Everything that we do is for free. But it kind of makes more sense now to actually look at how we sell this data.
That sentence captured a shift I've been watching unfold across hundreds of conversations.
- Niche data publishers
- Vertical SaaS companies
- Industry research outfits
- Business-to-business marketplaces
They're all sitting on structured data that's been a cost center for years.
And they're starting to realize that AI just turned that cost center into a product.
A reference site nobody could query
The founder I was talking to runs a small media and SaaS operation in a specific vertical.
His company has built a reference site for the industry.
- Statistics
- Leaderboards
- Regional breakdowns
- Year-over-year comparisons
Most of it pulled together over years of effort and partnerships.
Until now, it's been content marketing. Free to access.
The reasoning he gave me was straightforward: they just put high-level stuff up there, used it as content marketing, and called it a day.
Honestly?
That was the right call for a long time.
If your audience is regular users, dumping a raw database in front of them is worse than dumping nothing.
As he put it, the underlying data is:
Kind of confusing for just a regular user to decipher.
Tables nobody can read are just noise.
So they made a tradeoff: simplify, summarize, give it away, hope it drives some leads.
Pain points I keep hearing
Here's where it gets interesting.
The full data set he's sitting on, three databases, the largest with seventy million rows, has actually been useful the whole time.
Researchers, installers, financiers, regulators, brokers, all of them want to query it.
They want to slice it
- By region
- By year
- By category
- By size
They want answers, not raw rows.
The problem was always the interface
Building a custom dashboard for every type of question is a project. A big one.
And the moment your pre-canned reports don't show exactly what someone wants, they're stuck.
Or they email you. Or they leave.
This is the part I keep seeing on calls. Three pain points show up over and over:
1. The data is too big and too messy to expose raw
Seventy million rows on an analytics-unfriendly database, in his words:
Takes freaking ages to get a query out of.
Most data publishers are running on operational databases that were never designed for ad-hoc analytical queries.
2. The dashboard approach doesn't scale to questions
You can build five reports, ten, twenty.
You'll still miss the question your most valuable user actually wants to ask.
3. The free-versus-paid line is impossible to draw with static content
- Put your best charts behind a paywall and casual visitors bounce.
- Keep everything free and you have no business on top of the data.
For years, the only realistic answer was: pick one. Most people picked free.
What actually changed
What changed is the front end.
Specifically: a chat interface that can answer ad-hoc questions against a real database, without you pre-building every possible query.
That's not a small change.
It's a category shift in how data products work.
Think about what that unlocks.
The casual visitor still gets the high-level free charts.
But the researcher, the broker, the regulator, the consultant, anyone who needs to ask three layers of follow-up questions, now has somewhere to do that.
And because nobody else in their vertical has done it yet, the willingness to pay is real.
I asked him what he thought users would pay for it. He wasn't sure.
We talked through what other customers are charging for similar products.
- Some run it as a premium tier inside an existing subscription.
- Others spin it out as a standalone product at over a hundred dollars a month.
The pricing is wide open because nobody's set the floor yet.
That last point is what I keep telling founders. The window for being the first AI-powered data product in your vertical is open right now. In two years it'll be a feature.
Today it's a product.
The moment the math flipped
During the call, he started off describing the project as:
Let's see if we can charge a few bucks for this.
By the end, he was talking about it as a real revenue line.
Nothing in his business had changed.
- The data was the same.
- The audience was the same.
- The brand was the same.
What changed was the realization that the cost of building the front end had collapsed.
For most of my career, putting a chat-driven analytical layer on top of a real database was a six-month engineering project, minimum. Years of building Birst taught me what that infrastructure costs.
That's the part most founders haven't internalized yet.
The barrier to building a data product on top of your own data isn't engineering anymore.
It's deciding to do it.
Every company is sitting on a data product
There's a wider pattern here.
A lot of companies are sitting on data that's quietly valuable.
- CRM data.
- Transaction data.
- Industry statistics.
- Public market intelligence.
- Internal analytics.
For years that data lived in dashboards used by twelve people internally, or got published as a free PDF report once a quarter, or sat in a database nobody queried because the answers weren't worth the SQL.
AI changed the unit economics of that question.
It's now cheap to put a competent analyst in front of a structured data set. So the question for every company sitting on data is no longer "is this worth building a dashboard for" but "is this worth charging for."
- Some answers will be no.
- Some answers will be a casual yes.
- Some will be serious business.
The interesting part is that nobody knows which is which until they try.
If you're running a SaaS company: the data your customers create inside your product is probably worth surfacing through chat to those same customers.
If you're running a media operation: the proprietary data behind your content is probably worth selling.
If you're running an ops-heavy business: your operational data is probably worth charging for as a benchmark service to peers.
None of this is theoretical.
We're seeing it work right now.
Across data source integrations covering:
- CRMs
- Finance platforms
- Operational databases, and
- Proprietary research sets
Companies discover their data is more valuable than they thought, because it's finally easy for someone to ask a question of it.
The good problems
The founder I talked to is now figuring out his columnar database setup, his pricing tier, and how to position the launch.
The hard parts. The good problems.
If you're sitting on a data set you've been treating as a cost center, the question worth asking is whether you've been giving away a product.
Not every dataset clears that bar.
But the ones that do are quietly becoming some of the most defensible businesses I've seen build this year.
If you want to see what an AI-driven data product actually feels like in practice, you can ask your data on Scoop right now and watch the chat layer work against a live data set. For a deeper read on the broader shift, how AI is transforming business intelligence covers the macro picture, and the pillar piece on AI data analysis walks through the mechanics. If your situation looks like the one I described above, the fastest path is to walk us through it directly. Request a demo and we'll show you what the embedded use case looks like end-to-end.
