Prepare your content for the AI knowledge base
What the knowledge base is for, and how to get your documents ready before the feature ships.
Step-by-step
- Step 1
Collect the questions customers ask most
Look back over the last one to three months of conversations and collect the questions that repeat most often, together with the answers your team actually gives. The first 20–30 questions are enough to start with.
TipThe questions your team retypes most often are where AI saves the most time — rank them by frequency.Screenshot pendingCollect the questions customers ask most - Step 2
Collect your policies and terms
Gather the policy documents your team references often — admission criteria, processing times, refund policy, or the documents a customer has to prepare — and always note when each was last updated.
Screenshot pendingCollect your policies and terms - Step 3
Collect product and service details
Write up your products, services or programmes in full: name, price, conditions and exceptions. Incomplete details lead to vague answers that the team has to correct afterwards.
Screenshot pendingCollect product and service details - Step 4
Write it in a form the AI can use
One entry should answer one thing. Keep sentences short and direct, avoid internal abbreviations or spell them out, and delete outdated information every time you revise.
TipIf a new hire can read an entry and answer a customer straight away, it is in good shape for the AI too.Screenshot pendingWrite it in a form the AI can use - Step 5
Give each area an owner
Assign someone to review and update each area of the content, and set a review cadence. A knowledge base with no owner drifts out of date and starts producing wrong answers.
Screenshot pendingGive each area an owner