things.
260
00:19:08.315 --> 00:19:16.698
You can go deeper into with Vaz, where we talked about what is a forward deploy engineer, where we've talked about paid audits.
261
00:19:17.098 --> 00:19:18.799
But I want you to pick one vertical.
262
00:19:19.339 --> 00:19:29.182
B2B SaaS is obvious, but you can do Shopify apps, law firms, health care clinics, financial advisors, insurance brokers, home services, whatever you know.
263
00:19:31.624 --> 00:19:36.488
20 to 50 buyer intent prompts across major AI tools.
264
00:19:36.929 --> 00:19:40.872
So you're going to ask questions like, what is the best software for this use case?
265
00:19:41.213 --> 00:19:43.615
Compare this company to top alternatives.
266
00:19:44.055 --> 00:19:45.817
What does this company cost?
267
00:19:46.237 --> 00:19:47.298
Who is this product best for?
268
00:19:47.338 --> 00:19:48.660
What are the risks of choosing this vendor?
269
00:19:48.680 --> 00:19:51.022
What are the risks of choosing this vendor?
270
00:19:51.442 --> 00:19:55.005
Would you recommend this product for a 20-person company?
271
00:19:55.505 --> 00:19:57.787
What integrations does it support?
272
00:19:58.287 --> 00:20:00.569
Then you show the company the answers.
273
00:20:00.969 --> 00:20:07.734
And this is really the sales moment because you might show a founder, when buyers ask AI for your category, you do not show up.
274
00:20:08.174 --> 00:20:10.756
Or you show up, but AI gets your pricing wrong.
275
00:20:11.096 --> 00:20:15.540
On your website, it's $20 a month, but AI for some reason is getting $8 a month.
276
00:20:16.338 --> 00:20:20.241
Or the AI recommends your competitor because their docs are cleaner.
277
00:20:20.981 --> 00:20:26.005
Or your website has the answer, but it's buried in a PDF from 2002.
278
00:20:26.385 --> 00:20:30.248
That gets their attention, especially if you're doing cold email and stuff like that.
279
00:20:30.588 --> 00:20:31.629
Then you sell the fix.
280
00:20:32.169 --> 00:20:35.172
The fix is an agent-readable source of truth.
281
00:20:35.752 --> 00:20:42.217
So that might include a cleanllms.txt file, which is a file that helps...
282
00:20:45.859 --> 00:20:47.581
LLMs crawl your website.
283
00:20:48.581 --> 00:20:50.343
A better documentation structure.
284
00:20:50.363 --> 00:20:54.006
A pricing page agents can parse.
285
00:20:54.046 --> 00:20:56.427
Comparison pages that are honest and specific.
286
00:20:57.188 --> 00:21:02.272
Use cases pages written in just plain, simple language.
287
00:21:02.532 --> 00:21:04.454
Customer proof organized by segment.
288
00:21:04.854 --> 00:21:14.087
Structured FAQs around real buyer questions, schema markup, a product feed, a change log, a lightweight MCP server or search endpoint.
289
00:21:14.247 --> 00:21:19.374
If the company has enough useful content, if they don't, then probably not that.
290
00:21:19.995 --> 00:21:23.458
And then the recurring product is the measurement loop.
291
00:21:23.698 --> 00:21:38.450
So every month you rerun the prompts, you see what's changed, you see whether AI answers are more accurate, whether the company appears more often, whether the competitor comparisons improve or get worse, and you see where the website needs more structured proof.
292
00:21:38.870 --> 00:21:41.693
And this could start off as an easy services business.
293
00:21:42.233 --> 00:21:46.794
So you can charge something like $3,000 to $10,000 for the audit and cleanup.
294
00:21:47.215 --> 00:21:52.256
And then for larger B2B companies, it could be something like $10,000, $15,000, $20,000.
295
00:21:52.756 --> 00:21:57.698
And after you do 10 clients in the same niche, you're going to see a bunch of repeated work.
296
00:21:57.718 --> 00:22:03.560
And this is why I love starting with services businesses and then productizing it with software as you go.
297
00:22:04.120 --> 00:22:06.964
You're going to learn things like the same docs are missing.
298
00:22:07.405 --> 00:22:09.507
The same pricing pages are unclear.
299
00:22:09.568 --> 00:22:10.990
The same questions matter.
300
00:22:11.270 --> 00:22:13.413
The same structured files need to be created.
301
00:22:13.693 --> 00:22:16.577
The same monthly report needs to be delivered.
302
00:22:16.978 --> 00:22:19.141
This is when you turn it into software.
303
00:22:19.541 --> 00:22:20.222
And the way...
304
00:22:20.883 --> 00:22:23.145
To sell this is very simple.
305
00:22:23.185 --> 00:22:26.288
You're not selling the future, you're selling the screenshot.
306
00:22:26.688 --> 00:22:29.891
You show them what AI says about their company today.
307
00:22:30.411 --> 00:22:32.333
And that becomes the whole sales deck.
308
00:22:32.433 --> 00:22:38.918
For local businesses, this eventually becomes let AI assistants book appointments with you.
309
00:22:39.439 --> 00:22:46.025
For e-commerce, it becomes something like make your product catalog easy for shopping agents to compare and buy.
310
00:22:47.179 --> 00:22:52.602
For B2B SaaS, it becomes make your product easy for procurement agents to evaluate.
311
00:22:53.082 --> 00:22:58.745
And for publishers, it becomes make your archive easy for AI systems to understand and license.
312
00:23:01.023 --> 00:23:07.792
I think that there's going to be venture-backed companies, and you're already starting to see it happen, doing these horizontal products that do this.
313
00:23:07.852 --> 00:23:10.536
The opportunity here is to go extremely vertical.
314
00:23:10.556 --> 00:23:14.141
Obviously, B2B SaaS is too big, right?
315
00:23:14.241 --> 00:23:18.366
You want to pick a specific niche, go deep into that.
316
00:23:19.467 --> 00:23:26.589
But I think this is a wonderful business that's cash flow on day one that could productize that could eventually sell at some point in the future.
317
00:23:27.489 --> 00:23:33.170
Basically help businesses become easy for agents to understand trust compare and recommend.
318
00:23:33.590 --> 00:23:41.692
There's a ton of demand from businesses right now for this because they're feeling this pain.
319
00:23:42.652 --> 00:23:43.512
So you just solve it.
320
00:23:43.992 --> 00:23:50.076
Last but not least, startup idea number three is turning expert archive into agent tools.
321
00:23:50.096 --> 00:23:51.417
I'll explain what I mean by that.
322
00:23:52.417 --> 00:24:07.186
This one is probably the most fun for creators, media companies, analysts, consultants, researchers, people like that that are sitting on years of valuable content or can get their hands on valuable content.
323
00:24:07.586 --> 00:24:09.628
What do I mean by valuable content?
324
00:24:10.408 --> 00:24:18.375
I'm talking about things like YouTube videos, podcasts, newsletters, templates, things like that, community posts.
325
00:24:19.395 --> 00:24:29.123
Right now, most of that content makes money through ads, sponsorships, sometimes subscriptions, sometimes communities, sometimes consulting.
326
00:24:29.683 --> 00:24:32.886
But in the agent internet, that archive could become a tool.
327
00:24:33.426 --> 00:24:35.108
So imagine a founder agent
328
00:24:35.708 --> 00:24:39.932
That can access a startup's expert archive and critique your idea.
329
00:24:40.372 --> 00:24:46.518
Imagine a sales agent that can use a specific sales trainer's framework to rewrite your cold email.
330
00:24:46.958 --> 00:24:53.364
Imagine a fitness agent that can use a coach's training philosophy to build a personalized plan.
331
00:24:54.285 --> 00:24:57.086
The startup is archive to API.
332
00:24:57.646 --> 00:25:02.308
You take someone's expertise and you package it so agents can use it.
333
00:25:02.909 --> 00:25:04.269
The key is to start with one job.
334
00:25:04.629 --> 00:25:08.971
Don't go to a creator and say, we're going to turn your whole brain into AI.
335
00:25:09.411 --> 00:25:15.574
That honestly sounds a little creepy, a little vague, and honestly like a SaaS landing page that should be illegal.
336
00:25:16.074 --> 00:25:17.795
You got to say something that's specific.
337
00:25:17.895 --> 00:25:19.456
So what do I mean by that?
338
00:25:19.856 --> 00:25:22.277
Something like you have 300 videos about sales.
