What is my company's data worth?
There is no fixed price list. What your internal records are worth depends mostly on how many people used them, for how long, across how many systems, and how complete and varied the history is. Buyers publish their own ranges, and this site shows those ranges on each buyer's page.
Why do AI companies pay for ordinary business records?
AI developers are building software agents that are meant to do office work: answer tickets, chase invoices, coordinate shipments, review code. To teach that, they need real examples of how work actually gets done inside a company. Public internet text does not show that.
As of October 2026, press reports describe AI labs, and vendors that build simulated workplaces for them, buying email, chat, documents, tickets and code from companies. Reporting from MIT Technology Review links the demand to labs reaching the limits of public and synthetic data. The same reporting notes that nobody has published tests proving real archives train better models than the alternatives, so buyers are still working out what they value most.
- Mode$100K–$5M
- Handshake AI$100K–$4M
- Telegraph Lab$100K–$4M
- Polyshares$100K–$2M
- micro1$100K–$2M
- Mercor$20K–$1.5M
- Ooak Data$20K–$1M
Show as a table
| Buyer | Published range | Buyer's label |
|---|---|---|
| Mode | $100K–$5M | indicative |
| Handshake AI | $100K–$4M | headline (buyer's own figure, not an offer) |
| Telegraph Lab | $100K–$4M | target |
| Polyshares | $100K–$2M | typical |
| micro1 | $100K–$2M | tiered (open-ended '+') |
| Mercor | $20K–$1.5M | by industry |
| Ooak Data | $20K–$1M | ceiling |
What drives the value up or down?
Buyers tend to look at the same handful of factors. None of them is a formula, but together they explain most of the gap between a small offer and a large one.
- Headcount. More people means more conversations, more hand-offs and more decisions on record.
- Years of history. A long, continuous record shows how work changed over time. One wind-down service quoted by MIT Technology Review said it values assets partly on head count and years in operation.
- Number of connected systems. Email plus chat plus tickets plus documents lets a buyer follow one piece of work across tools, which is closer to how a real job looks.
- Depth and completeness. Gaps, deleted channels and missing years lower value. A complete, well-kept archive is easier to use.
- Variety and uniqueness. Records that show disagreements, judgment calls and unusual situations are often more useful than routine traffic. A niche industry workflow few buyers have seen can matter more than sheer volume.
- How the team used its tools. A company that ran its work in its ticketing system and chat, rather than in hallway conversations, leaves a richer trail.
Does size beat quality?
Not always. MIT Technology Review quoted one buyer saying a single company's chat history can be worth more than another company's chat, email and calendars combined, if it has enough variety and depth. Volume helps, but a buyer is really estimating how many useful training tasks it can build from your records.
That estimate is why offers for similar-sized companies can differ widely.
Does it matter if the company is still operating or closing?
Yes, in a few ways. A closing or closed company usually cannot add new data, so the archive is fixed and the deal is often a one-time payment. Much of the reported activity in 2026 has involved shut-down startups.
An operating company may be able to offer ongoing or refreshed data, which some buyers value. It also has more to protect: live customer contracts, current employees and trade secrets. That often means more exclusions and more review, which can shape what is on offer. See the guides on what can be excluded and on employee and customer privacy.
What lowers value or blocks a deal?
Some things reduce what a buyer will pay, or stop a deal entirely.
- You do not clearly hold the rights to license the records (for example, customer contracts forbid it).
- The archive is heavy with regulated data, such as health or consumer financial records, that cannot be cleanly removed.
- Large parts of the history are missing or cannot be exported.
- The records are mostly personal or HR material that would have to be stripped out anyway.
Where can I see real numbers?
Several buyer programs publish their own payout ranges or pricing approach. This site shows those figures on each buyer's page, labeled as the buyer's own figure, with the date we checked it. They are not offers. The actual amount depends on the buyer's review of your records.
Because the market is new and moving, published ranges change. Check the date shown next to any figure, and compare more than one buyer before you decide. Talk to your accountant about how any payment would be taxed.
Common questions
Can you give me a price for my company's data?
No. Only a buyer can make an offer, after looking at what you have. This site shows each buyer's published range so you can see where you might land.
Is more data always worth more?
Not necessarily. Buyers care about variety, completeness and how useful the records are for training, not just size.
Does my industry matter?
It can. Workflows that buyers rarely see, such as niche operations or specialized back-office work, may be more interesting to some buyers than common ones.
Will removing sensitive material lower the price?
It may reduce volume, but buyers generally expect personal and regulated data to be removed. Clean, low-risk records are easier for a buyer to accept.
Are the ranges on buyer pages guaranteed?
No. They are the buyers' own published figures as of the date shown, not offers.
Sources
- AI labs are buying data from dead startups to train their models (MIT Technology Review, Oct 13, 2026), accessed 2026-10-15
- Failed Companies Are Selling Old Slack Chats and Email Archives to Train AI (Gizmodo, Apr 17, 2026), accessed 2026-10-15
- Shuttered startups are selling old Slack chats and emails to AI companies (Fast Company, Apr 17, 2026), accessed 2026-10-15
General information, not legal, tax or financial advice.