AI features
AI inside your product, built to be trusted.
An assistant that answers from your content. Search that gives the answer and the source. A draft that is right the first time. We build the feature into your product, ground it in your data, and measure it before your customers see it.
What we build
Features that make a product easier to use, with the facts checked against the source.
Assistants and chat
Answer customer or staff questions from your own content, with the sources shown.
Search that answers
Retrieval over your documents and data, so people get an answer and the page it came from.
Drafting and summaries
Replies, reports and summaries in your tone, checked against the facts they are based on.
Classification and extraction
Route tickets, tag content, and pull structured data out of text.
How we build it
Grounded in your data
Answers come from your documents and systems, with citations, so that anyone can check them.
Evaluated before launch
A test set of real questions and hard cases, scored for correctness, groundedness, tone and safety.
Guarded
Prompt-injection and data-leakage checks in an agreed scope, and a clear answer when the model does not know.
Built in your stack
Your language, your cloud, your CI. The code and the test set live in your repository.
How a build runs
The same four steps as everything we do, with a fixed price agreed after the first one.
01Design
A kickoff call, then a design sheet: the job, the users, the risks, what the AI may and may never do, and what a good result looks like. We agree the scope and a fixed price.
02Build
In your stack and your repository, with the test set written alongside the code. You see working software early, not at the end.
03Prove
Every change is scored against the test set, with safety checks in an agreed scope. You get a report and a go or no-go recommendation before launch.
04Run
After launch, every prompt, model or tool change is re-tested. You get a monthly report on quality, cost and what changed.
Questions we get asked
Do we need our own model?
Usually not. Most features work well with a hosted model and good retrieval. Where data must stay in your environment, we use an open model that you host.
How do you stop made-up answers?
Grounding, citations, and a test set that scores groundedness on every change. When the model has no source, the feature says so instead of guessing.
Can you add AI to an existing product?
Yes. That is the usual case. We work inside your codebase and your release process.
Does it work in Arabic?
Yes. We build and test for Arabic and mixed Arabic-English use, with native speakers.
Arabic & bilingual AIWhich feature would help your users most?
Tell us about your product and the job the feature should do. We will come back with questions, then a proposal.
Founding clients get a lower price on their first project.