Know a company with years of records? Refer it and earn up to $100K
SOPs, wikis and process documents

SOPs and Knowledge Bases: Do AI Companies Pay for Them?

Yes. Buyer programs name SOPs, process documents and internal wikis, and one lists them first. What they want, what you can license, and what no one prices.

Short answerChecked

Yes, buyers ask for SOPs and knowledge bases by name, though no deal or price for SOPs on their own has been made public. At least four buyer programs name them, and micro1 lists "standard operating procedures (SOPs), process documentation, internal knowledge bases" first among the data it wants, for companies with 30 or more full-time employees. The published ranges, such as $100K to over $3M, cover a company’s operational data as a whole.

Programs naming SOPs or wikis
4+micro1, Appen, SuperAnnotate, Handshake (19 Nov 2026)
micro1 size floor
30+ staffFull-time employees; English; US first
Public price for SOPs alone
None foundChecked 19 Nov 2026
Rolled drawings, a hard hat and a tape measure in a site office
Key points
  • SOPs, process documents and knowledge bases are now a data type in their own right: micro1, Appen, SuperAnnotate and Handshake all name them.
  • Buyers want the real working version with its revision history and exceptions, de-identified, not a cleaned-up rewrite.
  • SOPs are worth most when they come with the records of the work they govern, such as tickets, work orders and approvals.
  • No buyer has published a price for SOPs alone. Treat any figure for documents alone as a negotiating position.

Do AI companies pay for a company’s SOPs and internal knowledge base?

Yes. Standard operating procedures, process documents and internal wikis have become a data type buyers name on their own, next to support tickets and code. The reason is simple: AI agents are being trained to do company work, and an SOP is the written version of how that work is supposed to run. The table shows who names these records, what each says about price and size, and the rules that decide what you can license. Every row was read on 19 November 2026.

Buyer or sourceStated range or ruleRecords asked forDate checked
micro1 (data partnerships FAQ)$100K to over $3M for operational data as a whole; 30+ full-time employees; US first"standard operating procedures (SOPs), process documentation, internal knowledge bases, workflows, templates"19 Nov 2026
Appen (enterprise data page)Terms in writing before extraction; payment on acceptance; no SOP price"Procedures, knowledge bases and templates that define how a task is meant to run"19 Nov 2026
SuperAnnotate (data partnerships page)Indicative $100K to $1M+ for a partnership as a whole"Process docs, runbooks, internal wikis, onboarding and training material, checklists, templates."19 Nov 2026
Handshake (data partnerships page)No price publishedWork instructions (manufacturing); process documentation (finance, legal, professional services)19 Nov 2026
Offer posted on LinkedIn (property management)Up to $2M, as the poster reported it; unverified"lease-administration procedures and work-order histories"Posted 20 Oct 2026; read 19 Nov 2026
18 U.S.C. 1839 (trade secrets)Procedures are protected only if the owner took "reasonable measures" to keep them secretAny SOP you treat as a trade secret19 Nov 2026

The program pages were read on 19 November 2026 and are quoted without a link, because they are intake pages. The trade secret definition is in 18 U.S.C. 1839. The offer is a public post, not a confirmed deal: LinkedIn, 20 October 2026.

Which companies pay for SOPs, process documentation and knowledge bases for AI training?

Four programs name these records on their own pages. micro1’s FAQ answers "What types of data are you looking for?" with standard operating procedures first, then process documentation, internal knowledge bases, workflows and templates. It also lists "Operational SOPs, logistics processes, fulfillment workflows, inventory management, and internal operations" for operations teams, and financial SOPs for finance teams.

Appen scopes three categories of company data, and the first is operational documentation: "Procedures, knowledge bases and templates that define how a task is meant to run". It adds a line owners rarely expect: "Most companies underestimate this category. It is usually the fastest to scope." SuperAnnotate lists "Process docs, runbooks, internal wikis, onboarding and training material, checklists, templates." Handshake names work instructions for manufacturers and process documentation for finance, legal and professional services firms.

Who sits behind these programs: frontier AI labs and companies building AI agents for business work. The programs collect and prepare the records and license them on. A name here is a market fact, not a recommendation, and terms change. For how SOPs fit with the other records an agent learns from, see licensing workflow data for AI agents.

How much is a company’s internal documentation, SOPs and process manuals worth to AI companies?

