AI labs are paying for the way your company works.$100K to $3M+.

Not your product and not your customer list. The written record of how the work actually happens: email, chat, tickets, job notes. That is the written-record test, and it is what this assessment asks.

Eligibility assessment Question 1 of 6
The business
Is the business still active? Records from a company that has closed are a category here, not a problem.
6 questions · about a minute6 to go

Free assessment · 60 seconds · No obligation

Trillions will be spent building the next generation of AI … it lives inside real companies like yours.
micro1.ai
25+ employees required 2+ years of records Based in US, CA, UK, IE, AU, NZ Records in English
What it pays

What they pay, and what moves it.

These are the tiers micro1 publishes. What you land on depends on how much you have, how good it is, and how rare it is.

$100K–500K
Qualified enterprise data partnerships
The usual fit. A company that runs properly and writes things down.
$500K–2M
Large-scale operational datasets
Records from several teams, or taking part across more than one of them over time.
Top tier
$2M+
Highly unique, proprietary operational data
Something unusual that an AI lab can't get anywhere else.

The program reaches $3M+. These three bands are the tier structure micro1 publishes, and its own payout estimator runs well past the top one: $4.2M for a company with two hundred people and fifty years of records, which micro1 labels an indicative figure rather than an offer. Every deal is judged on its own, and we will never put a number on yours.

Don't believe us. Believe them.

We are a referral partner, which means we have an interest in you saying yes. So don't believe us. Every figure below was published by somebody else, and we've linked it.

micro1 is worth $4B
Following a raise of over $100M, not announced by micro1 and not by us. Forbes, 22 Sep 2026
It has committed $1B to buying data
Over twelve months, financed with capital from Citi and Hercules Capital. Forbes, 4 Dec 2025
It tried to outbid Google for one company's records
Spirit Airlines' internal data went to auction in its bankruptcy. micro1 moved to top Google's winning bid. Business Insider, Aug 2026
Microsoft and Amazon are named customers
On micro1's own materials. What it does not publish is which labs receive which datasets. micro1's own published pages

Also covered by Fast Company, The Information, NBC Bay Area and Inc. None of these publications is affiliated with Atrayu and none endorses it.

They're not buying your documents.They're buying the decision.

These models have read Wikipedia, Reddit and every book anyone could get hold of. What they have never seen is how a real business actually makes a call. A customer complains by email. Your team argues about it in chat. Someone decides, and writes it in a ticket. The books show the refund.

That's one whole decision with the reasoning still attached, and there is nothing like it anywhere on the public internet.

Which is why two companies the same size can be worth completely different amounts. If your team sorts things out over the phone, nothing is left behind and there is nothing to license. If they write it down, there's a full trail, and the trail is what's bought.

We've now exhausted basically the cumulative sum of human knowledge … in AI training.
Elon Musk Reported by TechCrunch, 8 January 2025
2024-03-11 EMAIL 2024-03-11 CHAT 2024-03-12 CALL 2024-03-12 DOCS 2024-03-14 CRM 2024-03-15 BOARD Email customer complains Chat team argues the call Ticket resolution recorded Ledger refund posted One complete decision with the reasoning attached SCATTERED CONNECTED
  1. EmailA customer complains
  2. ChatThe team argues the call
  3. TicketSomeone decides, and records it
  4. LedgerThe refund gets posted
  5. One complete decisionwith the reasoning attached

How the process works, in six steps.

Step01
Check your eligibility
Six questions about size, age, systems, location and who owns the records. A minute, and no email.
Step02
Apply
Clear the bar and we hand you to micro1's own form. About five minutes. They come back to you, usually within a week.
Step03
Scope and price
micro1 works out which of your records it can license and puts an offer against them. You decide what goes in, what gets left out and what never leaves.
Step04
Sign and hand over
The agreement sets the terms, then you export from the systems you already use. micro1 publishes a commitment not to resell or distribute your data until identifying information has been removed.
Step05
Your data is cleaned
Names, emails and card numbers are stripped before any AI lab trains on a thing. micro1 says it does this itself, and that you do not have to clean anything first. Who pays for that work is still a term worth agreeing in writing.
Step06
Get paid
micro1 reviews the data for completeness, then publishes payment “in full within 30 days of acceptance”. That is 30 days from them accepting the data, not from the day you sign.

Only the first step is yours to take today. Six questions, about a minute, and you'll know whether the other five are worth your time. Nothing binds you until step four. You get there by choosing to, twice.

Check my eligibility →

What you need to have, in systems you already run.

Buyers don't ask what's inside your files. They ask which systems you run and how far back they go. micro1 says companies with records across several business systems are the stronger fit, so the number of these you recognize matters more than any one of them.

