If your company's history lives in Google Drive, SharePoint or Notion instead of a database, you can still license it to AI labs. Most companies with 50 or more people don't have a data warehouse or a codebase worth talking about. What they have is ten years of proposals, SOPs, project folders, specs, meeting notes and spreadsheets, and that's a record of how the company does its work. AI labs are paying for that record now, because the public internet explains what a business does and almost never shows how. Read on to find out which documents are worth licensing, which have to stay out, why your folder structure matters more than you'd think, and how the export works.
Why buyers want documents from companies like yours
AI labs are training models to do work, not just answer questions. A model that's going to draft a proposal, write a runbook or review a spec needs to have seen thousands of real ones, with the messy edits and the version that got approved. Your Drive holds that. The public web has templates and advice about proposals. It doesn't have your 400 real proposals, the ones that won and the ones that lost, with the comments in the margins.
The buyers Troveo works with (AI labs and AI startups, 40+ of them) are paying six figures for company data because it shows real work in a real business. Documents are one of the nine categories on our business data page, alongside chat, email, CRM, support, project tools, finance systems, code and internal databases. You don't need all nine. Plenty of companies only have the documents, and that's where this guide starts.
What in your Drive is worth licensing
Go folder by folder. The question for each one is whether it shows how your company does something, or whether it's just a file.
| Folder or file type | What it shows a model | Worth licensing? | What has to come out |
|---|---|---|---|
| SOPs, runbooks, playbooks, checklists | How your company does a task, step by step, in your industry's terms | Yes, among the most valuable | Names, credentials, anything specific to a named client |
| Proposals, quotes, statements of work | How you scope, price and pitch work, and what a finished pitch looks like | Yes | Client names and contacts, pricing tied to a named client |
| Specs, designs, project plans, postmortems | How work was planned, what went wrong, what changed | Yes | Named third parties, confidential partner material |
| Meeting notes and decision memos | Why decisions were made, who had to sign off, what the tradeoffs were | Yes, if they're real and dated | People's names, HR matters |
| Reports and deliverables you produced | Finished professional work product in a specific field | Depends on your client contracts (see below) | Client identity, anything the contract says is theirs |
| Templates and internal training material | The standard your company holds itself to | Yes | Usually nothing |
| Marketing PDFs, brochures, public web copy | Nothing a model can't already find | No | Skip it |
| Contracts and legal documents | Terms other parties agreed to with you | Rarely, and only your own templates | Every counterparty, every signature |
| Personal files, HR folders, payroll, medical | Nothing you're allowed to license | No | All of it, before anything else |
The first three rows are where the value sits for most companies. A construction firm's SOPs, an agency's briefs and postmortems, a clinic's operating procedures, a logistics company's dispatch playbooks. These are hard to find anywhere else and they're written by people who do the job. Our guide to enterprise operational data explains why buyers value that kind of record over generic text.
Why your folder structure and version history matter
A single finished document is a snapshot. A folder is a story. The draft, the three revisions, the comment thread, the version that got sent, and the follow-up six months later show how work moved from an idea to a result. Models learn from that sequence, and it's the part no template site can offer. So don't flatten your Drive into a pile of PDFs before you scope a license. Keep the folder tree, keep the file history where your tools support it, and keep the dates. Our glossary entry on workflow trajectories covers why the sequence, and not the file, is the unit buyers pay for.
The same logic applies across tools. If a proposal in Drive links to a deal in HubSpot and a thread in Slack, those three together are worth more than any one of them. You don't have to license all three on day one. It helps to know they connect, because the scoping call will ask.
What has to stay out
Start with your customer contracts. If you're a services firm, a lot of what's in your Drive was produced for a client, and many client agreements say who owns that work and what you can do with it. Those documents are usually out of scope, and saying so up front saves everyone time. Your own methods, templates, internal processes and the way you run the business are yours to license.
Then the people. Employee and customer names, contact details, credentials, and client identities are removed to a documented standard before a buyer sees anything. That's our work, not yours, and it happens before the data is put to work. Anything under a confidentiality agreement with a third party, anything from a partner, and anything in an HR, payroll or medical folder stays out entirely. If a document falls in a grey area, leave it out of the first scope and revisit later. Nothing about this process starts with "give us everything."
Buyers pay for licensed data because it comes with a clean record of where it came from and who's allowed to use it. Our guide to AI data provenance covers what that record looks like, and it's the reason the exclusions matter as much as the inclusions.
Documents alone, or documents plus the systems around them
Here's the honest part. A folder of documents on its own sits at the lower end of what company data pays, because documents rarely carry outcomes. A proposal doesn't say whether the deal closed. A spec doesn't say whether the feature shipped on time. What lifts the value is context from the systems around the documents (the CRM record that shows the deal closed, the ticket that shows the runbook worked, the email that shows the decision), and the years of history behind it.
So if documents are all you have, you can still license them, and a company with a decade of specialized SOPs has something buyers want. If you also run HubSpot, Zendesk, Jira or a similar tool, tell us on the scoping call, because documents plus those records are where a company's data starts to look like the six-figure deals on our business data page. Our guide to how much your company's data is worth to AI labs walks through what moves the number.
How the export works
Documents are the easiest system to export. Google Workspace, Microsoft 365 and Notion all have admin export tools, and a full export usually takes a few hours of your time and no engineering. Nothing gets installed and nothing connects to your systems. Troveo's process runs in four steps. Scope, where you take the data value assessment and we agree what's in and what's out. Agree, where you sign standard terms that cover the license and your earnings. Package, where we remove names and personal details to our documented standard, document what's included in a manifest, and get it ready for buyers. Get paid, where the first sale lands and you share in every sale after it.
You keep ownership the whole way through. Your company is never named publicly, and buyers are introduced to you only after an agreement is signed. If you want the step-by-step version for a company with more systems than just documents, our guide to licensing company data for AI has it.
What it pays
Typical full-company deals start at six figures. Documents alone will usually land below that, and how far below depends on your industry, how many years you've got, and how specialized the work is. A 40-person engineering consultancy with 12 years of specs and postmortems is a different asset from a 200-person company with three years of generic memos. There are no fees, no setup costs and nothing taken out of your payouts. And one dataset turns into several products over time (an export, an enriched dataset, a continuous feed, AI performance data), so the first sale is the floor. Our guide to exclusive vs non-exclusive data licenses explains why the same documents can earn more than once.
Where to start
Take the free data value assessment. It's ten questions about your systems, history and industry, takes about five minutes, and gives you an estimate of what your company's data could license for. Answer "documents and files" as one of your systems and be honest about the rest. The estimate is directional, and a scoping call gets you a real number. If you'd rather see the wider picture first, our guide to selling data to AI companies covers who's buying and what sells.
Where Troveo fits
Troveo helps companies understand what data they hold, protect what matters, and license what's valuable, selectively. We've paid rights holders more than 20 million dollars across video, audio, gaming, robotics and now company data, and documents and files are one of the nine categories we license. We handle the removal of personal details, the packaging and every buyer conversation. You decide what's in and what's out. Start with the data value assessment, or talk to us about what's in your Drive.