On October 9, San Francisco AI training company micro1 announced a plan to spend $1 billion over the next 12 months acquiring and licensing operational data from companies. The program, called Company Data Partnerships, is backed by capital from Citi and Hercules Capital.
The proposed deals are aimed at US companies with at least 30 employees and documented workflows. Materials may include standard operating procedures, knowledge bases, CRM records and internal documents. The aim is to turn that information into environments where AI agents can practice navigating workflows and making business decisions. Advertised compensation starts at $100,000 and can exceed $1 million, depending on the data’s quality, uniqueness and value.
Those figures are not a guaranteed payout for every business. But the announcement points to a bigger shift: the know-how embedded in ordinary company processes is becoming something the AI market may be willing to pay for.
That should prompt a new question for owners and operators: What knowledge has your business quietly built, and what would it take to share it safely?
Your processes hold more than instructions
A procedure document can look like a humble checklist. But it may capture how an experienced team handles an unusual order, resolves a support problem or decides when an invoice needs extra review. A CRM history might show how sales conversations move from first contact to renewal. A project archive can reveal how teams adapt when plans change.
This is valuable because real work rarely follows a perfect script. People work with incomplete information, exceptions and competing priorities. A well-organized record of how a company handles those situations can hold lessons that a generic template cannot.
That does not mean every spreadsheet is a gold mine. Data needs to be relevant, clear enough to interpret, and legally and ethically appropriate to share. But it does mean that operational knowledge deserves a more deliberate place in business strategy.
For years, many companies have treated process documentation as an internal housekeeping task: useful when onboarding staff, preparing for an audit or keeping the day moving. Now it may also be intellectual property with commercial value.
First, decide what you would actually be selling
“Selling your data” can sound like handing over the keys to your business. In practice, a deal could involve licensing certain material for a defined purpose, providing a prepared dataset, or sharing examples of workflows under agreed limits. The distinctions matter.
Before considering an offer, decide what your company is willing to share and what must stay in-house. A useful starting point is to sort potential material into three groups:
- Potentially shareable: Established workflows and process documents that don’t expose sensitive people, customers or trade secrets.
- Shareable only after careful preparation: Records that include personal details, customer references, pricing, proprietary methods or other confidential information.
- Not for sharing: Information your company is contractually, legally or ethically obligated to protect, or knowledge that would create an unacceptable competitive risk.
Then ask whether a one-time payment is worth the trade. Could the same information support future partnerships, product development, staff training or your own AI tools? If your process is a key competitive advantage, unrestricted or perpetual rights could cost more than the immediate payment is worth.
Licensing can sometimes preserve options, but the label alone is not enough. The contract determines what a buyer can do, for how long, with which materials, and whether it can share or reuse what it learns. Have qualified legal and privacy advisers review proposed terms before you sign.
Run a practical data audit
You don’t need a sophisticated data warehouse to begin. Start with a small inventory of where operational knowledge lives:
- List your sources. Include shared drives, knowledge bases, CRM and ticketing systems, project folders, training materials, templates and team-maintained checklists.
- Name the process, not just the file. “Customer onboarding” is more useful than “Q3-final-v7.” Record who uses each item, how current it is and what decisions it supports.
- Look for the real workflow. A polished SOP may describe the official process, while comments, revisions and examples reveal what people do when something goes off-script. Identify whether those records are appropriate to include.
- Mark ownership and obligations. Note whether content includes customer or employee information, third-party material, confidential terms or restrictions in existing agreements.
- Keep a clean source of truth. Assign an owner, date, version and access level to approved documents. Remove obsolete copies where appropriate, and restrict access to the originals.
A clear inventory helps even if you never license anything. It can make onboarding smoother, reduce dependence on one person’s memory and help your team spot inconsistent or outdated processes.
De-identification needs details, not assurances
Removing names is a start, not a complete privacy plan. People and businesses can sometimes be recognized from combinations of dates, locations, job titles, unusual events or transaction details. The more specific the material, the more important it is to understand what could still identify someone.
Before sharing, get precise answers to questions such as:
- Which fields will be removed, redacted or transformed, and who checks that work?
- Could someone reasonably infer a customer, employee or business identity from the remaining details?
- What exactly will the data be used for: model training, evaluation, reinforcement-learning environments or another purpose?
- Who can access the material, including contractors and service providers?
- How is it protected in transit, in storage and during processing?
- How long will the original and prepared copies be kept? What triggers deletion, and how is deletion confirmed?
- Can your company review representative samples before use?
- What rights apply to derived materials, such as synthetic rewrites, workflow representations or model improvements?
- Can the buyer reuse the material for new purposes or pass it to another party?
Also check whether your company has permission to share the information in the first place. Customer contracts, employee privacy commitments, industry rules and third-party licenses can all matter. A buyer’s safeguards do not replace your own obligations.
Ownership is not the same as control
A company can retain ownership of its underlying data and still grant another party meaningful rights to use it. That is why the contract should spell out the boundaries in plain terms: permitted uses, prohibited uses, duration, security standards, retention, deletion, audit rights and what happens if either side breaches the agreement.
Pay special attention to what counts as “data.” Does it include raw documents only, or also cleaned datasets, examples, derived workflows and future versions? Can the buyer use what it learns to improve systems beyond the specific project? Are there limits on combining your material with other sources?
These aren’t reasons to reject every opportunity. They’re the questions that help you distinguish a thoughtfully scoped partnership from a vague promise.
Treat your know-how like an asset, whether you sell it or not
The $1 billion commitment is a signal, not a verdict on what any one company’s data is worth. Some businesses may find a compelling opportunity. Others may decide the privacy, legal or competitive risks outweigh the potential payment. Both can be sound decisions.
Either way, the strategic move is to understand what you have. When teams document how work actually gets done, they protect expertise that might otherwise disappear when an employee leaves. They can make training more consistent, discover where processes break down and build a stronger foundation for responsible automation.
Your company’s everyday data may have commercial value. More importantly, it contains the record of how your people solve real problems. Start by making that knowledge visible, organized and governed. Then, if an opportunity comes along, you’ll be deciding from a position of clarity, not scrambling to figure out what you own.