No buyer has published a price for SOPs or a knowledge base on their own, and we found no reported deal for documentation alone as of 19 November 2026. The figures you see are ranges for whole partnerships. micro1 shows $100K to $500K for "Qualified Enterprise Data Partnerships", $500K to $3M for large datasets across several teams, and $3M or more for highly unique data. SuperAnnotate shows an indicative $100K to $1M+. Those prices cover everything a company licenses, usually several systems and years of history.

Inside a deal, documentation tends to add value in three ways. It explains the records: an SOP for handling a warranty claim tells a lab what the claim tickets mean. It shows change: a procedure with ten years of revisions shows how the business learned. And it carries exceptions: the notes on when to break the rule are the part models get wrong. A wiki of 200 pages that nobody edited since 2019 adds little on its own.

So the useful question is not "what is our wiki worth" but "what is our wiki plus the records it governs worth". A property manager’s lease procedures with the work-order history behind them, as in the offer below, is a stronger package than either part alone. To see a figure for your own records, check what your data may be worth.

Behind the number: do buyers want the clean version of our SOPs or the real one?

The real one. A commenter on the property management post asked exactly this: "do they want the clean version or the real one? Because those are two very different histories". The buyers’ own pages answer it. Appen wants the procedures "including the exceptions nobody ever wrote into the official version", and lists "Revision history and rationale" among what it extracts. SuperAnnotate says: "The reasoning matters more than the finished output."

In practice "the real one" means the working documents as your team uses them: the page history, the comments, the version that says "do not use, see v3", the checklist with handwritten workarounds scanned in. Do not rewrite or tidy documents before a buyer sees a sample. Rewriting removes the signal and can look like the records were made for the sale.

"Real" does not mean raw personal data. Employee names, customer details and anything a client gave you in confidence come out or get replaced before delivery. Several programs say they remove personal information themselves, but your license should still say who does it, to what standard, and that you approve a sample first. See how to de-identify company data before licensing.

How can a services company license its SOPs and internal wiki to AI companies?

We help you license your data. Origin is an independent data partner. Your firm decides what is shared and signs any license itself, and a buyer may pay Origin when a deal closes. For a services company the work usually looks like this.

  • Map where procedures live: Confluence, Notion, SharePoint, Google Drive, a ticketing system’s help center, training slides, scanned binders. Note the years each covers and whether page history survives.
  • Link procedures to the work they govern. A dispatch SOP next to two years of dispatch tickets is the package buyers describe; the SOP alone is the start of one.
  • Pull out what you do not own: vendor manuals copied into the wiki, client-supplied procedures, standards documents you licensed, and pages that quote client contracts.
  • Mark personal data: staff names in page history, customer names in examples, home phone numbers in on-call pages.
  • Prepare a small de-identified sample, then compare offers. Read what to check in a data licensing agreement before you sign anything.

If your procedures include runbooks for IT or operations, the runbook template guide shows what a complete runbook holds and why its history matters. When you are ready, start with a description of what you hold.

Can a company license its Confluence and Notion knowledge base for AI training?

Yes, the content is yours to license, subject to what is inside it. The Atlassian Customer Agreement says: "As between the parties, Customer owns all intellectual property and other rights in Customer Data and Customer Materials provided to Atlassian or used with the Products." In plain terms: Atlassian does not own your Confluence pages; your company does. The same holds for most workspace tools, but read your own plan’s terms.

Owning the workspace is not the same as owning every page in it. Pages may quote vendor manuals, client procedures or paid standards, and those belong to someone else. Contractors’ pages may belong to the contractor unless their contract assigned them to you. Under the Copyright Office’s Circular 30, work an employee does in the regular course of the job is a work made for hire, so "the party that hired the individual is considered both the author and the copyright owner of the work". Outside contractors are a different case.

Exports have gaps to plan for. Notion’s help center says a workspace export covers "all your pages as HTML, Markdown, or CSV (for databases), along with any files you’ve uploaded", but "Pages that the exporter doesn’t have access to, such as private pages of other users, will not be included in the export." For Confluence, a space export and a full site backup behave differently, and page history needs the right export type. The guide to exporting Jira and Confluence data covers each option.

Will licensing our SOPs give away trade secrets or help a competitor?

