Skip to the content

An AI agent on first-line support

Before a cohort started, support was taking 900 requests a day. The agent answers from the knowledge base with a link to the source.

Client
Online school, 12,000 students
Industry
Education
Year
2026
Duration
4 weeks
The problem

On the days a cohort started, the support queue reached 6 hours. 70% of the questions repeated: access to lessons, payment, deadlines, refunds.

Solution

What we did
and why we did it that way.

01

Cleaning up the knowledge base

We collected the internal policies and found 40 contradictions between documents. That gained more quality than the choice of model did.

02

No answer without a source

If the agent cannot cite a source, the answer is not shown to the student—the request goes to a human.

03

Training in the shadows

For two weeks the agent only suggested answers to the operator. That produced the labelled data, and no student ever noticed it was there.

04

Injection testing

We ran AI red teaming before switching it on: we found a way to extract someone else’s email address and closed it at the permissions level.

Result

What changed
and what we measured it with.

68%
of requests closed without an operator
40sec
average time to first response
94%
accuracy on a labelled set of 300 questions
11
average cost of a conversation
Stack
ClaudeLangChainQdrantFastAPIPostgreSQLLangfuse
Timeline

How it went
day by day and week by week.

  1. Week 1

    Knowledge base

    Collecting the documents and resolving the contradictions.

  2. Week 2

    Agent and tools

    Search across the documents, access to payment status.

  3. Week 3

    Shadow mode

    Suggestions to the operator, labelled data collected.

  4. Week 4

    Red team and launch

    Injection testing, switched on for 30% of traffic.

Screens

How it looks
in schematics.

These are screen schematics. We do not publish client interfaces without permission

Next

A similar problem
on your side?

Describe it in the brief. In working hours we come back with an estimate of time and cost within two hours.