What the human-data boom signals for company data and private equity portfolios

What does the growth of AI data companies tell investors?

It tells investors that AI developers now pay for human-made data as a core input, and that what they want is moving closer to real work. As AI shifts from models that answer questions to agents that carry out multi-step tasks, developers need records of how work is actually done: tickets and their resolutions, deal histories with outcomes, approvals and exceptions. For a private equity owner, that makes operating records already sitting inside portfolio companies a potential licensable asset, with no new product build and no dilution.

Vendors that recruit domain experts to write, label and grade training data have drawn wide business-press coverage. This page does not repeat vendor fundraising or valuation figures, because we have not verified them against primary filings. For an investment committee, the more useful evidence is what developers have paid for on the record, and why.

What does the dated public record show?

The table pairs two early, documented content-licensing agreements with two 2026 private equity reports on why new value levers matter. It describes what was announced or reported, not what any company's records would fetch.

SourceX is not a party to either content agreement; both are cited only as public market evidence. Disclosed figures from listed data holders are collected in which public companies disclose AI data licensing revenue.

Where is AI data spend going?

It helps to separate AI data spend into three layers, because only one of them involves portfolio companies directly.

The middle layer is what most headlines describe; its building blocks are explained in what human-in-the-loop data is. Vendor-written tasks are reconstructions: a contractor imagines a procurement exception and writes how it should be handled. An operating company's records show the exceptions that actually happened, in sequence, across years, with the outcome attached. That authenticity is why developer spend reaches the third layer. The full path from record to model is mapped in the AI data supply chain.

Which portfolio companies are likely to fit?

Fit depends on how work is recorded more than on sector glamour.

Industry-by-industry detail sits in vertical AI companies and the industry data they need.

The three-signal portfolio scan

Run each portfolio company through three signals plus the baseline at the next portfolio review. If any line is a clear no, set the company aside for now.

• Workflow depth: records sit across many connected systems (strong companies often run 10-15+), with archived systems and five to ten years or more of history.
• Outcome labels: cases end in a visible result, such as a ticket resolved, a deal won or lost, a claim paid or an approval refused.
• Clean rights and a sponsor: the company created the records, its contracts and notices allow licensing, and the CEO, CFO or owner will sponsor the process.
• Baseline: a US company with 50+ full-time employees at peak (contractors excluded) and several years of documented operations.

To test one company before the review, use the company fit checker, a preliminary screen that needs no contact details and commits no one.

What the human-data boom does not tell you

A hot vendor market is a signal, not a valuation input for your portfolio.

• Vendor valuations do not price any company's records. There is no public price list; each company agrees its own price and terms.
• Demand can shift as developers change methods, and synthetic data fills some gaps.
• A license is typically a one-time payment for an exclusive AI-training license over an agreed term, so treat it as non-recurring in any plan.
• Companies whose data mainly belongs to clients, is mainly consumer personal data or PHI, was deleted, or was already licensed for AI training are unlikely to qualify.
• No deal is guaranteed, even for a company that screens well.

How does an operating partner make the introduction?

The operating team opens the door; the portfolio company and SourceX do the work.

1. Raise the idea with the portfolio CEO at an operating review or board pre-read, and get the CEO's agreement to explore it.
2. Register, then send the CEO your referral link or enter the company in the referral form yourself.
3. SourceX qualifies the company with its sponsor on size, history, data breadth and rights.
4. The company builds a data inventory and agrees de-identification and redaction rules before any preparation begins.
5. Price and terms are agreed, AI labs and data buyers review the opportunity, and the company signs only if the terms work.
6. After delivery, the company receives a one-time payment, typically within about 60 days of invoicing once the buyer selects the data.

Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company, and the reward becomes payable only after the buyer pays and SourceX receives its fee. Because the reward is a share of SourceX's fee, the portfolio company's proceeds are untouched. Check your firm's policies and fund documents on fees connected to portfolio companies first; the private equity operating partner page covers the role in more depth.

Next step

Add the three-signal scan to your next portfolio review and pick the company with the deepest records. Register as a partner to introduce it, or ask the CEO to apply directly at sourcex.si/apply with your referral link.