339
00:25:22.477 --> 00:25:27.700
We're going to turn them into a tool your audience could use to improve cold emails.
340
00:25:28.260 --> 00:25:34.364
Or hey, you've got 500 episodes about startups, podcast episodes about startups.
341
00:25:34.404 --> 00:25:37.005
We're going to turn them into a startup idea feedback tool.
342
00:25:37.285 --> 00:25:38.586
Or maybe it's to a designer.
343
00:25:38.786 --> 00:25:41.508
It's like you have a decade of these design teardowns.
344
00:25:41.548 --> 00:25:44.730
We're going to turn them into a landing page critique tool.
345
00:25:45.470 --> 00:25:49.391
You've got this one archive with one painful job, one workflow.
346
00:25:49.891 --> 00:25:52.752
And here's how I would build this startup idea.
347
00:25:52.852 --> 00:25:57.454
So I would pick an expert with a deep archive and a specific audience.
348
00:25:57.834 --> 00:26:05.156
Me, in particular, I'd pick someone who is more in the B2B space, but it could work for B2C.
349
00:26:06.583 --> 00:26:15.674
Specific matters just because a general business creator, for example, is just going to be harder to do.
350
00:26:16.234 --> 00:26:24.784
But maybe a creator known for cold email or Shopify growth or local business acquisitions, tax strategy, fitness programming,
351
00:26:25.345 --> 00:26:27.207
Or like what we talked about, design teardowns.
352
00:26:27.227 --> 00:26:28.868
That would work really well.
353
00:26:29.509 --> 00:26:31.831
Second, you want to go and collect the archive.
354
00:26:31.851 --> 00:26:39.499
So you're going to want to go and transcribe the content, if that's videos, if it's podcasts, pull the newsletters, clean the docs.
355
00:26:39.759 --> 00:26:41.260
And then third, you're going to do tagging.
356
00:26:41.280 --> 00:26:48.988
So you're going to tag the archive by job, by topic, by audience, by example, by framework, and by outcome.
357
00:26:49.328 --> 00:26:51.511
A lot of people get lazy at this part.
358
00:26:51.531 --> 00:26:54.776
They throw everything into a vector database and then they just call it a day.
359
00:26:55.137 --> 00:26:57.841
That usually gives you a search box with confidence
360
00:26:58.883 --> 00:27:00.985
But a real product needs structure.
361
00:27:01.185 --> 00:27:12.256
So if it's a sales archive, you want to tag by prospecting, by subject line, by offer, by objection, by follow-up, personalization, deliverability, and close.
362
00:27:12.816 --> 00:27:21.064
If it's a startup archive, you want to tag by the idea, by the market, the wedge, the distribution, the pricing, the MVP, the community, the mode, and examples.
363
00:27:22.467 --> 00:27:24.389
So you're very specific, right?
364
00:27:24.549 --> 00:27:31.556
I think the best tools here are going to be very specific and that's going to help get the best outcome for people ultimately.
365
00:27:33.221 --> 00:27:36.303
Fourth, what you want to do is build one useful workflow.
366
00:27:36.444 --> 00:27:41.568
So for a sales expert, that workflow could just be paste your cold email.
367
00:27:42.068 --> 00:27:47.013
The agents are going to critique it using that expert's principles.
368
00:27:47.173 --> 00:27:48.994
Maybe it's Alex Hermosi, let's say.
369
00:27:49.054 --> 00:27:50.836
Alex Hermosi is going to go and critique it.
370
00:27:51.416 --> 00:27:52.998
It cites the source lessons.
371
00:27:53.138 --> 00:27:54.299
It rewrites the email.
372
00:27:54.619 --> 00:27:55.700
It gives you a score.
373
00:27:56.301 --> 00:27:57.882
It gives you one test to run.
374
00:27:58.182 --> 00:27:59.724
And that's the product.
375
00:28:01.225 --> 00:28:14.072
For a startup expert, the workflow could be paste your idea, the agent gives you the wedge, it gives you the customer, it writes the first offer, it suggests the first distribution channel, and it tells you what to validate this week.
376
00:28:14.092 --> 00:28:17.314
That's kind of what we're doing with ideabrowser.com.
377
00:28:17.414 --> 00:28:18.055
You know, you...
378
00:28:19.118 --> 00:28:21.780
one of our most popular features is adding the MCP.
379
00:28:22.240 --> 00:28:27.864
And it's so good because it takes your LLM and just makes it better, right?
380
00:28:27.904 --> 00:28:31.846
And it's got all this data that we've cleaned to help you do that.
381
00:28:32.106 --> 00:28:34.508
So I'm practicing what I'm preaching here.
382
00:28:36.315 --> 00:28:38.757
For a real estate expert, what could that be?
383
00:28:38.817 --> 00:28:50.087
Well, it could be something like you pace the deal, the agent checks the assumptions, it identifies the risk, it compares it to the expert criteria, and it tells you whether to ask the broker.
384
00:28:50.227 --> 00:28:51.788
That would be super, super useful.
385
00:28:53.209 --> 00:29:00.093
What's cool about doing something like this, the creator already has the distribution, so you don't have to worry about the marketing, the customer acquisition.
386
00:29:00.414 --> 00:29:04.956
They've got all that trust already built in, but the audience wants the expertise.
387
00:29:05.717 --> 00:29:10.039
And the creator might not want to do consulting with everyone, right?
388
00:29:10.079 --> 00:29:12.521
So this democratizes that.
389
00:29:12.941 --> 00:29:14.702
So you can charge a lot cheaper.
390
00:29:14.782 --> 00:29:16.583
You could be something like $19 a month or $50 a month
391
00:29:18.144 --> 00:29:29.411
You can bundle into a paid community, or you can even make it a lead magnet for the consulting, or you can license it to agencies or software companies, that sort of thing.
392
00:29:30.992 --> 00:29:44.280
This is really where the Cloudflare-style monetization becomes interesting, because if the archive becomes this resource and agents can pay per request, the creator gets paid when the knowledge is used, the builder gets this trusted resource
393
00:29:44.939 --> 00:29:51.985
expert layer, the person consuming it gets this expert layer and gets better output.
394
00:29:53.006 --> 00:30:00.593
That's way better than hoping someone watches a pre-roll ad before a 47-minute interview from seven years ago.
395
00:30:01.233 --> 00:30:09.901
I think the biggest mistake in this category that I've noticed is I've seen a lot of chat with an expert products, but that's too broad.
396
00:30:09.961 --> 00:30:11.002
So I think the specific
397
00:30:13.223 --> 00:30:17.127
The specific use case is way more interesting and it's way more outcome based.
398
00:30:17.167 --> 00:30:20.890
It's not chat with the sales person or chat with the sales creator.
399
00:30:20.970 --> 00:30:24.173
It's rewrite the cold email using the sales system.
400
00:30:25.913 --> 00:30:36.121
So if you can turn expert archives into job-specific agent tools, I think that could be really cool and someone should do it.
401
00:30:36.701 --> 00:30:38.623
So those are the three startup ideas.
402
00:30:39.564 --> 00:30:41.185
Number one, the niche data refinery.
403
00:30:41.245 --> 00:30:43.407
Number two, the agent readiness for business.
404
00:30:43.427 --> 00:30:46.970
And number three, the expert archives turned into agent tools.
405
00:30:48.030 --> 00:30:54.894
What connects all three of these ideas is agents need clean, trusted, and useful resources to do really good work.
406
00:30:54.974 --> 00:31:02.718
So that resource can be data, but it could be structure, it could be access, it could be expert knowledge, it could be a tool, it could be a payment rule.
407
00:31:03.398 --> 00:31:12.586
The Cloudflare news matters because building part of the access and payment layer, that's what they're building, and that's a huge thing.
408
00:31:13.046 --> 00:31:21.634
But you don't have to wait for Cloudflare's whole new agent internet to mature in order to start building out some of this stuff.
409
00:31:22.154 --> 00:31:26.978
So you can start by building the manual version first.
410
00:31:26.998 --> 00:31:28.980
You can sell the human version now.
411
00:31:29.040 --> 00:31:30.361
You can build the data now.
412
00:31:31.822 --> 00:31:33.422
and you can package this up now.
413
00:31:33.642 --> 00:31:42.764
So as agents become more capable and agent payments become more and more common, you're not scrambling to start building this stuff.
414
00:31:43.205 --> 00:31:47.025
This is the early, you're early, you know?
