ServicesAI & automation
AI that answers from your content and shows its sources.
Assistants, search and automation grounded in your own content, with citations a user can check.
What we build
We build assistants, search and automation that work from your own documents and data.
Responses are grounded in sources you approve and cite them where appropriate, so users can check what they are told. Each feature is evaluated against representative queries before release, with a fallback for questions the sources cannot answer confidently.
- 01
Retrieval over your documents, catalogue or knowledge base
- 02
Responses grounded in approved sources, with citations where appropriate
- 03
Evaluated against representative queries before release
- 04
Fallbacks for when there is not enough evidence to answer confidently
Who it is for
Teams with a body of content, such as documents, a catalogue or a knowledge base, that users need answers from.
Working together
What an engagement involves
- Agreeing the approved sources and a set of representative queries
- Evaluation before launch, against a quality bar agreed with you
- Usage limits and cost monitoring, with running costs estimated in the quote
- A plain recommendation when a conventional feature would serve users better
Shipped work
In our own products
DeenGuide's assistant answers from the Quran and authentic Hadith and cites what it drew from.
DeenGuide
An Islamic companion app with an assistant that cites its sources.
Read the DeenGuide case study
How do you reduce wrong answers?
Responses are grounded in sources you approve and cite them where appropriate. We evaluate against representative queries before launch and design a fallback for questions the sources cannot answer confidently. No system of this kind is error-free.
Is our data used to train AI models?
We choose providers and settings with you, and document where your content is stored and processed and whether it may be used for training.
What does it cost to run?
Running costs depend on usage. We estimate them with your quote and build in limits and monitoring.
What happens next
Three steps to a plan
- 01
We review your message
One of the founders reads it, whether it came through the form or by email.
- 02
We ask the questions we need answered
Usually a short call, to understand the goal, the users and the constraints.
- 03
We recommend the next step
A written quote, a discovery sprint, or a reason not to build it at all.
Considering an AI feature?
A few sentences are enough to start.
- Where we are
- Zero Horizon Ltd, United Kingdom