FAQ assistant that answers
from your knowledge base
Half of any support chat is the same ten questions on repeat. Load your answers into Telm once — the bot recognizes those questions in any wording and replies instantly, strictly from your materials. If your knowledge base has no answer, it stays silent instead of making things up.
- Instant answers 24/7
- Only your own materials
- Never invents answers
How the assistant works
Your knowledge base
Add articles, rules, how-tos and answers in the dashboard. The assistant treats them as the single source of truth for your group — update an entry once and every future answer reflects it.
Understands rephrased questions
Semantic search matches meaning, not keywords: 'how do I get my money back', 'refund?' and 'can I cancel my order' all find the same answer, in any language your members write.
No answer? No guessing.
If the knowledge base doesn't cover a question, the assistant stays silent and lets humans take over. It never improvises — so it can never mislead your members with confident nonsense.
Flood-proof by design
Per-member and per-group limits keep the assistant from being spammed into noise. It answers real questions, not attempts to turn the bot into a toy.
Why groups turn it on
- Admins finally sleep - the same ten questions get answered instantly at any hour, in any timezone, without pinging a human.
- One consistent answer - no more three different admins giving three different explanations — everyone gets the canonical answer from your knowledge base.
- Answers you can stand behind - because every reply is grounded in your own materials, you never have to apologize for something the bot invented.
Grounded answers, not chatbot improvisation
Generic AI chatbots are risky in a community: they answer everything, including things they don't know, and one confidently wrong answer about your refund policy costs more trust than a hundred correct ones earn. Telm's assistant is built the opposite way. It only answers when your knowledge base actually contains the answer — the AI's job is to recognize the question and pick the right materials, not to compose new facts.
Under the hood, each question is matched semantically against your entries, the top matches become the grounding for the reply, and anything below the confidence bar means silence. The result behaves like your best moderator on their best day: instant, accurate, on-message — and honest enough to say nothing when the answer isn't written down yet.
Frequently asked questions
Will the assistant answer questions it doesn't know?
No — that's the core design rule. If your knowledge base has no matching entry, the assistant stays silent and the question remains for humans. It never composes answers from general AI knowledge.
Does it understand differently worded questions?
Yes. Matching is semantic, not keyword-based: 'how to get a refund', 'money back?' and 'cancel my purchase' all resolve to the same knowledge-base entry, even across languages.
How do I add or update answers?
In the dashboard, on the group's FAQ page: each entry is a short title plus the answer text. Changes apply immediately — the very next question uses the updated material.
Can members spam the assistant?
Rate limits per member and per group keep it under control. The assistant is there to answer genuine questions, and the limits make flooding it pointless.
Related features
Explore more ways Telm can help your community