415
00:31:47.185 --> 00:31:48.846
Not many people are talking about this.
416
00:31:49.566 --> 00:31:56.027
And if you're looking for ideas in the space, the questions you should ask yourself are, what decision is expensive?
417
00:31:56.487 --> 00:31:58.008
What information is messy?
418
00:31:59.328 --> 00:32:00.689
What changes often?
419
00:32:01.289 --> 00:32:03.550
Who already pays for help?
420
00:32:03.710 --> 00:32:07.491
And what would an agent need to do to do the job better?
421
00:32:07.612 --> 00:32:11.353
That's basically the map to be thinking about and the questions to be thinking about.
422
00:32:11.873 --> 00:32:20.217
So you can spend the next 6, 12, 18 months building up this data, curating the data before it becomes hyper-competitive.
423
00:32:21.458 --> 00:32:26.902
I truly believe the internet is shifting from pages human visit to resources agents use.
424
00:32:27.302 --> 00:32:28.903
And I don't think it's crazy to say that.
425
00:32:29.464 --> 00:32:32.106
And if you made it this far, you probably agree too.
426
00:32:33.827 --> 00:32:42.254
The best opportunities look really small right now because the agent internet is small relative to where it's going to be.
427
00:32:43.275 --> 00:32:47.478
But I think that's how a lot of the biggest businesses started, right?
428
00:32:47.618 --> 00:32:49.760
You know, this is like building an app.
429
00:32:50.305 --> 00:32:54.408
when the App Store came out in 2009.
430
00:32:54.548 --> 00:33:05.395
This cohort of businesses starting in this era, 2026, 2027, I think there's going to be just an incredible amount of businesses created here.
431
00:33:05.815 --> 00:33:08.577
It's very clear that the agent internet is the next wave.
432
00:33:09.518 --> 00:33:17.223
So when you see Cloudflare talking about AI crawlers, X402, paid access, MCP tools, and monetization,
433
00:33:18.043 --> 00:33:20.466
Don't just think, that's a little interesting thing.
434
00:33:20.526 --> 00:33:23.810
Or don't just think, oh, publishers can now charge bots now.
435
00:33:24.391 --> 00:33:25.572
It's way bigger than that.
436
00:33:25.833 --> 00:33:27.294
Agents are becoming buyers.
437
00:33:27.435 --> 00:33:29.077
Websites are becoming resources.
438
00:33:29.757 --> 00:33:31.459
And the next great internet businesses...
439
00:33:34.403 --> 00:33:39.804
It might be these tiny paid doors that agents walk through all day that you're going to own.
440
00:33:40.065 --> 00:33:43.866
It's an asset and I can't wait to see what you build, what you do.
441
00:33:44.626 --> 00:33:50.947
If this has been helpful or got your creative juices flowing, shoot me over a comment, like and subscribe for more of this in your feed.
442
00:33:51.368 --> 00:33:53.088
I read every single comment, by the way.
443
00:33:54.268 --> 00:33:55.449
And I'll see you on the next time.
444
00:33:55.889 --> 00:33:58.229
Thank you for listening to the Startup Ideas podcast.
445
00:33:58.770 --> 00:33:59.790
Have a creative day.
446
00:34:00.510 --> 00:34:01.150
I'll see you next.
447
00:34:01.470 --> 00:34:02.991
And wow, I already said I'd see you next time.
448
00:34:04.716 --> 00:34:05.758
I'm missing you already.
449
00:34:06.319 --> 00:34:08.302
I can't wait to the next episode.
450
00:34:08.602 --> 00:34:08.843
See ya.
1
00:00:00.459 --> 00:00:03.341
So Cloudflare launched something huge around AI agents.
2
00:00:03.421 --> 00:00:05.984
And I think a lot of people haven't really paid attention.
3
00:00:06.184 --> 00:00:13.870
But I think the people that pay attention are going to be able to create businesses that monetize in a completely new way that I think is really interesting.
4
00:00:14.130 --> 00:00:16.732
So I wanted to do an episode breaking down everything.
5
00:00:17.613 --> 00:00:20.454
By the end of this episode, I want you to understand three things.
6
00:00:20.914 --> 00:00:25.915
The first is what is Cloudflare actually doing here in plain English around AI agents?
7
00:00:26.495 --> 00:00:30.275
The second is why does this create a new business model for the Internet?
8
00:00:30.795 --> 00:00:32.576
And the third is startup ideas.
9
00:00:32.776 --> 00:00:35.016
I think you should actually build on top of this.
10
00:00:35.556 --> 00:00:36.857
I want to show you the wedge.
11
00:00:36.957 --> 00:00:43.938
I want to show you who the customer is, what the first version looks like, how I would sell it and how I would become a real company.
12
00:00:44.438 --> 00:00:47.979
The opportunity is way bigger than Cloudflare launched a thing here.
13
00:00:48.419 --> 00:00:53.760
The opportunity is agents are going to need clean, trusted, useful resources to do their jobs.
14
00:00:54.080 --> 00:01:00.142
And the people that figure this out and launch things in the next 18 months are going to do incredible things.
15
00:01:00.402 --> 00:01:04.083
I'm going to show you exactly how everything works in plain English.
16
00:01:04.543 --> 00:01:06.424
I have no affiliation with Cloudflare.
17
00:01:06.444 --> 00:01:07.644
I just want to see you win.
18
00:01:07.944 --> 00:01:08.884
I'm rooting for you.
19
00:01:09.684 --> 00:01:10.605
And I'll see you in there.
20
00:01:10.705 --> 00:01:11.625
Enjoy the episode.
21
00:01:20.106 --> 00:01:21.447
Let's start with the old bargain.
22
00:01:21.487 --> 00:01:25.130
That's the way I think about it, the old bargain of the internet.
23
00:01:25.770 --> 00:01:32.456
For a long time, if you owned a website, you let search engine crawl it because search engines sent you traffic.
24
00:01:32.996 --> 00:01:41.823
And then Google would read your site, Google would index your site, someone would search, your page would hopefully show up, and then a human being would click through.
25
00:01:42.404 --> 00:01:46.127
Once the human being landed on your website, you would monetize that attention.
26
00:01:46.687 --> 00:01:52.531
Maybe you'd show ads, maybe you'd capture their email, maybe you'd sell them a subscription or do some affiliate thing.
27
00:01:53.272 --> 00:01:54.113
That was the trade.
28
00:01:54.233 --> 00:02:02.219
So the crawler got the content, the website got the visitor, the visitor got the answer, and the model funded a massive part of the internet.
29
00:02:02.719 --> 00:02:14.989
Now, AI agents are changing this flow because an AI agent or AI system can read your page, pull the useful answer, give that answer to the user, and the user might not visit your website.
30
00:02:15.349 --> 00:02:19.572
So the content still created value, but the website may lose the visit.
31
00:02:20.052 --> 00:02:21.654
And if the website loses the visit...
32
00:02:22.705 --> 00:02:27.688
it loses the ad impression, the email capture, the affiliate click, all that stuff.
33
00:02:27.808 --> 00:02:28.669
It just loses it.
34
00:02:29.149 --> 00:02:31.411
So that's why a lot of publishers are upset.
35
00:02:31.951 --> 00:02:35.514
But I actually think the publisher conversation is only the first inning.
36
00:02:36.034 --> 00:02:41.578
The bigger idea is that agents are going to use the internet in a way software uses infrastructure.
37
00:02:42.018 --> 00:02:48.042
So they'll request things, they'll call tools, they'll compare products, they'll retrieve data.
38
00:02:48.763 --> 00:02:51.785
By the way, I'm going to get into how Cloudflare plays into all this.
39
00:02:53.235 --> 00:02:57.097
in a second, they'll buy access, they'll take action.
40
00:02:57.757 --> 00:03:02.560
And that basically means that the internet needs a new pricing for machine usage.
41
00:03:03.440 --> 00:03:09.423
Obviously, a human being doesn't want to pay 0.0003 cents to read a page.
42
00:03:09.443 --> 00:03:10.544
That would feel ridiculous.
43
00:03:11.104 --> 00:03:16.287
If I clicked on a recipe and it said, please pay one third of a cent to see the sauce,
44
00:03:16.847 --> 00:03:20.708
I would close a laptop and probably order tacos out of spite.
45
00:03:21.209 --> 00:03:25.090
But a machine doesn't care if the payment is tiny and automatic.
