AI labs pay for real records of how work gets done. The emails, quotes, orders, drawings and customer chats on your old server are exactly that, whatever your trade. We buy them, take out the names, and sell them straight to the labs.
reported paid to a 70-person engineering firm for its records
reported paid to a logistics and distribution business
Figures reported by US data firms Telegraph Lab and Red Deer Data, converted from dollars at the 20 October 2026 rate. Not McKays deals, and not a valuation of yours.
Buyers are asking for these today. If this sounds like your business, or a business you look after, we’d like to talk.
WANTED
About 50 to 500 staff
Complete engagement histories, ideally with:
WANTED
About 50 to 1,000 staff
What the data team built and ran, ideally with:
Any size, any sector
Engineering and machine files, lab and research data, field and delivery records, customer service histories, and plenty more.
Tell us what you’ve got and we’ll tell you straight if there’s a buyer.
Tell us what you’ve gotAny UK business with digital records. Emails, orders, quotes, drawings, chats, machine files: if it shows how work gets done, labs want it.
Every business has one. Old emails, orders, quotes, drawings, chats and machine files nobody’s opened in years. That’s exactly what AI labs are short of.
The real value is in how they link up: one customer’s order, the emails about it, and how it got sorted. Click the other folders.
Every business leaves a trail of how work gets done, whatever it does. Pick a record to see what it looks like once the names are out.
Talking
Customers and money
Making things
Science and labs
Out in the field
Consulting and advisory
Software and data teams
Supplier says the 316 bar is six weeks out. Can we use 304 for this batch?
Not for the marine brackets. 304 pits in salt spray inside a year. Air-freight 40 bars of 316 instead.
Agreed. Tell the customer it’s +£1,180 or nine days late, their choice.
Real decisions with the reasons attached: who knew what, what it cost, and what they chose.
Where it lives: Gmail, Outlook, Google Workspace, Microsoft 365
the reasoning, not just the answer
Line 2 tripped again. Same alarm as Tuesday, E-47.
Check the coolant level first. Last three times it was low, not the sensor.
Coolant was low. Topped up, back running. 17 minutes down.
Nice one. Adding it to the checklist.
Problems solved in real time. The quick fixes that never make it into a manual.
Where it lives: Slack, Microsoft Teams
fault, fix, and how long it took
Customer: Third time it’s cut out this week.
Engineer: What does the pressure gauge say? Is it below one?
Customer: It says 0.6.
Engineer: That’s it. Top it up to 1.5, I’ll talk you through it.
Outcome: fixed on the phone, no callout.
Real conversations where problems get solved, word for word.
Where it lives: Phone systems, call recording software, CRM
fixed without a visit
Main contact: Hannah Doyle
Lost them to a cheaper quote in 2021, and again in 2022.
Won them in 2023 with 60-day payment terms, not a lower price.
Their buyer gets judged on cash flow, not unit cost.
Why customers buy, why they leave, and what won them back.
Where it lives: Salesforce, HubSpot, spreadsheets
why they really said yes
Same customer late every Q4. Chase on day 20, not day 30.
Real cash flow: who pays late, what gets chased, and what actually works.
Where it lives: QuickBooks, Xero, Stripe, Ramp, Shopify
patterns you only see over years
Supplier’s total liability is capped at the contract value twice the annual fees.
Comment from Alison Grant:
We accepted twice the fees after they dropped the indemnity in clause 11.4.
Comment from Tom Reekie:
Fine. 11.4 was the real risk.
How deals really get negotiated: what was given up, and what was won back.
Where it lives: Word files, document systems, email attachments
the trade-off is the gold
Every revision, with the reason it changed. That’s years of design judgement in one folder.
Where it lives: AutoCAD, SolidWorks, Revit, Fusion and shared drives
why it changed is the gold
Machine programs with the operator’s notes. Every feed rate someone slowed down, and why.
Where it lives: Machine controllers, CAM software and shop-floor folders
years of trial and error, in brackets
Found: porosity in 6 of 40 castings
Cause: ladle stood 9 minutes before pouring, metal too cool
Fix: pour within 4 minutes, log the temperature every time
Signed off: Michael Okafor
What went wrong on the shop floor, why, and the fix that stuck.
Where it lives: Quality systems, ERP, shared drives, scanned paper
cause and fix, written down
T-03 was cut from the weld zone. Too much heat, procedure changed to rev B.
Results tied to what was tested, and why things failed.
Where it lives: Test machine software, spreadsheets, PDF reports
the fail is the useful bit
14 March. CHO cells, passage 23.
Viability down to 71%. Normally 95% plus.
Swapped to a new serum batch.
Old serum batch was the problem. Back to 96% by day 3.
Initials: JLR
Experiments as they really happened, including the ones that went wrong.
Where it lives: Paper notebooks, electronic lab notebooks, shared drives
the failed runs teach the most
Edge wells read high on this reader. Use the inner 60 only.
Checked by Dr Priya Nair
Raw results with the judgement calls: what got retested, what got ignored, and why.
Where it lives: Plate readers, lab systems (LIMS), Excel
the judgement calls
FLOWCELL A3K29 SAMPLES 48
ORGANISM E. COLI STRAINS
Q30 87.2% PASS FILTER 79%
LANES 3-4 LOW YIELD: LIBRARY OVERLOADED
RERUN AT 1.4 NM: Q30 93.1%
OPERATOR D. KOWALSKI
Instrument runs with the fixes that made them work.
