What AI buyers want from a knowledge base
An AI agent doing real work needs the same reference material your staff use. Knowledge bases supply that domain knowledge.
| Type | Examples | Why buyers want it |
|---|---|---|
| External help center | FAQs, troubleshooting, product guides | Answers agents must give customers |
| Internal wiki | Confluence, Notion, SharePoint pages | How the company really works |
| Agent macros | Saved replies and templates | Approved language for common cases |
| Technical docs | Runbooks, architecture notes, design docs | Engineering knowledge for coding agents |
Public help-center pages may already be on the web. Private internal content is scarcer and usually more useful. Runbooks and troubleshooting trees are especially useful, because they spell out decisions step by step. Stale pages still have value if they carry dates, since they show how a process changed.
What makes a knowledge base worth more
- Depth. Detailed, specialized content in a niche industry.
- Version history. Edits over time show how processes changed.
- Links. Articles tied to the tickets that used them. Sunset prices on how well records link across systems (Forbes).
- Coverage. Many teams documented, which supports the published $500K+ multi-team tier.
Price signals
We found no public price for a knowledge base sold alone. The closest published figures:
- Enterprise programs: $100K+ qualified, $500K+ large-scale or multi-team, $1M+ highly unique, for a company's data package.
- Troveo: business-data deals "typically start at six figures" (Troveo, company claim).
- Defined.ai says several marketplace partners earn $1M+ per year licensing datasets (Defined.ai, company claim).
Risks specific to wikis
- Secrets. Wikis often hold passwords, API keys and network details. Scan for them.
- Copied content. Pages pasted from vendor manuals, standards or paid courses may not be yours to license.
- Client material. Client-specific pages may fall under NDAs. See client data vs. company data.
- HR pages. Frankfurt Kurnit recommends excluding HR, health, privileged and union data entirely (Frankfurt Kurnit).
- Personal data. Staff directories, org charts and meeting notes name employees, who count as consumers under the CCPA (Cal. Civ. Code 1798.140).
General information only, not legal advice.
How to prepare a knowledge base
- Export spaces with page history and links intact.
- Exclude HR, legal, finance-confidential and client spaces.
- Run a secrets scan and remove credentials.
- Remove staff directories and replace names with roles.
- Flag pages copied from outside sources.
- Bundle with the SOPs and tickets the articles support.
Related: selling SOPs and how to anonymize company data.
Who buys knowledge bases
Several kinds of buyers want documentation. Each works differently.
| Buyer or program | What it says it takes | How it pays |
|---|---|---|
| Enterprise data programs | Knowledge bases, internal docs, SOPs and more from operating companies | Published tiers from $100K+ |
| Turing Project Lazarus | Internal wikis and design docs from defunct startups | No public prices |
| Protege | Governed marketplace for training data | Revenue share per use |
| Scale AI Dataset Collection Initiative | Non-public datasets submitted by companies | Depends on quality and use case |
A flat license fee gives you certainty. A revenue share can pay more over time if your data gets used often, but income is less predictable. Buyers in our network differ on which model they offer, so compare terms with the data licensing agreement checklist.
Operating vs. shutting-down companies
Knowledge bases fit operating companies well. You keep using them after licensing a copy, and enterprise programs state the company keeps ownership. For shutting-down startups, internal wikis often go with the codebase, as in Turing's Project Lazarus (Turing).
See what your documentation may be worth: take the 60-second estimate.
Frequently asked questions
Can I sell my company's internal wiki to AI companies?
Generally yes, if you own the content. Remove secrets, HR pages, client material and copied third-party content first.
Is a public help center worth anything to AI buyers?
Less than private content, since public pages may already be on the web. Private internal docs and version history are usually more useful.
Do I keep my knowledge base after licensing it?
Under typical licensing terms, yes. You license a copy and keep ownership. Check the agreement.