46
00:03:25.730 --> 00:03:30.532
If the data helps the agent complete the job, the agent can pay.
47
00:03:31.192 --> 00:03:32.833
And that's really the mental model.
48
00:03:33.513 --> 00:03:40.576
The human web monetized attention and the agent web monetizes useful resources.
49
00:03:41.750 --> 00:03:43.992
So what is Cloudflare actually doing here?
50
00:03:44.493 --> 00:03:47.536
Well, there's a few pieces I want to go over.
51
00:03:47.656 --> 00:03:54.983
The first is AI crawl control, which gives site owners visibility and control over AI crawlers.
52
00:03:55.463 --> 00:03:57.645
You can see a crawler activity.
53
00:03:58.146 --> 00:04:00.568
You can allow certain crawlers to do things.
54
00:04:00.668 --> 00:04:02.290
And you can block certain crawlers.
55
00:04:02.570 --> 00:04:04.452
You can understand who's accessing your content.
56
00:04:04.852 --> 00:04:09.015
They also announced pay per crawl, and that's the direct monetization piece.
57
00:04:09.035 --> 00:04:11.216
That's the piece that's going to get people paid.
58
00:04:11.837 --> 00:04:16.880
A site owner now can charge AI crawlers when they access content.
59
00:04:17.321 --> 00:04:26.327
The crawler could either present payment intent in the request or receive a 402 payment required response with the pricing.
60
00:04:26.527 --> 00:04:30.588
Then monetization gateway is even a bigger version of that.
61
00:04:30.788 --> 00:04:35.790
So Cloudflare is basically saying this should not just apply to crawlers reading pages.
62
00:04:36.390 --> 00:04:46.874
Any resource behind Cloudflare could have payment rules, a web page, a data set, an API, an MCP tool call, basically a premium endpoint.
63
00:04:47.414 --> 00:04:48.835
a file, a search index.
64
00:04:49.316 --> 00:04:58.162
And the payment rail they're talking about is X402, which uses the HTTP 402 payment required status code.
65
00:04:59.083 --> 00:05:00.764
The agent requests the resource.
66
00:05:01.184 --> 00:05:05.848
The server says, this costs this amount, usually a fraction of a penny.
67
00:05:06.408 --> 00:05:09.631
And the agent pays and retries with proof of payment.
68
00:05:10.131 --> 00:05:14.575
Cloudflare can verify at the edge before the request hits the origin.
69
00:05:15.195 --> 00:05:20.881
Basically, in plain English, CloudFair is trying to make paid access feel like part of the internet itself.
70
00:05:21.282 --> 00:05:23.123
So you don't have this giant checkout flow.
71
00:05:23.143 --> 00:05:25.146
You don't have to create an account first.
72
00:05:25.206 --> 00:05:26.347
There's no sales call.
73
00:05:26.627 --> 00:05:28.709
There's no enterprise procurement stuff.
74
00:05:29.170 --> 00:05:30.571
Basically, there's a request.
75
00:05:31.232 --> 00:05:33.453
that becomes the transaction.
76
00:05:33.533 --> 00:05:35.535
And that's a really, really big idea.
77
00:05:36.675 --> 00:05:45.581
And it's a big idea because the request becomes the transaction and then there's these tiny resources on the internet that could become businesses.
78
00:05:45.881 --> 00:05:46.922
And that's what we're going to get into.
79
00:05:46.962 --> 00:05:48.323
That's why I'm doing this episode.
80
00:05:48.343 --> 00:05:51.505
I'm doing this episode not just because Cloud4 launches thing.
81
00:05:53.046 --> 00:05:56.288
It's interesting because of all the businesses that could be built on top of it.
82
00:05:57.448 --> 00:05:59.649
A dataset could charge per lookup.
83
00:05:59.709 --> 00:06:02.190
An API can charge per successful call.
84
00:06:02.370 --> 00:06:05.871
A research archive could charge per answer.
85
00:06:06.252 --> 00:06:09.053
A product catalog can charge per comparison.
86
00:06:09.533 --> 00:06:10.273
You get the idea.
87
00:06:11.053 --> 00:06:17.036
And someone in the comment section is basically going to say something like, yeah, but agents don't have wallets.
88
00:06:17.536 --> 00:06:21.057
Well, we're moving into a future where agents are going to have wallets.
89
00:06:21.077 --> 00:06:22.418
They're going to have email addresses.
90
00:06:23.625 --> 00:06:32.641
And assuming that's true, and I think there's a good chance that that happens, because it's already starting to happen, they're going to make transactions.
91
00:06:34.575 --> 00:06:38.476
And Cloudflare's AI index points in the same direction.
92
00:06:38.556 --> 00:06:50.020
So they've talked about making websites easier for AI systems to use things like MCPs, LLM.txt, search APIs, and bulk data APIs.
93
00:06:50.380 --> 00:06:54.141
That matters because agents need access, but they also need structure.
94
00:06:54.781 --> 00:06:57.682
So humans can tolerate messy websites.
95
00:06:57.722 --> 00:07:00.663
It might not be a good experience, but they can tolerate it.
96
00:07:02.590 --> 00:07:07.891
We click around, we zoom in, we read the FAQ from 2019, we open a PDF.
97
00:07:07.911 --> 00:07:09.451
We're good at suffering human beings.
98
00:07:09.851 --> 00:07:12.292
But agents need really clean doors.
99
00:07:12.412 --> 00:07:15.952
So they need information in a format they could trust and use.
100
00:07:16.613 --> 00:07:29.295
So the shift is websites become resource layers, content becomes indexes, expertise becomes scalable, data becomes metered, and tools become agent accessible.
101
00:07:29.315 --> 00:07:29.935
Okay.
102
00:07:30.455 --> 00:07:34.158
And that's what I'm paying attention to as a founder because I can build on top of it.
103
00:07:34.979 --> 00:07:40.323
So there's this stack that's forming around this whole new internet and I want to go through it.
104
00:07:40.463 --> 00:07:43.085
It's actually simpler than we think.
105
00:07:43.626 --> 00:07:45.888
So first you have a messy internet.
106
00:07:46.368 --> 00:07:52.072
That is PDFs, pricing pages, old blog posts, YouTube videos, support docs,
107
00:07:53.674 --> 00:07:59.959
comparison sites, websites that look like they were designed when people still said Web 2.0 without any irony.
108
00:08:00.839 --> 00:08:04.422
Then someone cleans that into structured data.
109
00:08:05.123 --> 00:08:16.532
Then someone makes it agent readable through an API, MCP tool, search index, feed, llms.txt, or some other clean access point, but it needs to be cleaned.
110
00:08:17.232 --> 00:08:18.933
Then someone adds payment rules.
111
00:08:19.353 --> 00:08:24.634
Some things are going to remain free because free creates a lot of distribution.
112
00:08:24.654 --> 00:08:28.296
But some things are going to be paid because they're extremely valuable.
113
00:08:28.656 --> 00:08:31.877
And some things are going to be blocked because they should stay private.
114
00:08:32.737 --> 00:08:35.078
And then someone adds trust in analytics.
115
00:08:35.578 --> 00:08:37.059
Like is the data fresh?
116
00:08:37.559 --> 00:08:38.839
Is the source reliable?
117
00:08:39.259 --> 00:08:40.600
Which agents are using it?
118
00:08:41.360 --> 00:08:45.088
which requests are worth money, which resources drive outcomes.
119
00:08:45.829 --> 00:08:51.160
And that stack is what's going to create this whole generation of thousands of companies.
120
00:08:53.200 --> 00:09:04.745
When you're looking for ideas here, the question really is, what resource does an agent need badly enough and often enough and reliably enough to pay for?
121
00:09:05.065 --> 00:09:07.586
That's basically the biggest question of all of this.
122
00:09:07.626 --> 00:09:08.887
That's what I'm asking myself.
123
00:09:09.787 --> 00:09:16.770
And let's get into the three ideas that'll get your creative juices flowing around building businesses on top of that.
124
00:09:16.870 --> 00:09:20.191
But before I get into it, I saw a lot of chatter on X,
125
00:09:22.092 --> 00:09:25.644
Levels.io, who's an incredible indie hacker, was talking about how
126
00:09:26.565 --> 00:09:32.707
you know, how hard it is now to build a business and how a lot of people's traffic and revenue has gone down.
127
00:09:33.088 --> 00:09:45.352
I think in large part, you know, because of AI overviews and how basically a lot of software is just not, it's not useful anymore because people are vibe coding it themselves and stuff like that.
