Tenali AI Docs
Admin & Teams

How Tenali Learns From Your Calls

Tenali builds your team's answer library from the meetings it records, with no admin action — and how to curate what it learns.

Live answer quality depends on what Tenali knows. That knowledge used to come only from what an admin uploaded or imported. Now every meeting Tenali records feeds the answer library automatically.

What Happens After a Call

Once a recorded meeting has been summarised, Tenali looks back over the transcript for questions your customers asked and your team answered well, and adds those pairs to your workspace's FAQ library.

This runs off the back of the summary, so a meeting only contributes when it has both a usable transcript and a completed summary. A silent recording, a call too short to be a real conversation, or a meeting whose summary never generated produces no FAQ candidates at all.

It draws on two things:

  • Answers your rep actually gave on the call. If a customer asked something and the rep answered it clearly in the conversation, that becomes the canonical answer.
  • Questions Tenali already detected live. Questions surfaced during the call that ended up with a real answer are promoted into the library too.

A rep's in-conversation answer wins over a generated one for the same question, on the basis that the person on the call said it best.

Where Curated Answers Fit

Automatic learning doesn't override what your team has written. The order of preference is:

  1. Answers an admin wrote on the FAQ Browser — always win.
  2. Answers a rep gave on a call — captured automatically as above.
  3. Answers Tenali generates from your connected sources at the moment the question is asked.

Curated and promoted answers are now used during live calls, not just in chat afterwards, so writing a good answer once changes what reps see on the next call.

What It Won't Do

  • Duplicate what you already have. Every candidate is checked against your existing FAQs first, so a question that's already answered isn't added again.
  • Learn from questions you've suppressed. Anything on the admin suppression list stays out. See Question Intelligence.
  • Learn from a call that wasn't really a conversation. Very short transcripts are skipped, as are recordings that captured no speech and meetings whose summary failed to generate.
  • Flood your library. There's a cap on how much any single meeting can contribute, so one unusual call can't swamp the set.
  • Add a weak answer. Only clearly-answered questions qualify; a vague or partial answer is left out rather than becoming canonical.

What Admins Should Still Do

Automatic learning gets you a useful library without effort, but it learns what your team says, not necessarily what you'd want said. Worth doing periodically:

  • Review the recurring questions in Question Intelligence — the ones asked most often are the ones worth getting exactly right.
  • Write the canonical answer yourself for anything pricing, security, legal, or competitive. Don't leave those to whatever a rep said on a Tuesday.
  • Suppress questions that shouldn't be answered from the library at all.
  • Watch for drift. If an answer is out of date because the product changed, edit it — the library won't notice on its own.

Was this page helpful?