Monetize your company’s data
Your engineering records can earn new revenue while advancing frontier AI.
01
New revenue from files you already have
Your existing records of engineering work become training environments for AI research labs, and all of your private information stays protected.
You get paid for data that shows how your work was done: from a customer requirement through revisions, reviews and shop-floor feedback to a released, manufacturable part.
No upfront risk
No buyer sees your files until you agree to license them.
No fees
You pay no fees.
You stay in control
You choose the projects and approve every package.
02
3 steps to start earning.
You tell us what you have and approve the packages. We handle everything else, from data preparation to lab negotiations.
- 1
Tell us what you have.
Fill in a short form. No files needed. We reply within 3 days with what is in scope and an estimate.
- 2
We find a buyer.
Labs see a de-identified summary without your name. Nothing is signed until a lab wants your data.
- 3
Sign, prepare, get paid.
We sign an NDA and licence, de-identify your files, and you approve every package. You are paid when a lab accepts it.
03
What data can I license through Phasing?
We work with engineering and design firms, contract manufacturers and product companies with roughly 20 to 300 people. You should have 50 or more completed projects in the last three years and hold the rights to the design files.
We look for the following files that show your work history:
- 01Native CAD with feature history: SolidWorks, Inventor, Fusion 360, Creo, NX, CATIA, Onshape
- 02The requirement that started the project: RFQ, customer spec, sketch, email, meeting note.
- 03Revisions: every saved version between first draft and release, with change notes, ECOs, review comments, DFM feedback from the shop floor.
- 04Drawings with tolerances and the final released state.
- 05Outcomes: first-article inspection result, customer acceptance, rework or scrap notes, field returns.
- 06CAM programs paired with the model, simulation setups and results, assembly BOMs.
04
How your data is protected
De-identified first
Names, emails, phone numbers, customer and supplier details, prices and anything under a third-party NDA come out before a buyer sees anything.
Encrypted and deleted
Your data is encrypted while we work on it, and we delete the originals once processing is done.
Not sold to competitors
It is not published or sold to your competitors.
You stay in charge
You approve every package and can leave out any system, folder or record type.
Background reading
Independent reporting on why AI labs pay for real work records.
- AI's New Training Data: Your Old Work Slacks And Emails ↗Forbes, April 2026
- Silicon Valley bets big on 'environments' to train AI agents ↗TechCrunch, September 2025
05
FAQ
How is the price set?
Every licence is negotiated individually. Labs pay most for real design tasks across several work steps with a known result: requirement, revisions, drawings and outcome. More completed projects and more years of history raise the price.
What is the catch?
Getting in touch commits you to nothing, and nothing is signed unless a lab wants your data. You keep ownership, approve every deal in writing, and can leave out any system, folder or record type. Licences are usually non-exclusive and typically run one year, so you can license the same data again afterwards.
How is our data protected?
Before a lab sees anything we de-identify it: names, emails, phone numbers, customer and supplier details, prices and anything under a third-party NDA come out. Data is encrypted, and we delete the originals once processing is done.
How much work is this for our team?
Under two hours: a short form, one inventory call, and approving packages. We handle de-identification, packaging and lab negotiations. Your files and documents do not need to be in English.
When are we paid?
When a lab accepts a package. No advance payments. Every deal is different, so we do not promise timelines.
What do AI labs use the data for?
Labs train models on tasks with a checkable answer. The model attempts each task many times, and attempts closer to the correct result are reinforced. Your data teaches models how engineering work moves from requirement to released part.
Phasing was founded by serial founders and engineers, previously in management roles at McKinsey and SAP. We work with a team of highly qualified AI researchers and reinforcement learning engineers.
Your company’s data could be worth $100k to $2m.
Fill in the form. We reply within 3 days.