128
00:09:45.813 --> 00:09:50.154
My take is it's an incredible time to be building.
129
00:09:50.194 --> 00:09:51.655
In fact, it's the best time ever.
130
00:09:52.295 --> 00:09:56.757
Now, I think the types of businesses to create are businesses like this.
131
00:09:56.857 --> 00:09:59.578
They're not like little tools that could be vibe-coded.
132
00:09:59.758 --> 00:10:08.042
These are startup ideas that could be profitable on day one, that have some moat, that are...
133
00:10:09.102 --> 00:10:13.247
are reinvented for an internet that's AI native and agent ready.
134
00:10:13.307 --> 00:10:22.217
So I don't want you to go on X and just be devastated that there's no entrepreneurship is done, RIP startups.
135
00:10:22.337 --> 00:10:26.101
No, the world is moving, the internet is changing, you have to move with it.
136
00:10:26.181 --> 00:10:28.824
And here's three startup ideas that move with it.
137
00:10:29.084 --> 00:10:29.425
Let's go.
138
00:10:30.102 --> 00:10:33.945
So the first startup idea is a niche data refinery.
139
00:10:34.325 --> 00:10:38.408
This is probably the one I would start with because it's the most practical.
140
00:10:38.688 --> 00:10:39.668
The idea is really simple.
141
00:10:39.728 --> 00:10:49.095
So pick one niche where valuable information is messy, fragmented, changing, annoying to collect,
142
00:10:49.935 --> 00:10:53.938
then turn that information into clean fuel for agents.
143
00:10:54.638 --> 00:10:59.201
I'm calling it a data refinery because the raw material already exists.
144
00:10:59.441 --> 00:11:01.362
So the internet already has that data.
145
00:11:02.083 --> 00:11:07.226
It's sitting in Google Maps, it's sitting in job posts, it's sitting in reviews and local directories.
146
00:11:08.046 --> 00:11:10.289
PDFs, it sits in pricing pages.
147
00:11:10.569 --> 00:11:14.733
It's in all these places and your job is going to be refining it.
148
00:11:15.314 --> 00:11:17.757
So let's make it concrete.
149
00:11:17.797 --> 00:11:20.279
I'll give you an example so it drives the point home.
150
00:11:21.253 --> 00:11:30.180
So imagine you pick med spas because, you know, they're growing a lot in the U.S. And that's the niche you want to pick.
151
00:11:30.240 --> 00:11:30.501
Great.
152
00:11:30.801 --> 00:11:47.875
A med spa owner wants to know what competitors are charging, what treatments they offer, what reviews complain about, which clinics are hiring, which services are trending, what offers are working, and how the local market is changing.
153
00:11:48.355 --> 00:11:51.296
So that information today lives everywhere, really.
154
00:11:51.916 --> 00:12:02.360
It lives on Google reviews, on competitors' websites, on Instagram, on job posts, meta ad libraries, in the owner's head, in employee's heads.
155
00:12:03.620 --> 00:12:09.362
An agent can do incredible work for a med spa owner if it had that information cleanly.
156
00:12:09.482 --> 00:12:10.723
It would be huge.
157
00:12:11.263 --> 00:12:19.092
It could say something like your Botox pricing is above the local median, but your reviews do not support premium positioning yet.
158
00:12:19.332 --> 00:12:20.513
That would be super valuable.
159
00:12:21.174 --> 00:12:28.062
It could say three competitors near you started promoting exosome treatments in the last 60 days.
160
00:12:28.923 --> 00:12:30.324
Super valuable information to know.
161
00:12:31.125 --> 00:12:36.370
It could say the most common complaint in local reviews is you're confusing pricing.
162
00:12:36.750 --> 00:12:38.552
So your offer should lead with simplicity.
163
00:12:39.093 --> 00:12:39.733
Great to know.
164
00:12:40.594 --> 00:12:47.421
It could say too fast growing competitors are hiring injectors, which probably means they're expanding capacity.
165
00:12:47.621 --> 00:12:48.322
Really good to know.
166
00:12:48.362 --> 00:12:49.783
So all useful information.
167
00:12:50.323 --> 00:12:53.747
And more importantly, the usefulness comes from the data.
168
00:12:54.187 --> 00:12:56.108
not from a generic AI wrapper.
169
00:12:56.668 --> 00:13:01.569
So here's how I would build a wedge into this space, into this business.
170
00:13:01.649 --> 00:13:03.910
I would pick one niche and I would pick one city.
171
00:13:04.410 --> 00:13:06.911
I would track 100 businesses, not more.
172
00:13:06.931 --> 00:13:09.372
I would do it manually at first.
173
00:13:10.252 --> 00:13:27.657
I would create a spreadsheet with your business name, website, services, prices, review count, review rating, top review complaints, Instagram links, recent posts, visible ad changes, hiring signals, and booking flow.
174
00:13:28.137 --> 00:13:30.358
Then I would create 10 outputs from that data.
175
00:13:30.718 --> 00:13:36.360
It could be like a local pricing map, a competitor gap report, a list of offer ideas,
176
00:13:37.380 --> 00:13:45.349
A services to ad recommendation, a review complaint summary, a hiring signal report, a monthly market movement report.
177
00:13:46.150 --> 00:13:48.312
Now you have something to sell here.
178
00:13:49.113 --> 00:13:52.597
And the first customer actually might not be a med spa owner.
179
00:13:53.197 --> 00:13:56.220
And that's a part I think people are missing a little bit.
180
00:13:56.881 --> 00:14:01.805
In the early days, your first customer is often the person already selling into the niche.
181
00:14:02.005 --> 00:14:17.476
So instead of trying to sell agent readable competitive analysis to a med spa owner, which is a phrase that is just confusing to the average person, sell it to med spa marketing agencies, consultants, freelancers, clients.
182
00:14:18.056 --> 00:14:20.777
software companies, and even AI implementation people.
183
00:14:21.237 --> 00:14:31.861
You could say, I built local market intelligence for med spas, and you can use it to create better audits, better offers, better landing pages, and better campaigns for your clients.
184
00:14:31.921 --> 00:14:33.802
They're going to be able to charge more.
185
00:14:34.962 --> 00:14:39.164
So what you've built is super valuable.
186
00:14:39.524 --> 00:14:41.005
And that's just a way easier sale.
187
00:14:42.385 --> 00:14:48.670
A MedSpa marketing agency might sell a client for like $5K a month as a growth package.
188
00:14:49.170 --> 00:14:57.216
And if your data helps them close one more client or improve their work, they can pay you $300, $500, $800 a month.
189
00:14:57.576 --> 00:15:01.178
And over time, you turn your spreadsheet into a real product.
190
00:15:01.238 --> 00:15:07.523
So first it's a report, then it's a dashboard, then it becomes an API, then maybe an MCP tool.
191
00:15:08.023 --> 00:15:12.006
then agents can pay per lookup or per report when the rails are ready.
192
00:15:12.046 --> 00:15:15.589
So you're doing basically a crawl, walk, run strategy.
193
00:15:16.130 --> 00:15:19.352
I used med spas as an example, but it doesn't need to be med spas.
194
00:15:19.893 --> 00:15:23.376
I live in Miami and there just so happens to be a lot of med spas here.
195
00:15:24.136 --> 00:15:26.098
But you can do it for roofing.
196
00:15:26.318 --> 00:15:30.922
You can do it for, let's say you were to do it for roofing.
197
00:15:31.562 --> 00:15:34.204
You would track storm events maybe.
198
00:15:34.224 --> 00:15:41.549
You would track permit data, insurance signals, local reviews, competitor offers, and ad angles.
199
00:15:41.569 --> 00:15:43.370
You can do it for real estate investing.
200
00:15:43.730 --> 00:15:45.011
You track zoning changes.
201
00:15:45.031 --> 00:15:45.951
You track permits.
202
00:15:45.971 --> 00:15:47.332
You track ownership records.
203
00:15:47.352 --> 00:15:48.473
You track rent comps.
204
00:15:49.013 --> 00:15:50.654
You track tax delinquencies.
205
00:15:50.715 --> 00:15:52.256
And you track insurance shifts.
206
00:15:53.096 --> 00:15:54.557
You can do it for e-commerce.
207
00:15:54.637 --> 00:16:03.924
You can track competitor SKUs, pricing changes, review complaints, influencer rates, UGC hooks, Shopify apps, shipping promises.