Gmail
Outlook
Slack
Zendesk
HubSpot
Salesforce
Jira
Asana
Notion
Google Drive
QuickBooks
Xero

Don’t see your tools? That’s fine. These are common ones, not a list of what counts. micro1 works out what you actually run when it reviews your application.

Customer support
Ticket histories, escalation paths, QA notes and knowledge base articles.
Sales and CRM
CRM records, pipeline notes, playbooks and the reasoning behind won and lost deals.
Engineering and product
Technical docs, code review, bug tracking, internal wikis and planning threads.
Finance and accounting
Reconciliations, approval workflows, reporting processes and documented SOPs.
Operations and logistics
Dispatch, fulfilment, exception handling, inventory and vendor management.
Email and chat
The thread where the decision actually got argued out, which is the part they want most.

What an AI lab actually receives.

Names, emails and card numbers are taken out before any AI lab sees a thing. Here's what that looks like on one record.


          
Personal details highlightedIllustrative record
Where it goes, in micro1's own words
  • “Licensed exclusively to labs and enterprises based in the U.S. and allied countries.”
  • “Your data is kept separate from other companies’ data, with access restricted to authorized personnel.”
  • “Raw data is deleted according to an agreed schedule.”
  • What it does not publish: which labs receive which datasets. Microsoft and Amazon are named as customers on micro1’s own materials, and the rest of the list is not public. Ask for the named licensee in writing before you sign.

Quoted from micro1’s published pages. Checked 24 Oct 2026.

If the company is closing, closed or sold

Check the records before you cancel the accounts.

"I don't know how we are going to pay our bills" became "we can tie this up neatly with a bow and be able to walk away."
Shanna Johnson, CEO of cielo24 A transcription company that wound down and sold thirteen years of Slack, Jira tickets and email as AI training data. Forbes reports it received "hundreds of thousands of dollars." Forbes, 16 April 2026

A closed company is a category in this market, not a problem. micro1 accepts them from more countries than it accepts operating ones. The catch is practical rather than legal.

Keep these alive until you know what they're worth
  • Email and shared drives
  • Slack or Teams
  • CRM and helpdesk
  • Jira, Asana or whatever tracks the work
  • Code repositories
Two things that catch people out

Hosted systems get cancelled on a schedule. Exports expire. On Slack's free plan, older content is deleted. Once it's gone there is nothing to license, and that happens quietly while everything else is being wound up.

Whoever signs has to have the authority. That may be the former directors, or it may be a liquidator, administrator or trustee. If one has been appointed, they usually decide, not the founders. Settle that before you apply, not after.

Check the records →

Questions owners actually ask.

Where we don't know, we say so.

Honestly, is this a scam?

Good question, and the right instinct. Here's how to check it without taking our word for anything.

micro1 is a real company with a real address: 655 Montgomery St, San Francisco. It raised over $100M at a $4B valuation, and that was reported by Forbes, not by us and not by them. Look them up yourself.

Be careful about the middle layer, which is where we sit. At least twelve outfits are running ads into this program right now and most of them don't tell you they get paid. We do: micro1 pays us if you license, and it's at the bottom of every page here. A site that won't tell you how it makes money is the one to worry about.

Who actually ends up with the data?

The buyer is not the end customer. It licenses on to AI labs, and names Microsoft and Amazon among its customers. It does not publish which labs receive which datasets.

That gap matters, and it is one of the five things we'd put to them in writing before signing. Name the legal entity on the agreement, ask who they can pass the records to, and ask whether that list is fixed or open-ended. Get the answer in the agreement, not in the email thread.

Does any data leave my systems when I run the assessment?

No. The eligibility assessment runs in your browser, nothing is uploaded, and nothing you tap on this page is recorded or sent anywhere. There is no email field until after you have your result, and there is none at all if the answer is no. Run it →

We don’t qualify. Is there anything else?

Yes, and it is worth knowing about even if your own company never qualifies. micro1 pays a referral fee for every company that licenses its records, and you do not have to own one to collect it. Its published rates are $50,000 on a trading company and $75,000 on one that has wound down, both up to a deal size, then a percentage above it.

We have put the rates, the bar, the part about when you actually get paid, and what we take out of it on one page. The short version of the last bit: if you register through our link, micro1 pays us 20% of what it pays you, separately, not out of your fee.

Who is Atrayu, and why go through you?

When we sold one of our brands, the buyer's diligence went straight at the operational layer. Every account, every CRM, the support tickets, the whole file structure. We got through it, but we had never once thought of that record as something worth having, and it showed.

Knowing what we know now, we would have kept that business. The record it was generating would be worth something today. We handed it over without ever asking what it was.

That is who we are. Two operators, selling online since 2010. Multiple businesses and multiple exits across the years, from ecommerce to agencies to commercial real estate. Contract manufacturers, fulfilment houses, support teams, design and ad agencies. We have been in the mix with all of them, and the thing running underneath every one of them was the record.

