Turn your analytics into a better knowledge base

Analytics isn't just a scoreboard — it's a to-do list for your knowledge base. Every conversation the AI couldn't handle is a hint about something your knowledge base doesn't cover yet. Here's how to close that loop.

Why the two are connected

AcornReply drafts replies by pulling from your published knowledge base. So the quality of your KB directly sets the ceiling on how much the AI can handle:

When a good article exists, the AI can draft a grounded reply — or the customer self-serves before they even write in.
When it doesn't, the conversation falls through to your team, and shows up as needed a human in analytics.

That means your outcome mix is really a map of your KB's strengths and gaps.

A simple weekly routine

Open Analytics → Overview and set the range to the last 7 or 30 days.
Look at the "needed a human" slice. A large share here is where the AI is coming up short. Some of that is genuinely human work — but a lot of it is repeat questions a good article would have answered.
Check your self-served slice over time. If it's climbing, your articles are pulling their weight. If it's flat or low, customers aren't finding answers on their own.
Open the Channels page. Compare the outcome split across email, contact form, and chat. A channel with a low AI-handled rate often has its own kind of question your KB doesn't address yet.

From a gap to an article

Once analytics points you at a weak spot, go find the actual questions behind it:

Open a handful of recent resolved conversations that needed a human. What did your teammate end up explaining?
If the same answer keeps coming up, that's an article waiting to be written. Turn it into a knowledge-base entry — a short FAQ for a quick answer, or a longer article for something that needs steps or context.
Publish it. Only published entries are used to ground drafts, so an unpublished draft won't move your numbers.

Watch it pay off

Give it a week, then come back to Analytics. As your coverage improves, you should see the AI resolved and self-served slices grow and needed a human shrink — and your team spends its time on the conversations that genuinely need a person.

Tip: don't chase 100%. Some conversations should reach a human. The goal is to stop spending time on the ones that shouldn't.