Sell architecture firm data to AI companies

Why AI companies want architecture records right now

AI labs are training agents for professional work that runs on long, coordinated projects, and architecture is a textbook example. A single building involves a client, consultants, contractors, code officials and a design team, each exchanging drawings, questions and decisions over months or years. Design tools are adding AI that drafts, checks and coordinates models, and agents are being built to handle the correspondence and project management around them.

What those systems lack is a view of how a real practice works. Finished buildings are photographed and published; the process is not. How a schematic design changed after a client meeting, how an RFI from the contractor was answered, how a fee was scoped and tracked against hours – that record lives only inside the firm. That is why architecture firms are a good fit for AI data licensing today.

What carries the value

The core of a deal is the firm's operating record across its five main systems. CAD and BIM files come on top: a Revit model is far more useful when it arrives with the email thread where the client asked for the change, the RFI log and the hours the change took. That linked record is what AI companies pay for, which is also why leaving whole systems out lowers the offer. For more on the files themselves, see CAD and engineering data.

Typical starting size

Bigger teams get bigger offers. For architecture firms, count the architects, designers, project managers and office staff who work in the firm's systems every day. A mid-sized practice with many years of projects often has a deep record; smaller firms can still describe their data and get an honest answer on whether it fits.

History is important here because projects run long. Three years of records still in the systems is the usual floor, and five or more is better. Internal communication should mainly be in English.

The self-check guide goes through each point, and the industry overview shows how architecture compares with other sectors. Firms that also do construction management may want to read construction company data as well.

What is assessed first, and what is never part of a deal

• Client ownership of drawings. Copyright in drawings and models depends on each contract. Under AIA B101 the architect typically keeps it, but owner-drafted agreements often assign it to the client. Who owns CAD files explains how this generally works. Projects where the client owns the work are left out.
• Confidential and security-sensitive projects. Government, defense, critical infrastructure and some private projects carry confidentiality or security terms and are normally excluded.
• Third-party content. Consultant drawings, manufacturer families and licensed content libraries belong to others and are filtered out.
• Personal data and HR. Names, contact details, personnel matters and private conversations are filtered inside email and chat, and identifiers are removed before delivery.
• Payment details. Bank and card numbers are removed from accounting and email.

This describes how it generally works today, not legal advice. Every firm and every deal is unique, and what can be included is assessed case by case.

What companies typically receive right now

Smaller companies in high-demand industries are quoted individually. These are indicative ranges from what we see in the market right now, not offers – the real number comes in a written offer after a review, and depends on your industry, years of history, which systems are included and demand at the time. Our data value estimator gives a first estimate in about two minutes.

How a deal works

1. You describe the practice – team size, years of projects, the systems you use and roughly how your contracts handle drawing ownership. No files are uploaded.
2. We assess ownership, project restrictions and demand, and get back to you within 48 hours on business days.
3. You approve a written offer – or say no. Nothing is shared without your approval.
4. Filtering and delivery. Client-owned, restricted and third-party material is filtered out inside each system, identifiers are removed, and the agreed record is delivered.

The license is non-exclusive by default and limited in scope, time and use. Having your firm reviewed is free – we're paid only when a deal goes through. More on how it works. For the full picture across industries, see how to sell data to AI companies.