208
00:16:04.484 --> 00:16:05.485
You can do law firms.
209
00:16:05.945 --> 00:16:10.869
You can track local competitors, practice area positioning, ad copy, reviews, intake.
210
00:16:10.889 --> 00:16:11.669
You get the idea.
211
00:16:12.310 --> 00:16:14.712
The filter is pretty simple.
212
00:16:15.152 --> 00:16:16.513
The data should be valuable.
213
00:16:17.295 --> 00:16:22.819
The data should be repeatable, changing, fragmented, and annoying.
214
00:16:23.100 --> 00:16:28.824
And valuable data basically means better decisions make or save money.
215
00:16:29.645 --> 00:16:30.746
What does repeated mean?
216
00:16:30.786 --> 00:16:33.388
Well, repeated means the customer needs it again and again.
217
00:16:33.448 --> 00:16:34.669
It's not like a one-time thing.
218
00:16:35.129 --> 00:16:36.069
What does changing mean?
219
00:16:36.109 --> 00:16:39.311
It means that freshness matters in the data.
220
00:16:39.871 --> 00:16:42.032
What does fragmented mean?
221
00:16:42.132 --> 00:16:45.073
It means that one person can easily collect it.
222
00:16:45.993 --> 00:16:46.933
What does annoying mean?
223
00:16:47.173 --> 00:16:48.994
That basically means there's some margin there.
224
00:16:49.794 --> 00:16:51.635
And that's my first startup idea.
225
00:16:52.455 --> 00:16:57.837
It's basically one idea that can give you a lot of ideas, depending on what niche that you have some unfair advantage in.
226
00:16:59.918 --> 00:17:04.685
Take one niche's messy internet, turn it into clean fuel for agents.
227
00:17:04.945 --> 00:17:07.308
I don't want you to forget that phrase, clean fuel for agents.
228
00:17:08.149 --> 00:17:09.731
Let's get on to the next startup idea.
229
00:17:10.597 --> 00:17:14.779
So startup idea number two is agent readiness for businesses.
230
00:17:15.359 --> 00:17:25.783
This is like SEO for the agent internet, but I wanted to find it very specifically because AI SEO is already becoming a buzzy, fuzzy phrase.
231
00:17:26.303 --> 00:17:32.866
The real business is helping companies become easy for agents to understand, trust, compare, and recommend.
232
00:17:33.446 --> 00:17:35.107
Think about a B2B SaaS company.
233
00:17:35.547 --> 00:17:37.648
A human buyer lands on the homepage,
234
00:17:38.663 --> 00:17:50.232
reads the hero image, clicks around, looks at the pricing page, reads some docs, maybe books a demo, watches a case study, and maybe asks a friend about this particular product.
235
00:17:51.152 --> 00:17:54.055
Agents are compressing that whole process.
236
00:17:54.555 --> 00:17:58.358
If someone asks their AI assistant, find me the best payroll provider,
237
00:17:59.376 --> 00:18:04.260
For a 15-person company in California, the agent has to understand the market.
238
00:18:04.680 --> 00:18:07.962
So the agent is basically asking, who's this product for?
239
00:18:08.042 --> 00:18:09.003
What does it cost?
240
00:18:09.383 --> 00:18:10.364
What does it replace?
241
00:18:10.404 --> 00:18:11.605
What sort of integrations?
242
00:18:11.645 --> 00:18:12.666
What are the risks?
243
00:18:13.346 --> 00:18:15.068
What does implementation look like?
244
00:18:15.148 --> 00:18:16.209
What do customers say?
245
00:18:16.549 --> 00:18:18.070
How does it compare to alternatives?
246
00:18:18.430 --> 00:18:21.833
Most websites actually make this harder than it needs to be.
247
00:18:22.113 --> 00:18:23.194
They hide pricing.
248
00:18:23.234 --> 00:18:24.435
They bury docs.
249
00:18:25.475 --> 00:18:27.576
There's just not a lot of information there.
250
00:18:27.896 --> 00:18:33.278
They're not constantly updating their website, so they end up having stale websites, stale comparison pages.
251
00:18:34.379 --> 00:18:38.480
Some of the most important information is actually in PDFs.
252
00:18:38.680 --> 00:18:42.502
And finding policies is extremely hard to find in general.
253
00:18:43.624 --> 00:18:45.825
And there's just a lot of marketing speak.
254
00:18:45.845 --> 00:18:49.467
So you'll see things like unlocking operational excellence.
255
00:18:50.147 --> 00:18:52.789
So there's a lot of just like foggy copywriting.
256
00:18:53.589 --> 00:18:58.392
So the startup idea is basically to make these websites agent readable.
257
00:18:58.712 --> 00:19:00.393
Here's the wedge I would use.
258
00:19:01.113 --> 00:19:02.534
Start with a paid audit.
259
00:19:02.634 --> 00:19:07.135
And I've done episodes on the podcast with Corey Gannam around some of these things.
260
00:19:08.315 --> 00:19:16.698
You can go deeper into with Vaz, where we talked about what is a forward deploy engineer, where we've talked about paid audits.
261
00:19:17.098 --> 00:19:18.799
But I want you to pick one vertical.
262
00:19:19.339 --> 00:19:29.182
B2B SaaS is obvious, but you can do Shopify apps, law firms, health care clinics, financial advisors, insurance brokers, home services, whatever you know.
263
00:19:31.624 --> 00:19:36.488
20 to 50 buyer intent prompts across major AI tools.
264
00:19:36.929 --> 00:19:40.872
So you're going to ask questions like, what is the best software for this use case?
265
00:19:41.213 --> 00:19:43.615
Compare this company to top alternatives.
266
00:19:44.055 --> 00:19:45.817
What does this company cost?
267
00:19:46.237 --> 00:19:47.298
Who is this product best for?
268
00:19:47.338 --> 00:19:48.660
What are the risks of choosing this vendor?
269
00:19:48.680 --> 00:19:51.022
What are the risks of choosing this vendor?
270
00:19:51.442 --> 00:19:55.005
Would you recommend this product for a 20-person company?
271
00:19:55.505 --> 00:19:57.787
What integrations does it support?
272
00:19:58.287 --> 00:20:00.569
Then you show the company the answers.
273
00:20:00.969 --> 00:20:07.734
And this is really the sales moment because you might show a founder, when buyers ask AI for your category, you do not show up.
274
00:20:08.174 --> 00:20:10.756
Or you show up, but AI gets your pricing wrong.
275
00:20:11.096 --> 00:20:15.540
On your website, it's $20 a month, but AI for some reason is getting $8 a month.
276
00:20:16.338 --> 00:20:20.241
Or the AI recommends your competitor because their docs are cleaner.
277
00:20:20.981 --> 00:20:26.005
Or your website has the answer, but it's buried in a PDF from 2002.
278
00:20:26.385 --> 00:20:30.248
That gets their attention, especially if you're doing cold email and stuff like that.
279
00:20:30.588 --> 00:20:31.629
Then you sell the fix.
280
00:20:32.169 --> 00:20:35.172
The fix is an agent-readable source of truth.
281
00:20:35.752 --> 00:20:42.217
So that might include a cleanllms.txt file, which is a file that helps...
282
00:20:45.859 --> 00:20:47.581
LLMs crawl your website.
283
00:20:48.581 --> 00:20:50.343
A better documentation structure.
284
00:20:50.363 --> 00:20:54.006
A pricing page agents can parse.
285
00:20:54.046 --> 00:20:56.427
Comparison pages that are honest and specific.
286
00:20:57.188 --> 00:21:02.272
Use cases pages written in just plain, simple language.
287
00:21:02.532 --> 00:21:04.454
Customer proof organized by segment.
288
00:21:04.854 --> 00:21:14.087
Structured FAQs around real buyer questions, schema markup, a product feed, a change log, a lightweight MCP server or search endpoint.
289
00:21:14.247 --> 00:21:19.374
If the company has enough useful content, if they don't, then probably not that.
290
00:21:19.995 --> 00:21:23.458
And then the recurring product is the measurement loop.
291
00:21:23.698 --> 00:21:38.450
So every month you rerun the prompts, you see what's changed, you see whether AI answers are more accurate, whether the company appears more often, whether the competitor comparisons improve or get worse, and you see where the website needs more structured proof.
292
00:21:38.870 --> 00:21:41.693
And this could start off as an easy services business.
