# Buy expert domain data from professional firms

URL: https://sellmydatatoai.net/buy-ai-training-data/expert-domain-data

Expert domain data is the working record of professionals such as lawyers, accountants, engineers and clinicians: case files, workpapers, drawings and notes that show expert judgment applied to real problems. Mercor's published listings put company payouts at $50,000 to $1.2 million for law firms and $25,000 to $865,000 for accounting and tax firms.

## Key takeaways

- AI teams use expert domain data for (1) fine-tuning models for professional tasks, (2) building expert evaluation sets and rubrics and (3) agent training on professional workflows.
- Usually priced per company archive, as a flat license fee.
- Common delivery formats: Document sets in PDF, DOCX and native formats, structured exports from practice management systems and JSONL of tasks with expert reference answers.
- Check before you license expert domain data: (1) US: HIPAA de-identification for medical records, (2) US: IRC Section 7216 limits on tax return information and (3) client confidentiality, attorney-client privilege and professional duties.
- Every seller we shortlist passes 8 vetting checks before you see a sample.

## Buyer's spec

- Used for: Fine-tuning models for professional tasks; Building expert evaluation sets and rubrics; Agent training on professional workflows; RL environments for drafting and review tasks
- Formats: Document sets in PDF, DOCX and native formats; Structured exports from practice management systems; JSONL of tasks with expert reference answers; CAD and BIM files for engineering and construction
- Usually priced: Per company archive, as a flat license fee
- Check before you sign: US: HIPAA de-identification for medical records; US: IRC Section 7216 limits on tax return information; Client confidentiality, attorney-client privilege and professional duties; Privileged and regulated material excluded; Credentials of authors and reviewers; Files complete from instruction to outcome; Client contracts that restrict reuse; Clients and counterparties de-identified; UK and EU: client records are personal data under UK GDPR or EU GDPR unless anonymized

## What is expert domain data?

Expert domain data is the record of skilled professional work, kept by the firms that did it. It includes closed matter files at law firms, tax returns and workpapers at accountancy practices, project records and drawings at construction companies, and de-identified clinical notes at medical practices.

Labs buy it to teach models complete professional tasks, from instruction to finished output. MIT Technology Review reported in October 2026 that Surge AI builds reinforcement learning gyms where agents practice tasks such as drafting appellate briefs.

## How much do AI companies pay for expert domain data?

AI companies pay from tens of thousands of dollars to over $1 million per firm for expert domain data, depending on the profession and the size of the archive. Mercor's published listings, compiled by ddiy.co in October 2026, give these company payouts.

| Profession | Records licensed | Company size | Published payout |
| --- | --- | --- | --- |
| Law firm | Complete files from closed matters and completed deals | 5 to 300 attorneys and staff | $50,000 to $1.2 million |
| Accounting and tax firm | Tax returns, workpapers and bookkeeping records | 5 to 200 employees | $25,000 to $865,000 |
| Construction company | Project records, drawings and site photos | 50 to 2,000 employees | $115,000 to $1 million |
| Manufacturer | Production, maintenance and inspection records | 20 to 2,000 employees | $100,000 to $1.5 million |
| Medical practice | De-identified patient records with physician notes | 10 or more clinicians or 2,500 or more patients | $80,000 to $820,000 |
| Consulting firm | Client deliverables, models and drafts | 10 to 500 employees | $150,000 to $300,000 |

These are payouts to the firms supplying data, not buyer prices, and each range spans small and large firms. The Next Web reported that micro1 has committed $1bn over 12 months to buying and licensing operational company data, a sign of the competition for good archives.

Firms on the supply side can read our pages for [law firms](https://sellmydatatoai.net/sell-your-data/law-firms), [accounting and tax firms](https://sellmydatatoai.net/sell-your-data/accounting-firms) and [medical practices](https://sellmydatatoai.net/sell-your-data/medical-practices).

## What rights issues affect expert domain data?

Expert domain data carries professional duties as well as privacy law. Client files sit inside a relationship governed by confidentiality, and often by contract terms that restrict reuse.

Our seller vetting covers contract restrictions, the privacy notice in force when the records were made and the de-identification method, so shortlisted firms have already cleared those checks. You still need independent legal advice on your own intended use.

- Medical records: HIPAA allows de-identification by expert determination or by removing 18 listed identifiers under 45 CFR 164.514, and HHS says de-identified health information is no longer protected health information
- Tax files: US preparers face IRC Section 7216, which prohibits knowingly or recklessly disclosing or using tax return information
- Legal files: attorney-client privilege and client confidentiality
- Engineering drawings: client ownership of designs under some services contracts

If you are buying UK or EU records, client files are also personal data under UK GDPR or EU GDPR unless anonymized, and health records are special category data.

## How should expert domain data be delivered?

Expert domain data should be delivered as complete files grouped by matter, engagement or project, with documents in native formats where possible and a manifest listing each file's type, date and role in the workflow. Drafts add value: Mercor's consulting listing, for example, covers client deliverables and the drafts behind them.

Ask for a short note of each file's outcome, such as the matter result or the audit opinion, so the record can be turned into tasks with known answers.

## What should you check before buying expert domain data?

Before buying expert domain data, check that each file runs from instruction to outcome and that the authors held the relevant qualifications. Our guide to [data provenance](https://sellmydatatoai.net/guides/data-provenance-explained) lists the records that prove where each file came from.

Ask how clients and counterparties were de-identified, who reviewed the result and whether any files were excluded for privilege or regulation. Every seller we shortlist passes our [eight-point vetting standard](https://sellmydatatoai.net/how-we-vet). [Book a free consultation](https://sellmydatatoai.net/book-a-consultation?side=buyer) to source vetted professional archives.

## Frequently asked questions

### Why do AI labs want professional firm records?

AI labs want professional firm records because they show expert judgment applied to real problems from instruction to finished output. MIT Technology Review reported that Surge AI uses such data to build gyms where agents practice tasks like drafting appellate briefs.

### Can law firm files be licensed for AI training?

Law firm files can be licensed for AI training only after privilege, client confidentiality and contract restrictions have been cleared, and client identities removed. Mercor publishes payouts of $50,000 to $1.2 million for complete files from closed matters at firms of 5 to 300 attorneys and staff.

### Can tax returns be used as AI training data?

Tax returns can only be used as AI training data within the rules on tax return information. In the US, IRC Section 7216 prohibits preparers from knowingly or recklessly disclosing or using tax return information, with fines of up to $1,000 per violation. Take specialist advice before licensing tax files.

### How is medical data de-identified for AI training?

Medical data is de-identified for AI training under HIPAA in the US by one of two methods in 45 CFR 164.514: expert determination that re-identification risk is very small, or the safe harbor method of removing 18 listed identifiers. HHS says de-identified health information is no longer protected health information. In the UK, health data is special category data under UK GDPR.

## Sources

- [How to sell company data to AI companies (2026 guide)](https://ddiy.co/sell-company-data-to-ai/), ddiy.co, accessed November 2026
- [AI labs are buying data from dead startups to train their models](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models), MIT Technology Review, accessed November 2026
- [micro1 commits $1bn to buy company data for training AI agents](https://thenextweb.com/news/micro1-1bn-enterprise-data-ai-training-citi-hercules), The Next Web, accessed November 2026
- [Section 7216 frequently asked questions](https://www.irs.gov/node/3756), Internal Revenue Service, accessed November 2026
- [Guidance regarding methods for de-identification of protected health information](https://www.hhs.gov/hipaa/for-professionals/special-topics/de-identification/index.html), US Department of Health and Human Services, accessed November 2026

---

Book a free consultation: https://sellmydatatoai.net/book-a-consultation