Record Yield
UK company data licensing

Your company's records, turned into revenue.

AI companies pay to learn from real examples of everyday work, like support tickets and email threads. We help UK companies sell a cleaned, anonymised copy of theirs, safely and legally. You keep the originals.

No uploads or loginsDe-identified before transferYou approve every buyer
Your recordsIllustrative

What most teams already hold

@
Email and chatOutlook, Teams, Slack · how decisions get made
↗
✓
Support deskZendesk, Freshdesk · problems and their fixes
↗
≡
Documents and SOPsSharePoint, Drive · procedures and their revisions
↗
£
CRM and dealsSalesforce, HubSpot · pipelines with outcomes
↗

Start with a few questions. Nothing leaves your systems until you say so.

Your value

Get a rough number in 20 seconds.

Pick the systems you use, your team size and how far back the records go. It uses the same model as our full valuation. Systems count differently: see what you can sell below.

How far back your systems go8 years
Indicative first-year range
…
£10k£100k£1m£10m

A rough guide, not an offer. What buyers pay depends on your records, your rights to them and the licence you agree.

Get my full estimate →
How it works, step by step

From records to revenue, in four steps.

Scroll to follow one support ticket through the process. Your original records never leave your systems.

  1. 1. Your recordsYears of email, chat, tickets and documents on your own systems, full of names and contact details.
  2. 2. De-identifiedInside your systems, every name and contact detail is swapped for a consistent stand-in, then tested.
  3. 3. LicensedOnly the cleaned copy goes out. Buyers make offers, and you choose the terms you're happy with.
  4. 4. RevenueYou're paid, the originals never left, and you can license the same copy again.
Step 1 of 4 · Your recordsExample data
IN YOUR SYSTEMSSupport ticket #20417FromPeter JonesEmailpeter.jones@gmail.comPhone07700 900123IssueInvoice export failingFixReset template, re-ran exportPersonal detailsYears of email, chat,tickets and documents,sitting unused.CLEANED IN YOUR SYSTEMSSupport ticket #20417FromPeter JonesPerson AEmailpeter.jones@gmail.com[email: Person A]Phone07700 900123[phone removed]IssueInvoice export failingFixReset template, re-ran exportSame person,same stand-in3 IDENTIFIERS REPLACED✓ Testedmotivated intruder testBUYERS BIDCLEAN COPY → BUYERSCleaned copyBuyer A£240kBuyer B£360kBuyer C£500k ✓PAYMENT → BACK TO YOUOriginal recordsStill yoursLicense againLICENCE FEE£LICENCE FEE£LICENCE FEE£Paid: £500,000

Years of email, tickets and chat hold the real working language AI developers want, but they're full of names, emails and phone numbers, so nothing moves yet.

Scroll or drag
1 / 4
00 / The basics

AI training data, the fundamentals.

New to data licensing? This is the model in three parts: demand, supply and the licence. Every section below builds on it.

1

AI needs real examples

AI assistants learn to answer customers or follow a process by studying real examples of people doing it. The public internet has largely been used up.

2

Your company already has them

Years of support tickets, email threads, team chats and documents show exactly how real work gets done. Most companies never use them again.

3

They pay for a cleaned copy

Names and personal details are removed on your own systems first. An AI company then pays for a licence to learn from that cleaned copy. You keep the originals.

✓You keepyour original records, your customers and the right to sell again
✕Never sharednames, contact details, private messages, HR or health records
£You geta licence fee, typically four to seven figures, depending on size

Not a people-search data broker. We never build or sell profiles of individuals. What's licensed is how work gets done, with every person de-identified first.

You can sell it more than once. Licences are non-exclusive by default, so the same cleaned copy can earn from several buyers over the years.

Four words you'll see on this page
De-identified
Personal details replaced with stand-ins like "Person A", so no one can be identified.
Licence
Permission to use the cleaned copy for agreed purposes. You don't sell your business or your data outright.
Non-exclusive
You can license the same copy to other buyers too, and earn again.
Buyer
An AI company or data partner that pays to train or test its models.

Next: how the process works, step by step ↓

01 / The process

How licensing your data works

What most buyers look for. Each company is still assessed on its own merits.