293
00:21:42.233 --> 00:21:46.794
So you can charge something like $3,000 to $10,000 for the audit and cleanup.
294
00:21:47.215 --> 00:21:52.256
And then for larger B2B companies, it could be something like $10,000, $15,000, $20,000.
295
00:21:52.756 --> 00:21:57.698
And after you do 10 clients in the same niche, you're going to see a bunch of repeated work.
296
00:21:57.718 --> 00:22:03.560
And this is why I love starting with services businesses and then productizing it with software as you go.
297
00:22:04.120 --> 00:22:06.964
You're going to learn things like the same docs are missing.
298
00:22:07.405 --> 00:22:09.507
The same pricing pages are unclear.
299
00:22:09.568 --> 00:22:10.990
The same questions matter.
300
00:22:11.270 --> 00:22:13.413
The same structured files need to be created.
301
00:22:13.693 --> 00:22:16.577
The same monthly report needs to be delivered.
302
00:22:16.978 --> 00:22:19.141
This is when you turn it into software.
303
00:22:19.541 --> 00:22:20.222
And the way...
304
00:22:20.883 --> 00:22:23.145
To sell this is very simple.
305
00:22:23.185 --> 00:22:26.288
You're not selling the future, you're selling the screenshot.
306
00:22:26.688 --> 00:22:29.891
You show them what AI says about their company today.
307
00:22:30.411 --> 00:22:32.333
And that becomes the whole sales deck.
308
00:22:32.433 --> 00:22:38.918
For local businesses, this eventually becomes let AI assistants book appointments with you.
309
00:22:39.439 --> 00:22:46.025
For e-commerce, it becomes something like make your product catalog easy for shopping agents to compare and buy.
310
00:22:47.179 --> 00:22:52.602
For B2B SaaS, it becomes make your product easy for procurement agents to evaluate.
311
00:22:53.082 --> 00:22:58.745
And for publishers, it becomes make your archive easy for AI systems to understand and license.
312
00:23:01.023 --> 00:23:07.792
I think that there's going to be venture-backed companies, and you're already starting to see it happen, doing these horizontal products that do this.
313
00:23:07.852 --> 00:23:10.536
The opportunity here is to go extremely vertical.
314
00:23:10.556 --> 00:23:14.141
Obviously, B2B SaaS is too big, right?
315
00:23:14.241 --> 00:23:18.366
You want to pick a specific niche, go deep into that.
316
00:23:19.467 --> 00:23:26.589
But I think this is a wonderful business that's cash flow on day one that could productize that could eventually sell at some point in the future.
317
00:23:27.489 --> 00:23:33.170
Basically help businesses become easy for agents to understand trust compare and recommend.
318
00:23:33.590 --> 00:23:41.692
There's a ton of demand from businesses right now for this because they're feeling this pain.
319
00:23:42.652 --> 00:23:43.512
So you just solve it.
320
00:23:43.992 --> 00:23:50.076
Last but not least, startup idea number three is turning expert archive into agent tools.
321
00:23:50.096 --> 00:23:51.417
I'll explain what I mean by that.
322
00:23:52.417 --> 00:24:07.186
This one is probably the most fun for creators, media companies, analysts, consultants, researchers, people like that that are sitting on years of valuable content or can get their hands on valuable content.
323
00:24:07.586 --> 00:24:09.628
What do I mean by valuable content?
324
00:24:10.408 --> 00:24:18.375
I'm talking about things like YouTube videos, podcasts, newsletters, templates, things like that, community posts.
325
00:24:19.395 --> 00:24:29.123
Right now, most of that content makes money through ads, sponsorships, sometimes subscriptions, sometimes communities, sometimes consulting.
326
00:24:29.683 --> 00:24:32.886
But in the agent internet, that archive could become a tool.
327
00:24:33.426 --> 00:24:35.108
So imagine a founder agent
328
00:24:35.708 --> 00:24:39.932
That can access a startup's expert archive and critique your idea.
329
00:24:40.372 --> 00:24:46.518
Imagine a sales agent that can use a specific sales trainer's framework to rewrite your cold email.
330
00:24:46.958 --> 00:24:53.364
Imagine a fitness agent that can use a coach's training philosophy to build a personalized plan.
331
00:24:54.285 --> 00:24:57.086
The startup is archive to API.
332
00:24:57.646 --> 00:25:02.308
You take someone's expertise and you package it so agents can use it.
333
00:25:02.909 --> 00:25:04.269
The key is to start with one job.
334
00:25:04.629 --> 00:25:08.971
Don't go to a creator and say, we're going to turn your whole brain into AI.
335
00:25:09.411 --> 00:25:15.574
That honestly sounds a little creepy, a little vague, and honestly like a SaaS landing page that should be illegal.
336
00:25:16.074 --> 00:25:17.795
You got to say something that's specific.
337
00:25:17.895 --> 00:25:19.456
So what do I mean by that?
338
00:25:19.856 --> 00:25:22.277
Something like you have 300 videos about sales.
339
00:25:22.477 --> 00:25:27.700
We're going to turn them into a tool your audience could use to improve cold emails.
340
00:25:28.260 --> 00:25:34.364
Or hey, you've got 500 episodes about startups, podcast episodes about startups.
341
00:25:34.404 --> 00:25:37.005
We're going to turn them into a startup idea feedback tool.
342
00:25:37.285 --> 00:25:38.586
Or maybe it's to a designer.
343
00:25:38.786 --> 00:25:41.508
It's like you have a decade of these design teardowns.
344
00:25:41.548 --> 00:25:44.730
We're going to turn them into a landing page critique tool.
345
00:25:45.470 --> 00:25:49.391
You've got this one archive with one painful job, one workflow.
346
00:25:49.891 --> 00:25:52.752
And here's how I would build this startup idea.
347
00:25:52.852 --> 00:25:57.454
So I would pick an expert with a deep archive and a specific audience.
348
00:25:57.834 --> 00:26:05.156
Me, in particular, I'd pick someone who is more in the B2B space, but it could work for B2C.
349
00:26:06.583 --> 00:26:15.674
Specific matters just because a general business creator, for example, is just going to be harder to do.
350
00:26:16.234 --> 00:26:24.784
But maybe a creator known for cold email or Shopify growth or local business acquisitions, tax strategy, fitness programming,
351
00:26:25.345 --> 00:26:27.207
Or like what we talked about, design teardowns.
352
00:26:27.227 --> 00:26:28.868
That would work really well.
353
00:26:29.509 --> 00:26:31.831
Second, you want to go and collect the archive.
354
00:26:31.851 --> 00:26:39.499
So you're going to want to go and transcribe the content, if that's videos, if it's podcasts, pull the newsletters, clean the docs.
355
00:26:39.759 --> 00:26:41.260
And then third, you're going to do tagging.
356
00:26:41.280 --> 00:26:48.988
So you're going to tag the archive by job, by topic, by audience, by example, by framework, and by outcome.
357
00:26:49.328 --> 00:26:51.511
A lot of people get lazy at this part.
358
00:26:51.531 --> 00:26:54.776
They throw everything into a vector database and then they just call it a day.
359
00:26:55.137 --> 00:26:57.841
That usually gives you a search box with confidence
360
00:26:58.883 --> 00:27:00.985
But a real product needs structure.
361
00:27:01.185 --> 00:27:12.256
So if it's a sales archive, you want to tag by prospecting, by subject line, by offer, by objection, by follow-up, personalization, deliverability, and close.
362
00:27:12.816 --> 00:27:21.064
If it's a startup archive, you want to tag by the idea, by the market, the wedge, the distribution, the pricing, the MVP, the community, the mode, and examples.
363
00:27:22.467 --> 00:27:24.389
So you're very specific, right?
364
00:27:24.549 --> 00:27:31.556
I think the best tools here are going to be very specific and that's going to help get the best outcome for people ultimately.
365
00:27:33.221 --> 00:27:36.303
Fourth, what you want to do is build one useful workflow.
366
00:27:36.444 --> 00:27:41.568
So for a sales expert, that workflow could just be paste your cold email.
367
00:27:42.068 --> 00:27:47.013
The agents are going to critique it using that expert's principles.
368
00:27:47.173 --> 00:27:48.994
Maybe it's Alex Hermosi, let's say.
369
00:27:49.054 --> 00:27:50.836
Alex Hermosi is going to go and critique it.
370
00:27:51.416 --> 00:27:52.998
It cites the source lessons.
