The Sakuya journal

A plain‑English look at the new market connecting ordinary companies with the AI labs that want their records.

The public internet taught models to talk. Company records teach them to work.

Frontier labs are only one part of the demand. Here is who else is buying.

Volume alone does very little. These are the factors that move a deal.

Labs want to learn how work happens, not who did it. Here is how personal details come out.

The structure of a deal matters as much as the price. Know the difference before you sign.

Team size is a shortcut for the depth of records a company tends to hold.

Not exotic datasets. The everyday files your team already makes.

Engineering and design records are some of the hardest material for labs to find.

Documents show what was decided. Messages show how.

Every resolved ticket is a worked example of solving a real problem.

Our average deal size, and what tends to move a company toward the top of the range.

A simple checklist to know what you hold before your first call.

Before any deal, confirm the records are yours to license.

Labs do not issue a public spec for company data. Here is why outreach stalls.

The step between your first meeting and your quote, explained.

Why the final step moves so quickly once the groundwork is done.

A bankruptcy auction showed that operating records from a non‑tech company are an asset class.

Two very different ways to enter the AI data market.

A short checklist for choosing who represents your records.

Agents need to see real work, step by step, before they can do it.

The simulated workplaces where AI agents practice, and why they need real company data.

Design files, production planning and supplier records are among the scarcest data AI labs can find.

Every shipment leaves a trail of decisions. That trail has value.

Drawings, specs and project correspondence show how built environments come together.

Briefs, strategy decks and campaign threads capture how creative work gets made.

Bids, schedules, change orders and field reports show how projects actually get built.

Firm process and internal workflow can have value. Client financials need special care.

Privilege and confidentiality come first. Here is how to think about it.

Patient information is protected. Operational records are a separate question.

Code, tickets and engineering discussion show how software actually gets built.

Support threads, merchandising plans and operations data all tell a story.

Transactions involve dozens of steps. The records of those steps have value.

Process records are valuable. Candidate personal data requires care.

Quoting, servicing and claims workflows show how complex decisions get made.

HVAC, plumbing and electrical companies hold years of real‑world problem solving.

Email is still where much of the real work of a company happens.

Boards, tickets and task histories show how work moves from idea to done.

Pipeline records show how deals are won, lost and managed over time.

Years of proposals, plans and SOPs are a record of how your company works.

Every ticket is a worked example of diagnosing and fixing a problem.

Commits, reviews and bug reports show how software gets better over time.

Notes capture decisions. Transcripts raise extra consent questions.

The models a company runs on show how it thinks about money and operations.

Your documented processes are a manual for how work gets done.

Transparency builds trust. Here is how to frame the conversation.

It depends on your contracts and your relationships. Here is how to decide.

The simple idea behind why your records have value.

Years of records show change, and change is what labs want to learn from.

A ticket that links to a conversation and an outcome tells a complete story.

Common misconceptions that stop owners from exploring a deal.

Generally, yes, within your obligations. Here is what to review.

Ten minutes of preparation that saves days later.

Quality is less about polish and more about completeness and context.

Why invented examples cannot fully replace the real thing.

The open internet taught AI to talk. It cannot teach it everything.

What happens on the lab's side during vetting.

A pass from one lab is not the end of the road.

It depends on the terms of each deal.

A key term that shapes both price and future options.

From first call to payment, here is the typical timeline.

A little preparation makes anonymization cleaner and faster.

The final steps between signature and payment.

Timelines, terms and what affects them.

Licensing revenue is new for most companies. Get advice early.

Decades of history can be a quiet asset.

Scale makes evaluation easier for labs.

The middle of the market holds a large share of real operating knowledge.

Where licensing for company records may be heading.

Plain definitions for the words you will hear in a data deal.