25+staff at peak
2+ yearsof records, ideally 5+
2+ systemsemail, chat, tickets, CRM, docs
Any statustrading, restructuring or in insolvency
01

Tell us what you hold

How many people, how many years and which systems. No files, no logins.

02

We prepare it

A readiness review: what you own, what stays out, and de-identification inside your own environment.

03

Buyers compete

We aim to show an anonymised profile to 2–4 buyers together, then set their offers side by side.

04

You approve, you get paid

You sign only the terms you're happy with. Privacy checks and delivery are agreed before anything moves.

02 / The market

AI companies are spending billions. Data is the bottleneck.

Models learn from real examples of work, and good public data is running short. That's why the biggest AI developers now pay to license data, and why company records are a growing part of it.

$33.9bnprivate investment in generative AI in 2024Stanford AI Index 2025
$109.1bnUS private AI investment in 2024. The UK: $4.5bnStanford AI Index 2025
~$3.2bnspent on AI training data in 2025, growing about 20–25% a yearAnalyst estimates, 2025
$1m–$5ma year: OpenAI's reported offers to license one publisher's contentThe Information, 2024

Who's buying

Companies publicly reported as paying to license data. Shown to illustrate the market; they are not our partners.

Training-data partnersModel evaluation firmsVertical AI start-upsData marketplaces
03 / What you can sell

One system shows a slice of the work. Several show all of it.

The more systems, people and years of history behind your records, the more they're worth. Buyers pay most for work they can follow from the conversation to the ticket, the document and the code.

TETeamsSLSlackGCGoogle ChatZOZoom

Chat and messaging

Whole conversation threads, with who took part and every reply kept in order.

OUOutlookGMGmailEXExchangeCACalendar

Email and calendar

Who wrote to whom and how they replied, kept as whole threads, with the meetings that went with them.

SHSharePointONOneDriveGDGoogle DriveDRDropbox

Documents and files

The content and its metadata: who wrote it, every version and when it changed.

JIJiraCOConfluenceASAsanaMCMonday.comNONotionTRTrello

Work and knowledge

Tickets, sprints, wikis and the full history of each project.

SASalesforceHUHubSpotZEZendeskFRFreshdeskINIntercom

CRM, sales and support

Deals from first contact to close, account histories and the judgement calls along the way.

SEServiceNowWOWorkdaySASAPD3Dynamics 365

Business platforms

Requests, approvals and workflows from IT, HR and operations.

XEXeroSASageQUQuickBooksNENetSuiteSHShopify

Finance and operations

Invoices, ledgers and contracts that show where the money goes.

GIGitHubGIGitLabBIBitbucketSNSnowflake

Data and code

Repositories, reviews, schemas and warehouses: the deepest layer of how work gets done.

In-house and custom systems

Not listed here? If your team does its work in it, it is probably worth something.

Ask about your systems →
04 / Your data

What buyers look for

What every buyer checks, whatever the data
Clear rightsProof you can license it: your own records, and customer or vendor contracts that don't forbid reuse. We check this in the readiness review.
Real outcomesThreads that end in a resolution, a decision or a sale. A ticket marked solved, or a deal marked won or lost, tells a model what good looks like.
Depth over timeSeveral unbroken years from the same systems, so buyers can see how language, products and processes changed.
Clean, consistent formatExports that keep threads, timestamps and metadata together, with stand-ins applied the same way everywhere.
05 / What affects price

What makes records worth more

Buyers pay for context and outcomes, not raw volume. Examples are illustrative. See what buyers pay →

Communications

Threads that end in a decision.

The question, the discussion and what was decided, all in one place.

#operations08:17
"Supplier's late again. Do we switch to the backup for Friday?"
✓ Switched · procedure updated
Customer support

Tickets with the whole fix.

The customer's problem, the agent's reasoning and the fix that closed it.

Ticket 20417Resolved
Invoice export failing
RaisedDiagnosedFixed
Documents

Procedures with their history.

Earlier versions show how a process improved. Buyers pay more for that.