It can, which is why the license matters. Federal law lists "methods, techniques, processes, procedures, programs, or codes" among the kinds of business information that can be trade secrets, but only if "the owner thereof has taken reasonable measures to keep such information secret" (18 U.S.C. 1839). In plain terms: an SOP stays protected only if you keep treating it as confidential. A license with confidentiality terms, limits on use and deletion at the end is one of those measures. Sharing the same documents with no terms is not.

Competitors are the other worry. Under the property management post, one commenter asked whether the buyer was "a management company competitor looking to build their own native AI system". Some programs let you control it: SuperAnnotate says of buyers, "You approve each one, and can block competitors permanently." Ask for the same in writing from any buyer, together with a ban on resale or sublicensing without your consent, a field-of-use limit, and deletion of raw files at the end of the term.

Be realistic about what cannot be undone. Once a model has trained on your procedures, deleting the files does not remove what the model learned. If a procedure is the core of your edge, keep it out, or license it for evaluation only.

This page is general information, not legal advice.

Should we take an offer of up to $2M for our procedures and work-order histories?

Look at what the number covers before you look at the number. On 20 October 2026 a property management newsletter writer posted on LinkedIn that "An AI data lab wants to pay me up to $2M for" lease-administration procedures and work-order histories. That is a reported offer, not a confirmed deal, and "up to" is the top of a range. One commenter said offers like it start high and "end up offering 10-100x less".

Questions that separate a real offer from a fishing line: who the end buyer is, and whether you can approve it; what exactly is licensed (procedures only, or the work-order history too); whether the price is fixed per delivery or depends on what the buyer later accepts; whether resale is allowed; who removes tenant and staff details; and when you get paid. Run any offer through the offer checker and the guide to AI data broker offers.

Origin Data Partners works with companies that own data and helps get it bought. We are independent of the services on this page, and a buyer may pay us when a deal closes (how Origin is paid). See what your data is worth.

Common questions

An AI data lab wants to pay up to $2M for our lease-administration procedures and work-order histories. Should we take the deal?

Not before you know what the $2M covers. Ask who the end buyer is, whether the price is fixed or depends on acceptance, whether the work-order history is included, who removes tenant and staff details, and whether resale is barred. A public offer like this was posted on 20 October 2026 and remains unverified. Compare it with at least one other route before signing.

An AI data broker offered to buy our company’s old SOPs and wiki. Is that legit, and what should I check before I sign?

It may be. Check that the license names the end buyer or lets you approve each one, bars resale and sublicensing without your consent, fixes the price and payment date in writing, and lets you review a de-identified sample first. Keep vendor manuals, client procedures and pages with personal data out. Ask for deletion of raw files at the end of the term.

I am the lawyer for a services company. Can it license SOPs that quote client contracts or vendor manuals?

Usually not those parts. Client contracts are typically confidential, and vendor manuals and paid standards are owned by their authors, so the company cannot grant rights it does not hold. The usual approach is to strip quoted client terms and third-party text, keep the company’s own procedures and history, and record what was removed. Check each client agreement’s confidentiality clause. This is general information, not legal advice.

Is there a minimum company size for licensing SOPs and a knowledge base to AI companies?

It depends on the program. micro1 looks for "established companies with 30+ full-time employees", data mainly in English, and US companies first, though it says there is no fixed minimum number of files. SuperAnnotate says any size and any industry can qualify if the records built up over time and you hold the rights. Years of history matter more than page count.

Do we have to remove employee names from our wiki before licensing it?

Plan on it. Page history, comments and on-call pages carry staff names, emails and phone numbers. Several programs say they remove personal information themselves; SuperAnnotate removes identifying detail "to the standard your team sets". Your license should still say who does it, to what standard, and that you approve a sample before anything is used.

Are SOPs worth more alone or bundled with tickets and work orders?

Bundled, in almost every case. A procedure tells a model how work should run; the tickets, work orders and approvals show how it actually ran, including the exceptions. Appen describes both categories and says the decision and escalation records are what frontier labs ask for most often. No buyer has published a separate price for SOPs, so bundling is also where the published ranges apply.

Can we license procedures that a consultant wrote for us?

Check the consulting contract first. Under US copyright law a work made for hire belongs to the hiring party only in two situations: an employee’s work in the regular course of the job, or certain types of work under an express written agreement. A consultant’s procedures may stay the consultant’s unless the contract assigned them to you. An assignment clause settles it.

Sources and further reading

What is your company’s data worth?

AI companies pay $20K to $5M for company records. Six questions, no files.