APPLIED AI / KNOWLEDGE ASSISTANTS
Answers your people can check
We build assistants that answer from your own documents, cite their sources and respect existing access rights.
People spend too long finding the right policy, procedure or past piece of work. A knowledge assistant answers in plain language from your own content, shows where each answer came from, and only retrieves what the person asking can already see.
What we do
Permission-aware retrieval
Retrieval scoped to what each user can already see in the source systems, so the assistant never widens access.
Answers with citations
Every answer links to the passages it was drawn from, so people can check it before relying on it.
Quality and evaluation
Test questions from your teams, accuracy targets you set and feedback that improves answers over time.
Our solutions
A single place to ask about policies, procedures and guidance.
- Plain-language questions
- Cited answers
- Links to the source documents
Assistants for a specific team, grounded in that team’s documents and past work.
- Scoped content sources
- Templates and drafting help
- Feedback from users
The assistant where people already work, rather than another app to open.
- Embedded in your intranet or workspace tools
- Single sign-on
- Usage and quality reporting
How it runs
Assess
We review content sources, access rules and the questions people actually ask.
Pilot
An assistant on a defined set of your content, measured against an accuracy target you set.
Scale
More sources, permission sync, monitoring and integration into everyday tools.
Run
We keep content and quality current, or hand over to your team with a run guide.
What you get
- An assistant that answers from your content with citations.
- Retrieval that respects existing access rights.
- An evaluation set and accuracy report.
- Usage and quality monitoring.
- A content and run guide for your team.
Questions
Can users see documents they are not allowed to?
No. Retrieval is scoped to what each user can already see in the source systems, and we test that before go-live.
What happens when the assistant does not know?
It should say so. We design and test for that, and every answer shows its sources so people can check it.
Is our content used to train the model?
We do not use client data to train or fine-tune models, unless a client asks us to in writing for their own dedicated model.