Returns procedurev12
Step 4 added: check stock before refunding.
v1 → v12 kept+ history
Sales and CRM

Deals with the outcome recorded.

Won or lost, and why. Outcomes turn a log into training signal.

Engineering

Code with its reviews.

Commits, review comments and the bug each change fixed.

Time span

Years, not months.

Five or more years across several systems is worth far more than one clean export.

06 / Data safety

Personal data out. Value kept in.

This is step 2 of the process in detail. Data deals can fail on privacy, or lose their value in a clumsy clean-up. Ours is built for both: people stay protected under UK GDPR, and the work in the records stays intact for buyers.

Blanked out (typical)

"[PERSON] asked [PERSON] to call [PERSON] at [ORG] about the [PERSON] refund. [PERSON] agreed."

Who said what is lost. The thread is worth less.
Consistent replacement (ours)

Same person, same stand-in, every time. The conversation still makes sense.
01

Map and exclude

We list every system and agree what never leaves: private messages, HR, health, children's and legally privileged records, and anything under client confidentiality.

02

Assess first

A data protection impact assessment and legitimate interests assessment, done with your lawyer before any processing starts. De-identifying data is itself processing, so it needs a lawful basis too.

03

Detect and replace

Automated detection plus human review finds names, contact details, account numbers and rarer identifiers. Each is swapped for a consistent stand-in of the same type, inside your own environment.

04

Test like an intruder

We sample the output and try to re-identify people, following the ICO's "motivated intruder" test. Anything that fails goes back a step.

05

Lock it in contract

Buyers are named, barred from re-identification and resale, limited to agreed uses, and must delete on request with a signed certificate.

The key never leaves you

The table linking stand-ins to real people stays with you, or is destroyed. Pseudonymised data is still personal data under UK GDPR, so we treat it that way until the tests pass.

Re-identification is a crime

Under section 171 of the Data Protection Act 2018, knowingly or recklessly re-identifying de-identified data without consent is an offence. Our contracts say so too.

No live access, ever

Buyers get a tested, de-identified export. Never logins to your systems, and we don't hold your raw data ourselves.

Based on the ICO's guidance on anonymisation and pseudonymisation and the Data Protection Act 2018. This isn't legal advice: every deal is reviewed with your own lawyer. More on how we handle data →

07 / Competition

More buyers, better terms.

This is step 3 of the process. Going direct usually means one buyer and one number. We aim to put your anonymised profile in front of several buyers together, so they compete on price and terms.

  • We aim for 2–4 buyers on each profile
  • Weigh the whole offer: how long, how exclusive and when you're paid, as well as the price
  • Nothing is shared and nothing is signed without your approval
Example · made-up numbers

Four offers for the same data

Offer 1 Buyer A opens
£60k
Offer 2 Buyer B bids higher
£85k
Offer 3 Buyer A raises
£110k
Offer 4 Buyer C, non-exclusive
£130k

When buyers know others are bidding, offers rise. The best one here is also non-exclusive, which means you can still license the same data to other buyers later.

What we handle

Data deals take real work. Most of it lands on us.

Legal, security, technical and commercial work, run end to end, so your team spends hours, not months. You stay responsible for what's shared, and you approve every buyer and term.

Finding buyers

We keep a live list of who buys what, and approach them for you.

Readiness review

Inventory, rights check and a de-identification plan, agreed with your lawyer.

NDAs and data room

Buyers see only the samples and summaries you approve.

Pricing and terms

Every offer set out side by side, so price is never the only thing you see.

Contracts

Licence terms checked clause by clause: use limits, deletion and audit rights.

Delivery

Secure transfer after de-identification is tested. Buyers never get live logins.

08 / Why us

Five reasons to choose us.

Choosing an adviser is about trust as much as price. You've seen how we protect data and how deals run. So here's exactly how we're different, and every point goes into our engagement letter.

01

We only ever act for sellers

We never charge buyers. Anything a buyer pays on your deal, such as a referral reward, is disclosed to you and paid to you in full, so our only incentive is the best deal for you.

02

Every fee in writing, and no lock-in

12% of proceeds, only if a deal closes, with the readiness review included and nothing to pay up front. No 12-month exclusivity and no long tail clauses: walk away before a licence is signed and you owe nothing more.