Where it lives: Sequencer exports, lab systems, lab servers
what they changed to fix it
Spikes follow rain within 48 hours. Sample after storms, not just on schedule.
Years of readings, with the local knowledge that explains them.
Where it lives: Data loggers, monitoring software, spreadsheets
readings plus the why
Weather: heavy rain from 10:00
On site: 6 crew, 1 excavator
Stopped groundworks at 11:30, trench flooding.
Moved the crew inside to first-fix. Lost half a day, not a full one.
Signed: Craig Muir
What really happened on site each day, and the calls made on the spot.
Where it lives: Site apps, project software, paper diaries
on-the-spot decisions
Town-centre drops before 9 always run late. School traffic. Do them after 9:30.
Driver: Stevie Lang
Real routes against real traffic, and the driver know-how that beats the plan.
Where it lives: Vehicle trackers, route planners, ShipHero, spreadsheets
the driver knows best
River 4 went in late after flooding. Drill before 5 Oct or change variety.
Farm: Mains of Auchterless
Season after season of what was planted, what happened, and what it yielded.
Where it lives: Farm software, yield maps, spreadsheets
a season’s lesson in one line
Scope: freight and logistics spend across 4 sites
Interview with Fiona Baird, ops director:
“Half our freight goes same-day because nobody plans past Thursday.”
Finding: £1.05m a year on avoidable express freight
Six months on: weekly freight plan in place, express spend down 38%
The whole job, start to finish: what was asked, what people said, what was found, and whether it worked.
Where it lives: SharePoint, Google Drive, email, project folders
brief to result, all connected
Review comment from Ravi Sethi:
57% is too high. Take out the two hospital clients, they really need same-day.
Live formulas, every version, and the reviewer’s challenge that changed the answer.
Where it lives: Excel, Google Sheets, shared drives
v1 to v8, and why it changed
Saves £1.05m a year. Pays back in 4 months.
Comment from Helen Ross, partner:
Lead with late deliveries, not the saving. That’s what their board worries about.
Reply from Sam Kerr:
Moved to slide 3. The saving is now the second point.
Drafts and finals with the review comments: how advice gets sharpened before the client sees it.
Where it lives: PowerPoint, Google Slides, SharePoint
how the advice got sharper
models/finance/fct_revenue.sql
- where status = 'paid'
+ where status in ('paid', 'part_refunded')
Review from @jmcfarlane
Part refunds were dropping out. Revenue 2% low since March.
Added a test so it can’t happen again.
Every change to the data code, with the review and the reason. How real data problems get found and fixed.
Where it lives: GitHub, GitLab, dbt
the bug, the fix, and the test
Incident note: HubSpot pull moved to 01:00, with retries. No failures since.
Every job run, every failure, and the incident notes that explain the fix.
Where it lives: Airflow, Dagster, dbt Cloud, loading logs
what broke at 2am, and why
Raised by: Lewis Park, sales
Data team chat: the source system changed its date format, so the churn table stopped updating.
Fix: read both date formats, alert if a table goes 24 hours without updating
Closed: 2 days later
Questions, bugs and fixes between the data team and the rest of the business.
Where it lives: Jira, Linear, Slack, Teams, BI tools
the question behind the dashboard
If it’s digital and it shows how work got done, it counts. We agree exactly what gets exported, so there’s never a blanket handover of your inbox or drive.
tick the ones you use
Nothing ticked yet.
Tell us about themTwo minutes on the form below. Rough answers are fine. No files, no logins, no passwords.
Within a week we’ll tell you if it fits, and what we’d pay or what a licence could earn.
Once you’ve signed, the names come out, the cleaned files go to the lab, and you get paid.
Not every archive fits. If yours doesn’t, we’ll say so, and it won’t have cost you a penny.
Each box is a year of records.
1,440 jobs and orders
Each with its own emails, paperwork and back-and-forth. That’s a story labs can’t get anywhere else.
A sense of scale, not a valuation.
Sell or license a cleaned copy of your old job files and keep the originals. Nothing changes in your day-to-day.
Your records could be one of the last things the business has worth selling. Talk to us before anything is deleted or switched off.
If you kept the records when you sold up or retired, tell us what’s there. We’ll check what you’re free to sell.
Digital records can be a real asset in a liquidation. We’ll look quickly, tell you if there’s value, and work to your timetable.
Richard McKay has spent 30 years in business, first in construction and then in e-commerce, where he built Sprung Gym Flooring into the UK’s largest online rubber gym flooring shop. He was also an early adopter of AI in his own businesses.
Every business he’s run lived on its records: quotes, orders, and the emails where a price changed or a problem got fixed. Those records are what AI labs now pay for.
Nothing. The first look is free. If we buy your records, we pay you. If we set up a licence instead, any fee is agreed before you sign.
A short description of the business and the records you’ve got. No files, no logins, no passwords.
Not with a licence. The lab gets a cleaned copy and you keep the originals. If we buy outright, we’ll set out exactly what changes hands before you agree.
They come out before anything leaves. You can also leave out particular clients, jobs or folders.
AI labs building models that understand real work. We sell to them directly, and the buyer is named in your agreement.
You’ll hear back within a week. Got a deadline, like a server switch-off or a closing date? Tell us and we’ll work to it.
We’ll tell you straight. No charge, no obligation.
Two minutes. Rough answers are fine. We’ll reply within a week with a straight answer.
UK businesses and digital records only.
describe it, don’t send it
Prefer email? [email protected]
Not ready yet? Read our free guide, Before you license (PDF).