Compliance first: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.
Home / Insights / Founder thesis
Founder thesis

"Every Industry Is a Geek Away From Being Uberized": The Founder's Thesis on Operational AI

"Every Industry Is a Geek Away From Being Uberized": The Founder's Thesis on Operational AI

OpenDataPolicies founder Jeff Cline argues that every industry is one technical builder away from disruption, and that the fuel for that disruption is operational data. Here is the thesis, the evidence behind it, and what it means for business owners who hold that data today.

"Every industry is a geek away from being uberized." — Jeff Cline, founder of OpenDataPolicies.org. The line is short on purpose. It says that no industry is protected by tradition, regulation or complexity forever; sooner or later, someone technical figures out how to rebuild it in software. What has changed in the AI era is what that someone needs to succeed. It used to be an app and a lot of venture capital. Increasingly, it is data showing how the work is really done.

By OpenDataPolicies Editorial · November 11, 2026 · 6 min read

What "uberized" means here

"Uberized" has become shorthand for an industry being reorganized around software: a coordination layer that matches demand to supply, prices work, routes it and measures it, often without owning the underlying assets. Taxis were the famous case, but the pattern is general.

In the founder's thesis, the next version of this is not an app sitting on top of an industry. It is AI agents doing the operational work itself: answering the support ticket, scheduling the job, drafting the change order, chasing the invoice, triaging the request. The "geek" in the quote is whoever builds the agent. The question every owner should ask is: what will that builder train on?

Why the geek needs operational data

Large language models learned to write by reading the public web. But knowing how to write is not the same as knowing how a dental front office handles a same-day cancellation, how a dispatcher reshuffles a plumbing route when a job runs long, or how a bookkeeper closes the month when a client is late with receipts.

That knowledge lives in operational records:

  • Email and chat that show how requests are clarified and escalated.
  • Tickets that show problem, conversation and resolution together.
  • CRM, project and job records with natural outcome labels: won, lost, done late, reopened.
  • Industry software such as ServiceTitan, Procore, Clio or dealer management systems.

This is what we mean by operational data for machine learning, and it is fundamentally different from web text. We compare the two in operational data vs web data.

The industry is already saying it out loud. OpenAI co-founder Ilya Sutskever told NeurIPS in December 2024, "We've achieved peak data and there'll be no more," and predicted that next-generation models will be "agentic in real ways" (The Verge). Agents need examples of work, and the public internet has very few.

Unique data is king

The second half of the thesis is about who captures the value. Jeff Cline's view is that unique people and unique data are the future, and that whoever holds the keys to unique data profits while it is still unique.

That has a few consequences:

  • The open web is the commons; operational data is the private reserve. Public text is shared, increasingly restricted and largely already used. Researchers behind the Consent in Crisis study found that in a single year, about 5% or more of all tokens in the C4 dataset, and 28% or more of its most critical sources, became fully restricted by robots.txt (arXiv).
  • Synthetic data cannot fully replace the real thing. A 2024 Nature paper showed that models trained recursively on generated data degrade, a phenomenon the authors called model collapse (Nature). Real-world records stay valuable as an anchor. See synthetic vs real operational data.
  • The holders are ordinary businesses. The richest operational records are not in Silicon Valley. They sit in the systems of HVAC contractors, law firms, insurance agencies, dealerships and clinics' front offices.

So the same shift that threatens to uberize an industry also hands its incumbents something to sell.

Two ways to respond

An owner looking at this thesis has broadly two options, and they are not mutually exclusive.

1. Get paid for the data while it is scarce. Licensing a de-identified copy of operational records lets a business collect value from the AI transition rather than only being disrupted by it. According to founder Jeff Cline, small companies have gotten over $100,000 for their operational data. Results vary; not a guarantee.

2. Use AI in your own operations. The same data that trains outside models can help you understand your own workflows: where tickets stall, where jobs run over, where cash gets stuck.

What the thesis argues against is doing nothing. Data that is unique today may not be unique in a few years, once enough similar businesses have licensed theirs and models have learned the patterns. That is why the founder stresses urgency: the window is limited, and once models are trained, the value of this data drops. See why the AI data window is closing.

Which industries are closest to the geek

Every industry is in scope, but buyers ask most often for industries with high volumes of repeatable, text-heavy workflows. Each industry page below goes deeper. A few examples:

In each case, the agent builders' gap is the same: they need to see how the work is actually done.

What the thesis does not say

It is worth being precise about the limits of the idea.

  • It does not say every business will disappear. Uberization reorganizes industries; it rarely eliminates the underlying work. Plumbing still needs plumbers. What changes is who coordinates the work, who captures the margin and which tasks software handles.
  • It does not say the change happens overnight. Some industries move quickly; regulated and relationship-heavy ones move more slowly. But slow is not never.
  • It does not say owners are powerless. The central point is the opposite: the people who hold the operational records hold something the builders need.

Seen this way, the thesis is less a prediction of doom than a statement about leverage. Leverage is worth most before the other side has found an alternative.

Doing it without selling out your customers

Uberization stories often come with a privacy cost. This one does not have to. The rule is non-negotiable: we never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Buyers want the pattern of the work, not the identity of the people. Names, contact details, account numbers and other identifiers are removed or replaced before anything is transferred. Our de-identification techniques explainer covers how, and the data rights checklist covers what to confirm first. That material is informational, not legal advice.

How a business acts on the thesis

The practical path is short:

  1. Inventory your systems and how much history each holds. The data readiness checklist helps.
  2. Prepare a header sample: field names and counts, no content. See what a header sample is.
  3. Request bids. According to founder Jeff Cline, bids typically arrive within 72 hours if the data is correct and a header sample is provided, ODP seeks offers from multiple buyers so you can take the highest bid, and deals typically close in about 90 days. Results vary; not a guarantee.
  4. Decide. You can say no to any offer.

The geek is coming to every industry. The owners who hold the data get to decide whether they are only disrupted or also paid. Start with a free bid request.

Our compliance promise: We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Frequently asked questions

Who said "Every industry is a geek away from being uberized"?

Jeff Cline, founder of OpenDataPolicies.org. The line sums up his view that every industry will be reshaped by technical builders, and that operational data is the fuel for that change.

What does uberized mean for AI?

It means an industry being reorganized around software, now increasingly AI agents that do operational work such as scheduling, support and billing. Those agents need training data showing how the work is done.

Why is operational data valuable to AI builders?

Because it shows real workflows, decisions and outcomes that do not appear on the public web, and the public web itself is being used up and increasingly restricted.

Does selling data mean giving up customer privacy?

No. We never take HIPAA/PHI data or personal customer data. Everything is de-identified before it leaves your control.

Sources

  1. The Verge — Ilya Sutskever on peak data (Dec 2024)
  2. Longpre et al. — Consent in Crisis (arXiv 2407.14933)
  3. Shumailov et al. — AI models collapse when trained on recursively generated data (Nature, 2024)

Updated November 19, 2026.

OpenDataPolicies.org connects businesses with buyers of de-identified operational data. We are not a law firm, broker-dealer or investment adviser, and nothing on this site is legal, tax or financial advice. Statements about amounts, timing and outcomes are Jeff Cline's statements based on past transactions. Results vary; nothing here is a guarantee of any bid, price, buyer or closing. Every deal depends on the data, its quality, its rights and buyer demand at the time.