Selling vs Licensing Your Office's Slack, Email and Ticket History to AI Companies
Published · Updated
Disclosure: I run howmuchismycompanydataworth.org, which introduces companies to AI data buyers and earns a referral fee from the buyer when a deal closes. Dollar figures below are the buyers' own published ranges, not quotes or results.
If you run the office, you probably hold the admin keys. Slack or Teams. Google Workspace or Microsoft 365. The help desk. The CRM.
So when an AI company emails the owner about “licensing your company data,” the question lands on your desk next. What do they want, what would we be giving up, and is this even a good idea?
Short answer: almost nobody wants to buy your Slack history outright. They want to license it. That one word decides who owns the records afterward, who else you can deal with, and how you get paid.
Why AI companies want office records
AI labs are building agents that do office work. Answer tickets. Route requests. Update the CRM. Chase invoices.
Public web text does not show how that work really happens. Your ticket queue does. Your Slack threads show who decided what, and why something got escalated. That is the part they can’t scrape.
So the buyers are spending. micro1 announced a $1B, 12-month commitment to buy company data on October 9, 2026. Mode announced a $5B enterprise data initiative four days later. Scale AI, Prism by Labelbox, Handshake AI, Replay, Endless and others run their own programs.
Selling vs licensing: the basic split
A sale transfers the data and the ownership. You get paid once. You may lose the right to use or sell that data again. It can fit an old archive from a closed line of business. It rarely fits the systems you run on every day.
A license keeps ownership with you. The buyer gets the right to use the data for a set purpose, for a set time, under written rules. Nearly every AI data deal on offer right now is a license.
There are more shapes than those two. Data feeds and pure revenue-share partnerships exist too. If you want the full menu, this is a plain side-by-side of the five common data deals and how each one is priced.
Inside “license,” two more choices matter.
Exclusive or non-exclusive
An exclusive license means only one buyer may use the data. It is usually limited to a field of use or a time window. A non-exclusive license lets you license the same records to others.
Buyers start in different places. Prism’s FAQ says its partnerships are exclusive by default. Endless says its deals are non-exclusive.
Neither is wrong. But exclusive should come with three things: a higher price, a narrow field of use and a fixed end date. If it has none of those, you are giving away your other options for free.
Upfront cash or revenue share
Some deals pay a fee at signing. Others add a share of what the buyer earns from your data later.
- Replay pays upfront cash plus an ongoing revenue share.
- Prism pays an upfront licensing fee at signing plus a revenue share on downstream use.
- Scale AI describes its program as an ongoing partnership, not a one-time file transfer, with room for recurring revenue.
Revenue share looks good on a slide. It also ties your income to the buyer’s success, and you can’t see their books. If a deal includes a revenue share, ask for audit rights. No audit rights, no real revenue share.
What buyers actually ask for
Scale AI publishes the clearest list. Its data partnership page names:
- Chat: Slack, Microsoft Teams, Google Chat
- Email and calendar: Gmail, Outlook, Google Calendar
- Files: Drive, SharePoint, Docs, Sheets, Excel
- Work tracking: Jira, Confluence, Asana, Notion, Linear
- Business systems: Salesforce, HubSpot, Zendesk, ServiceNow, Workday, SAP
Its target industries include field services, staffing and recruiting, marketing agencies, law firms, insurance brokerage and property management. That is a lot of ordinary offices.
Prism says the most valuable data shows operational workflows, human decisions and escalation paths. In other words, how work actually moves. Volume matters less than how real it is.
Size matters too. Scale prefers 40+ employees. micro1 and Endless look for about 30+. Handshake AI asks for 20+ employees and 3+ years in business. Replay’s published tiers start at 20 to 50 employees.
What buyers publish about pay
These are the buyers’ own published or indicative ranges. None of them is a quote.
| Buyer | What they publish | How they pay |
|---|---|---|
| Scale AI | $10K to $1M+ (illustrative) | Ongoing partnership |
| Replay | $10K to $100K (20 to 50 staff), $100K to $1M (50 to 250), $1M+ (250+) | Upfront cash plus revenue share |
| Prism by Labelbox | About $50K to $1M+ | Upfront fee plus revenue share, exclusive by default |
| Endless | Up to $4M | Non-exclusive |
| micro1 | $200K to $2M+ | Not covered here |
What any one office could get depends on headcount, years of records, which systems you use, your industry and what you are legally free to share.
What you keep control of
The better programs give you real levers. Know which ones you have before you start.
- Scope. Scale says partnerships can start with a very limited scope. Prism says sellers can exclude any category, system or record type at any time. Start with one system or one team.
- Identities. Scale says it strips personal details inside your environment, before anything is extracted. It says employee and customer identities never leave. Prism excludes HR and employee identifier data by default.
- Approval. Scale says nothing ships without your sign-off on each data package. Prism has you approve the data brief before any buyer sees it.
- The exit. Prism’s licenses include a kill clause. It forces deletion within 30 days for material breach, a regulatory change or a change-of-control trigger you define up front.
One limit deserves a plain sentence. Prism’s own FAQ notes that a model already trained on your data cannot be untrained. Deletion stops future use. It does not reach back. Treat every license as permanent in that one sense.
The part ops should raise first
Your records include other people. Customers wrote those tickets. Employees wrote those Slack messages. Some clients have contracts with confidentiality clauses.
Check what your privacy policy and customer contracts say before anything is shared. The FTC warned in February 2024 that quietly changing your terms to use data for AI training could be unfair or deceptive. Rewriting the policy the week before you sign is the exact move it described.
And an office manager should not be the one who signs. That call belongs to an owner or executive with authority over company data, with a lawyer reading the contract. Your job is to make sure they ask the right questions.
Questions to ask before signing
- Is this a sale or a license? Who owns the data afterward?
- Exclusive or non-exclusive? If exclusive, for which field, for how long, and for how much more?
- Upfront, revenue share or both? When is each payment due?
- If there is a revenue share, can we audit it?
- Which systems, date ranges and record types are in scope, exactly?
- Who removes personal data, where does that happen, and is it before the data leaves our systems?
- Do we approve each data package before it ships?
- What is the permitted use? Can the buyer resell or sublicense our data?
- What happens when the deal ends or we revoke it? What can’t be deleted?
- Who pays if a privacy complaint or a breach starts on the buyer’s side?
- Do our customer contracts and privacy policy allow this at all?
Where to start
Take inventory. One page. List each system, how many years of history it holds, and what personal or client-confidential data sits in it.
That page answers half of what any buyer will ask. It also tells the owner, fast, whether this is worth a conversation or a hard no.
Then hand the owner the eleven questions above. A buyer that answers them clearly is worth talking to. One that won’t is telling you something too.