371
00:27:53.138 --> 00:27:54.299
It rewrites the email.
372
00:27:54.619 --> 00:27:55.700
It gives you a score.
373
00:27:56.301 --> 00:27:57.882
It gives you one test to run.
374
00:27:58.182 --> 00:27:59.724
And that's the product.
375
00:28:01.225 --> 00:28:14.072
For a startup expert, the workflow could be paste your idea, the agent gives you the wedge, it gives you the customer, it writes the first offer, it suggests the first distribution channel, and it tells you what to validate this week.
376
00:28:14.092 --> 00:28:17.314
That's kind of what we're doing with ideabrowser.com.
377
00:28:17.414 --> 00:28:18.055
You know, you...
378
00:28:19.118 --> 00:28:21.780
one of our most popular features is adding the MCP.
379
00:28:22.240 --> 00:28:27.864
And it's so good because it takes your LLM and just makes it better, right?
380
00:28:27.904 --> 00:28:31.846
And it's got all this data that we've cleaned to help you do that.
381
00:28:32.106 --> 00:28:34.508
So I'm practicing what I'm preaching here.
382
00:28:36.315 --> 00:28:38.757
For a real estate expert, what could that be?
383
00:28:38.817 --> 00:28:50.087
Well, it could be something like you pace the deal, the agent checks the assumptions, it identifies the risk, it compares it to the expert criteria, and it tells you whether to ask the broker.
384
00:28:50.227 --> 00:28:51.788
That would be super, super useful.
385
00:28:53.209 --> 00:29:00.093
What's cool about doing something like this, the creator already has the distribution, so you don't have to worry about the marketing, the customer acquisition.
386
00:29:00.414 --> 00:29:04.956
They've got all that trust already built in, but the audience wants the expertise.
387
00:29:05.717 --> 00:29:10.039
And the creator might not want to do consulting with everyone, right?
388
00:29:10.079 --> 00:29:12.521
So this democratizes that.
389
00:29:12.941 --> 00:29:14.702
So you can charge a lot cheaper.
390
00:29:14.782 --> 00:29:16.583
You could be something like $19 a month or $50 a month
391
00:29:18.144 --> 00:29:29.411
You can bundle into a paid community, or you can even make it a lead magnet for the consulting, or you can license it to agencies or software companies, that sort of thing.
392
00:29:30.992 --> 00:29:44.280
This is really where the Cloudflare-style monetization becomes interesting, because if the archive becomes this resource and agents can pay per request, the creator gets paid when the knowledge is used, the builder gets this trusted resource
393
00:29:44.939 --> 00:29:51.985
expert layer, the person consuming it gets this expert layer and gets better output.
394
00:29:53.006 --> 00:30:00.593
That's way better than hoping someone watches a pre-roll ad before a 47-minute interview from seven years ago.
395
00:30:01.233 --> 00:30:09.901
I think the biggest mistake in this category that I've noticed is I've seen a lot of chat with an expert products, but that's too broad.
396
00:30:09.961 --> 00:30:11.002
So I think the specific
397
00:30:13.223 --> 00:30:17.127
The specific use case is way more interesting and it's way more outcome based.
398
00:30:17.167 --> 00:30:20.890
It's not chat with the sales person or chat with the sales creator.
399
00:30:20.970 --> 00:30:24.173
It's rewrite the cold email using the sales system.
400
00:30:25.913 --> 00:30:36.121
So if you can turn expert archives into job-specific agent tools, I think that could be really cool and someone should do it.
401
00:30:36.701 --> 00:30:38.623
So those are the three startup ideas.
402
00:30:39.564 --> 00:30:41.185
Number one, the niche data refinery.
403
00:30:41.245 --> 00:30:43.407
Number two, the agent readiness for business.
404
00:30:43.427 --> 00:30:46.970
And number three, the expert archives turned into agent tools.
405
00:30:48.030 --> 00:30:54.894
What connects all three of these ideas is agents need clean, trusted, and useful resources to do really good work.
406
00:30:54.974 --> 00:31:02.718
So that resource can be data, but it could be structure, it could be access, it could be expert knowledge, it could be a tool, it could be a payment rule.
407
00:31:03.398 --> 00:31:12.586
The Cloudflare news matters because building part of the access and payment layer, that's what they're building, and that's a huge thing.
408
00:31:13.046 --> 00:31:21.634
But you don't have to wait for Cloudflare's whole new agent internet to mature in order to start building out some of this stuff.
409
00:31:22.154 --> 00:31:26.978
So you can start by building the manual version first.
410
00:31:26.998 --> 00:31:28.980
You can sell the human version now.
411
00:31:29.040 --> 00:31:30.361
You can build the data now.
412
00:31:31.822 --> 00:31:33.422
and you can package this up now.
413
00:31:33.642 --> 00:31:42.764
So as agents become more capable and agent payments become more and more common, you're not scrambling to start building this stuff.
414
00:31:43.205 --> 00:31:47.025
This is the early, you're early, you know?
415
00:31:47.185 --> 00:31:48.846
Not many people are talking about this.
416
00:31:49.566 --> 00:31:56.027
And if you're looking for ideas in the space, the questions you should ask yourself are, what decision is expensive?
417
00:31:56.487 --> 00:31:58.008
What information is messy?
418
00:31:59.328 --> 00:32:00.689
What changes often?
419
00:32:01.289 --> 00:32:03.550
Who already pays for help?
420
00:32:03.710 --> 00:32:07.491
And what would an agent need to do to do the job better?
421
00:32:07.612 --> 00:32:11.353
That's basically the map to be thinking about and the questions to be thinking about.
422
00:32:11.873 --> 00:32:20.217
So you can spend the next 6, 12, 18 months building up this data, curating the data before it becomes hyper-competitive.
423
00:32:21.458 --> 00:32:26.902
I truly believe the internet is shifting from pages human visit to resources agents use.
424
00:32:27.302 --> 00:32:28.903
And I don't think it's crazy to say that.
425
00:32:29.464 --> 00:32:32.106
And if you made it this far, you probably agree too.
426
00:32:33.827 --> 00:32:42.254
The best opportunities look really small right now because the agent internet is small relative to where it's going to be.
427
00:32:43.275 --> 00:32:47.478
But I think that's how a lot of the biggest businesses started, right?
428
00:32:47.618 --> 00:32:49.760
You know, this is like building an app.
429
00:32:50.305 --> 00:32:54.408
when the App Store came out in 2009.
430
00:32:54.548 --> 00:33:05.395
This cohort of businesses starting in this era, 2026, 2027, I think there's going to be just an incredible amount of businesses created here.
431
00:33:05.815 --> 00:33:08.577
It's very clear that the agent internet is the next wave.
432
00:33:09.518 --> 00:33:17.223
So when you see Cloudflare talking about AI crawlers, X402, paid access, MCP tools, and monetization,
433
00:33:18.043 --> 00:33:20.466
Don't just think, that's a little interesting thing.
434
00:33:20.526 --> 00:33:23.810
Or don't just think, oh, publishers can now charge bots now.
435
00:33:24.391 --> 00:33:25.572
It's way bigger than that.
436
00:33:25.833 --> 00:33:27.294
Agents are becoming buyers.
437
00:33:27.435 --> 00:33:29.077
Websites are becoming resources.
438
00:33:29.757 --> 00:33:31.459
And the next great internet businesses...
439
00:33:34.403 --> 00:33:39.804
It might be these tiny paid doors that agents walk through all day that you're going to own.
440
00:33:40.065 --> 00:33:43.866
It's an asset and I can't wait to see what you build, what you do.
441
00:33:44.626 --> 00:33:50.947
If this has been helpful or got your creative juices flowing, shoot me over a comment, like and subscribe for more of this in your feed.
442
00:33:51.368 --> 00:33:53.088
I read every single comment, by the way.
443
00:33:54.268 --> 00:33:55.449
And I'll see you on the next time.
444
00:33:55.889 --> 00:33:58.229
Thank you for listening to the Startup Ideas podcast.
445
00:33:58.770 --> 00:33:59.790
Have a creative day.
446
00:34:00.510 --> 00:34:01.150
I'll see you next.
447
00:34:01.470 --> 00:34:02.991
And wow, I already said I'd see you next time.
448
00:34:04.716 --> 00:34:05.758
I'm missing you already.
449
00:34:06.319 --> 00:34:08.302
I can't wait to the next episode.
450
00:34:08.602 --> 00:34:08.843
See ya.