03

Your raw data never touches our servers

De-identification runs inside your own environment. We coordinate the work; we never hold copies of your records.

04

Clean-up that keeps the value

We replace personal details with consistent stand-ins instead of blanking them out, so conversations still make sense to a model. That keeps the text far more useful than blanking it out.

05

UK compliance and buyer vetting built in

ICO-registered and built around UK GDPR, PECR and the Data Protection Act. Every buyer is checked for ownership, sanctions and links to hostile states before a data room opens.

→

See what yours is worth

A free, nine-question estimate. No files, no logins.

What it could look like for you

Example scenarios, not clients. The estimates are worked out live by the same model as our estimator.

Refer and earn

Advise a company sitting on years of records?

Introduce the owner. We agree your share in writing before anything moves, and it's disclosed to your client.

Make an introduction → How referrals work

We never pay or receive fees per referral or appointment from insolvency practitioners.

If their deal closes at£500,000
£50k£2m

You'd earn

£15,000

25% of our fee, and our fee is 12% of the deal (8% above £1m). Set in your referral agreement.

Questions

Straight answers

Anything else? Email us. A person reads every message.

Get a free valuation →
Quick answersTwo questions sellers ask first

Do you see our raw data?

No. De-identification runs inside your own systems. We coordinate the work and never hold copies.

Are staff messages included?

Never. Private messages, HR and health records are excluded by default.

Who buys this data?

AI labs and the specialist firms that supply them. Most of it trains AI agents to do real professional work: answering customers, following procedures, writing and reviewing code.

Will selling our records help our competitors?

No. Buyers are AI developers training general models, not companies in your market, and each licence names the buyer and bans resale or sublicensing. Pricing, strategy, client-confidential material and anything you consider sensitive stay out, and you approve every buyer before they see a sample.

Are you a data broker?

Not in the usual sense. People-search and marketing data brokers build and sell profiles of individuals. We never do that. We help companies license de-identified records of how work gets done, with names and contact details removed inside their own systems first.

Does my company qualify?

Usually if you have 25+ staff, 2+ years of records and at least two systems. The estimator says plainly if you don't, and what would change that.

How long does it take, from first call to payment?

We expect two to four months, though every deal differs. The estimate takes five minutes and the readiness review two to three weeks. Taking a profile to buyers and agreeing a licence usually takes another four to eight weeks, and payment follows delivery, sometimes in stages.

Do we need to tell our staff or customers?

Usually, yes. UK GDPR expects people to be told how their data is used, so most companies update their privacy notice and let staff know before de-identification starts. We help with the wording, your lawyer confirms it, and anyone can object through our objection page.

How much work is it for our team?

Usually a few hours a week from one person for two to three weeks: a call to list your systems, IT help with exports or vendor access, and sign-offs at each stage. We run the buyers, the data room, the negotiation and the paperwork.

Is this legal under UK GDPR?

It can be, done properly. De-identifying personal data is itself processing, so it needs a lawful basis and an assessment first. Customer personal data usually stays out. We run that assessment with your lawyer before anything moves.

Do you see our raw data?

No. Exports and de-identification happen in your environment, run by your IT team or a specialist you appoint. Buyers never get live logins. We coordinate; we don't hold it.

Will staff's private messages be included?

No. Direct messages, HR, grievance and health threads are excluded by default, and anyone can object through our objection page.

Can a closed or insolvent company sell its data?

Yes, through its liquidator or administrator, who controls the records. See insolvency practitioners.

How is payment worked out?

Buyers usually pay a licence fee per dataset, sometimes in stages on delivery. Non-exclusive licences let you license the same de-identified data again.

What does it cost, and what if no buyer wants it?

The estimate is free. We take 12% of licence proceeds (minimum £2,500, 8% above £1m), only from money received. The readiness review is included, so there's nothing to pay up front, and if no deal closes you owe nothing. We tell you early if we don't think your data will sell. See pricing.

Find out what your records could earn.

Free, about five minutes, and nothing to upload.

Get a free valuation →