■ Data licensing for AI
The internet is secondhand.
Your records are firsthand.
AI labs have trained on what people say about work. What they can't get is the original record of work actually being done: your emails, tickets, deals and docs. That's worth $25K to $2M+, and we license it for you.
You approve every deal · Nothing leaves until a contract is signed
Rec T-0417Source ZendeskStatus Raw
Shipment to Holloway GrocersCOMPANY_08 missed the dock window again. Third time this month.
Carrier cutoff moved to 2pm. Rebooking on the morning run and crediting the fee. Call me at (415) 555-0192PHONE.
Moved to AM pickup. 0 misses in 6 weeks. Card ending 4417 refunded.
View
Every record is reviewed and redacted before release. Names become tokens, contact and payment details are removed, and the work itself stays intact.
01 / Why it's valuable
Labs have the commentary. You have the source.
Every public page has already been scraped. Models learn the most from the real thing: how a problem was actually handled, step by step, by people who do it for a living.
Secondhand · already scraped
Firsthand · your records
A blog post on handling angry customers
2,400 resolved support tickets, start to finish
A forum thread on negotiating renewals
The actual email chain behind a $400K renewal
A tutorial on month-end close
Five years of your close checklists and the fixes in them
Advice on running a sales team
Every deal note, won and lost, with the reasons why
A Stack Overflow answer
The pull request, the review and the incident that caused it
02 / What we license
What's in your files
The tools your team already uses hold years of decisions, fixes and expertise. Open any file to see a made-up example of a cleared record.
Never released:PasswordsCredentialsLive system accessPayment card dataAnything you exclude
03 / Your terms
It's been off the record for years. Put it on the record.
On your terms, with your approval, and only after it's been cleared.
You own it
You license the data and keep ownership. Non-exclusive by default, so you can license it again.
You approve it
You see every buyer, every use and every price before anything is signed. You can say no.
We stand in between
You never deal with AI labs directly. We vet buyers, negotiate terms and handle contracts.
04 / Process
Four steps. One signature.
About two hours of your time. We handle the rest.
0130 min
Intake
A short call about your company, your tools and how far back your records go.
Received02~1 hr
Verify
Send record counts and screenshots. No raw data changes hands.
Verified03You decide
Approve
We bring vetted buyers. You set scope, exclusions, allowed uses and price.
Approved04After signing
Clear & release
We redact and review every record, deliver it, and pay you your share.
PaidRule Your records stay in your systems until a buyer signs. Buyers only see cleared samples first.
05 / Appraisal
What are your records worth?
Three questions for a rough range. More people, more years and more systems mean more of what buyers need.
Estimatenot an offer
Appraisal summary
Company size
History
Systems
Illustrative range
$0
Real value depends on your industry, data quality, current buyer demand and the terms you choose.
Get a real appraisal →Nothing is sent. This stays in your browser.
06 / In demand
Expert records are the rarest
The more skilled the work, the harder it is for labs to find, and the more it's worth.
07 / For buyers
Firsthand data, with a paper trail
Proprietary records from operating companies. Clear rights, consistent redaction, and documentation your legal team can sign off on.
- Pre-training
- Long-tail professional language and workflows that never reach the open web.
- Fine-tuning
- Threads from problem to resolution that show how real work gets done.
- Evaluation
- Fresh, held-out tasks from real businesses your model has never seen.
- Synthetic
- Realistic scenarios to ground synthetic data and agent environments.
Every dataset ships withForm D-1
- Rights
- Written license from the data owner, with scope and term
- Redaction
- Consistent tokens for people and companies; contacts removed
- Datasheet
- Source systems, date range, volume and known gaps
- Sample
- Cleared sample to review before you commit
- Format
- JSONL or Parquet, threaded and timestamped
- Exclusivity
- Non-exclusive by default; exclusive on request
08 / Questions