The Economics of Agentic AI: Rethinking Value of Data in a Non-Linear World

Why the death of the "workflow" is breaking our revenue model—and how to fix it.

For the last decade, we made money from a mistake.

We didn't call it a mistake. We called it a "Big Data business model". It is simple: we sold the haystack. If a customer needed a specific answer, they had to buy the pile to find the needle.

It was a good trade for us.

The Crash of the Process Economy: We used to charge for the assembly line. The Agent didn't just speed up the line; it dissolved it

I saw this firsthand. Years ago, at a global coffee retailer, we bought customer data from Experian. We paid by the record. Volume was the metric. We bought the rows, then we paid again to pipe them to LiveRamp.

It was a long, expensive, linear assembly line.

Later, running my own company, I felt the pain from the other side. I needed to find new customers. Every startup pundit told me to "define my ideal customer profile (ICP)". But they never told me how to execute it.

Try translating a feeling into a filter for a marketing team. It is a nightmare.

I had to translate my intuition into crude proxies: "Show me companies with $50M in ARR in zip code 12345."

I don't want a zip code. I wanted an ICP. I wanted a feeling. I wanted an intent.

But the current systems only understood filters. So, I bought the list, filtered it myself, and threw more than half of it away. And yes, I lost my shirt on Google Ads because of it.

The Big Data industry is built on monetizing that waste.

The End of the Filter is Near

Agentic AI breaks this model because it replaces filters with intent.

Today, I don't want to select checkboxes. I tell the Agent, "Find me CxOs in the Pacific Northwest who are worried about coffee supply chain risks."

The Agent does not download a million records. It scans the universe of data non-linearly. It reads news reports, checks financial filings, cross-references LinkedIn posts, and returns the twelve names that matter.

It skipped the download. It skipped the cleaning. It skipped the waste.

And if your pricing is attached to the volume of the download, then your revenue skipped town too.

The Economics of the Crash

This is not a sales slump. it is a fundamental break in Value Capture.

The agentic AI creates immense value. It saves the user hours of manual filtering. It delivers a higher quality result. But our pricing mechanism - "big data business model" - fails to capture that value.

We are seeing a brutal inversion of "The Javons Paradox".

In 1865, economist William Stanley Jevons observed that as steam engines become more efficient, coal consumption went up not down. Efficiency drove volume.

We bet our business that AI would do the same - that making data easier to use would lead to more downloads.

We are wrong.

Traditional efficiency (automation) meant doing the same steps faster. Agentic efficiency means skipping some of those steps entirely.

The Agent drives the transaction costs of search to near zero. We are witnessing the collapse of what economists call "Transaction Cost Economics". Nobel laureate Ronald Coase taught us that firms exist to manage the costs of finding, bargaining, and transacting. Your data business charged for the friction of finding information.

The Agent just removed (abstracted) the friction.

The Path Forward

This leaves us with a dangerous question:

If the customer is getting more value (a perfect answer) but consuming less product (fewer bytes), what are we actually selling?

Are we selling the raw material? Or are we selling the result?

And if we are selling the result, how do we price it?

The customer who asks that strategic question about CxOs has a high Willingness-to-Pay. They are getting a competing advantage. But our current model charges them pennies because the file size is small.

We are applying linear pricing (per record) to a non-linear asset (intelligence).

What Comes Next

We know the old math is broken. But what is the new math?

In my next blog in this series, we will dismantle the most dangerous metric in your P&L: Volume.

We will use Microeconomic principles to map the "Consumer Surplus Trap" that is leaking revenue from every AI query. We will build Econometric models to calculate the exact dollar value of a decision, rather than the cost of a row. And we will structure Two-Part Tariffs to capture that value without killing adoption.

We are going to stop asking, "What does this data cost to produce?" and start calculating, "What would they pay to not live without it?"

This is not a lecture. It is an open experiment.

I am rebuilding this model in real-time, and I want you in the lab with me. If you are tired of trading dollars for pennies, follow along. Let’s engineer a pricing model that survives the future.

The Big Data assembly line is closing. The non-linear age is here.

It is time to price like it.