AI agent for customer service managers
Script and Knowledge Article Gap Agent
Close the top gaps in scripts and articles and confirm that they reduce long calls and repeat contacts.
What it does
Agents give three different answers to the same refund question because the article is two years old. This agent reads the past month of calls and tickets and finds the questions that repeat and the calls that run long. For each topic, it checks whether a current article or script exists and whether it is recent and accurate. It drafts a new or updated article, using the best answers given in resolved cases and the current policy. It tests each draft against a sample of past questions to see if the article would answer them. If it would not, it revises the draft. After publishing, it watches the topic for handle time and repeat contacts to see if it improved. The supervisor approves publishing. Edge case: the policy itself is unclear, so the agent asks the policy owner.
How it works
Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.
Read the steps as a list
- Monthly review starts
- Read calls and tickets and group them by topic
- Rank topics by volume and handle time
- Search for a current article for each top topic
- Draft a new or updated article from resolved cases and policy
- Test the draft against ten past questions
- Does the draft answer at least 8 of the 10 questions?If not: revise the missing parts and test again. Back to step 4.
- Supervisor approves publishingThe agent waits here for your OK.
- After two weeks, read handle time and repeat contacts for the topic
- Did handle time or repeat contacts improve?If not: find what agents still ask and revise the article. Back to step 4.
- Updated articles and results
How it decides
It ranks topics by volume times handle time. A draft passes when it answers at least 8 of 10 sampled past questions.
- Rank topics by number of contacts times average handle time
- Treat an article older than 12 months as outdated
- Use only policy text that has an owner to confirm it
- Compare after two weeks against the prior four weeks
Make it yours
Every agent is a starting point. You choose these settings for your own situation.
- Review schedule
- Topic threshold for action
- Article format and tone
- Sources allowed for answers
- Results period (default 2 weeks)
What keeps you in control
It always asks you first
- Supervisor approves each article before publishing
Hard limits
- Never publish an article without approval
- Never state a policy that no owner has confirmed
It stops when
- Done: top gaps are closed and results are recorded
- Stop: policy is unclear and no owner answers
Set it up
We guide you through the set-up, step by step
Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.
- One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
- The agent then walks you through connecting your own data, one source at a time
- A downloadable copy with the flow chart, the rules and the full guide
