An AI knowledge assistant gives accurate answers when four things are in place. It only reads documents you've chosen, there's one current version of each, each section makes sense on its own, and every answer shows which document it came from. When an assistant gets things wrong, a lot of times the AI is fine and the documents are the issue.
So this is how to get the documents right.
1. Choose the sources deliberately
Don't point the assistant at your whole shared drive. It'll find the 2019 price list, the draft policy nobody signed off and someone's personal notes, and treat all of it as true.
Start with a short list:
- Standard operating procedures and process guides
- Current service descriptions and pricing
- Policies: refunds, returns, HR, security
- Frequently asked questions, from customers and from staff
- Templates and approved wording
For each one, name an owner who keeps it up to date.
2. One version of everything
If there are three versions of the onboarding checklist, the assistant might quote any of them. Before you load anything:
- Delete or archive old versions.
- Put a "last reviewed" date on every document.
- Where two documents say different things, fix it in the documents themselves.
This step on its own fixes a lot of wrong answers, and it helps your team whether you ever use AI or not.
3. Write so each section stands on its own
Assistants find and quote sections of documents, they don't read whole files. A section that says "as above" or "see the previous page" stops making sense when it's quoted by itself. Good habits:
- Clear headings that say what the section is about, like "Refunds for annual plans" instead of "Other".
- Name the thing instead of saying "it" or "this" across paragraphs.
- One topic per section. Long pages that mix three processes give muddled answers.
- Tables for facts like prices, limits and dates, so numbers aren't buried in paragraphs.
- Spell out acronyms the first time in each document.
4. Make every answer show its source
If your team can't check an answer, they can't trust it. The assistant should quote or link the document and section it used, so anyone can click through and check. It also shows you straight away when an answer came from the wrong document.
5. Teach it to say "I don't know"
The most important instruction is what to do when the documents don't have the answer. The assistant should say so plainly and suggest who to ask, and it shouldn't fill the gap with something that just sounds right. Test this properly by asking it questions you know aren't covered.
6. Control who can see what
Not everyone should get answers from every document. Salary bands, client contracts and disciplinary policies need the same access rules in the assistant as they have in your files. Set this up before you launch, not after someone asks the wrong question.
7. Keep it current
- When a document changes, the assistant should read the new version and stop answering from the old one.
- Look at the questions the assistant couldn't answer. Each one points to a gap in your documents.
- Re-test a standard set of questions after big changes, the same way you would test any AI agent.
A simple preparation checklist
- List the 20 questions your team asks most often.
- Find the document that answers each one. Where there isn't one, write it.
- Remove old versions and add review dates and owners.
- Fix headings and split up long pages that mix topics.
- Decide who can see which documents.
- Load them, then test with the 20 questions plus ten it shouldn't be able to answer.
What it gives you back
Done properly, your team finds answers without interrupting the one person who knows, new starters get up to speed quicker, and the knowledge stays in the business when people leave. Our AI knowledge assistants page goes through how we build them, including answers that show their source and a review step before anything is sent.