So we are not selling you anything. We found something worth knowing about, we read everything these buyers publish, and we published what we found, including the parts that should put you off.

One practical thing. If you apply, use the link on this page. micro1 credits an introduction only when the company arrives through a partner link, and it is never applied afterwards. It costs you nothing, it does not come out of what you are paid, and it is the only way this site earns anything. Start the eligibility assessment →

What counts as "our data"?

The written record of how the work got done. Emails, chat threads, support tickets, CRM notes, project boards, internal docs, SOPs, job sheets and dispatch records.

Not your product. Not your customer list. Not your code, unless you choose to include it. You set the boundaries before anything moves.

Do we lose ownership?

No. The buyer states that companies retain ownership of their underlying data. What you grant is a license to use a copy.

The thing to read carefully is not ownership, it is exclusivity. Perpetual use is normal and every buyer wants it. Perpetual exclusivity means you can never license those records to anyone else again, so there is no second deal in two years when more buyers are competing. Those two things are routinely written into the same sentence. Ask for the exclusivity to be termed, around two years, and for perpetual use to stand on its own.

What happens to customer and staff personal information?

It has to be stripped before anyone trains on it, usually by a specialist third party. The Before and After panel further up this page shows what that looks like on a single record.

Two things worth knowing. First, who pays for that work is a negotiated term, not a given, and it is real money. Second, the buyer's wording is not identical everywhere: one page says "all confidential information is removed" and another says personal information is removed "when relevant". Those are different promises. Ask which one is in your agreement.

How much would we actually get?

We won't give you a number, and nobody should before they've looked at your systems. Any site that spits out a dollar figure from six dropdowns is guessing and dressing it up as maths.

What we can tell you is what micro1 publishes: $100K to $500K for a qualified partnership, $500K to $2M for large-scale datasets across several teams, and $2M or more for highly unique proprietary data. That is a band across every deal they have ever done. It tells you the market is real. It does not tell you where you sit in it. See where you land →

When do we get paid, and what could delay it?

After signature and after the data has been extracted and accepted, not on the day you sign. Expect roughly sixty days from signing to the wire, and ask for the trigger in writing.

The clause to watch is when the price is fixed. If the number can still move after they have extracted and inspected everything, you have not agreed a price, you have agreed to find out the price later, after handing over the only thing you had to bargain with. Ask for a floor in writing before extraction begins, or a clean exit if the revised number lands below a line you set now.

Does this cost us anything?

Nothing to us, at any point. micro1 pays us a referral fee if you apply through this site and go on to license. It is a separate payment that does not come out of yours.

Costs you may actually face in the deal itself: de-identification, if the agreement puts that on you rather than the buyer, and your own lawyer. We'd argue with you about skipping the second one. It is a license of your records with terms that run indefinitely, and a few hours of review is a rounding error against a six-figure agreement.

The conflict, stated plainly: a commission on a completed deal gives any intermediary an interest in the deal closing and no interest in the terms being good. That is true of us too. It is why the parts of this that should put you off are on this page, in the same type as the rest.

Could a competitor end up seeing how we operate?

Not directly. The records go into training a model, not into a database anyone can read through. The buyer says its purpose isn't to copy businesses, products or whatever makes you competitive.

The honest version of the risk is quieter than that. Once something has trained a model, the model has learned it, and you can't take that back. If you have one method that is your whole edge, leave it out at the start rather than relying on a promise later.

Can we change our mind afterwards?

Not really, and anyone who tells you otherwise is wrong. Once records have been used to train a model, deleting the source files does not remove what the model learned. Retraining to strip it out is prohibitively expensive, so in practice these agreements are one way.

That is not a reason to avoid it. It is the reason to read the terms properly before signing rather than after.

What about NDAs and our client contracts?

This is the question that stops deals, and it is why the assessment asks it.

If your records are mostly about your clients' businesses, your patients, or your cases, you may own the file without holding the right to license what is in it. Those are two different tests. Agencies, consultancies, MSPs, law and accounting firms often hold the densest records in this market and the weakest right to license them.

Read your client contracts and your own terms of service before you read anything a buyer sends you. Your internal record, how your firm runs rather than what it did for a client, can be a different and cleaner question. The assessment asks this one →

How much work is this for my team?

More than a form, less than an IT project. Expect a call of under an hour to walk through what you have, then a stage where you decide what's in and what's out, then an export from systems you already use.

You're not rebuilding anything and nobody gets access to your live systems. The hardest part is usually deciding internally what to include, not the technical side.

Find out if you're one of the companies they'll pay for.

Six questions, about a minute. No email, nothing uploaded, and a straight answer either way.

Check my eligibility →
About a minute Nothing uploaded Free, always