AI Knowledge Assistants

AI customer support assistant

Help customers find answers from your approved support knowledge, with references where they help and a hand-off to a person when the assistant does not know.

The problem

The same questions, every day

Support answers questions that are already in the help centre.

Customers wait for simple answers

Easy questions queue behind hard ones.

What is included

Your approved knowledge

Answers come from the help centre and documents you approve.

References

Answers refer to the article they came from where that helps.

Clear disclosure

Customers are told they are talking to an AI assistant.

Escalation

Questions it cannot answer go to a person, with the conversation so far.

An example, step by step

Example · synthetic data A delivery question answered from the help centre

Input

A customer question

My parcel was marked delivered but I did not receive it. What should I do?

  1. Search the approved help articles for the matching policy.
  2. Answer with the steps from the policy and a link to the article it came from.
  3. If the policy does not cover it, or the customer asks for a person, hand over with the conversation.

Output

Answer, source or hand-off

A short answer with a link to the missing-parcel article; a question about a refund over the agreed limit goes to a person instead.

How it works

A customer asks

The question arrives in the chat or form on your site.

The assistant searches and answers

It finds the relevant passages and answers from them.

A person takes over when needed

Unclear or sensitive questions are handed to your team.

What you receive

A support assistant

Connected to your approved knowledge, on your site.

An evaluation

Results on a set of real questions before launch.

A list of unanswered questions

What customers asked that the knowledge does not cover yet.

What it works with

Your approved support content

Help articles, policies and FAQs you choose.

Your support channel

Website chat or helpdesk where the assistant is placed; named once confirmed.

Your hand-off process

How a conversation reaches a person, and when.

How we check it works

Grounded answers

A set of known questions checks that answers follow the approved content and cite it.

Unanswered questions

Questions it could not answer are listed, so content gaps can be filled.

Hand-off works

Requests for a person, and topics on the hand-off list, reach a person in testing.

What to keep in mind

  • The assistant answers from your knowledge; gaps in it show up as questions it cannot answer.
  • It can still be wrong; references and escalation let customers and your team check.

What shapes the scope and quote

  • How much support content there is and how current it is.
  • The number of languages and channels.
  • How the hand-off connects to your current support tools.

Results

Results from client projects

Published only with each client's approval.

Client project AI Agents & Automation
Chiropractic clinic: Rosa, an AI client assistant on WhatsApp and email
Chiropractic clinicAmsterdam
Client project AI Knowledge Assistants
A Dutch telephony provider: an AI customer-service platform for calls and email
A Dutch business telephony providerNetherlands

Frequently asked questions

Will customers know it is an AI?

Yes. The assistant says it is an AI assistant and offers a person when it cannot help.

What if the answer is not in our documents?

It says it does not know and hands the question to your team, so you can add the missing information.

Can it look up a customer's order?

Only if you choose to connect it and agree what it may see. A first version usually answers from your help content only.

What if our help articles are out of date?

The assistant answers from what you approve, so outdated content gives outdated answers. The evaluation shows where articles need updating.

Start a project

Have a business problem AI could help solve?

Tell us what you're trying to improve. We'll help you identify a practical next step.