Service 05
AI product development
AI agents, assistants and AI features built into your product, with guardrails, evaluation and clear costs.
We build AI into products: assistants that answer from your own documents, agents that complete multi-step tasks, and features like search, summaries and drafting.
We start with one task and a test set of real examples. Then we pick the model, design the prompts and tools, and measure answers against the test set before anything ships.
Every AI feature ships with guardrails, logging and a cost estimate per request, so you know what it does and what it costs.
Questions
Which AI models do you use?
We choose per task. We test models from Anthropic and OpenAI, and open-source ones where hosting matters, against your own examples, then pick the one that answers best for the cost.
How do you stop an AI assistant from making things up?
We ground answers in your own content, limit what the assistant may do, and test it against a set of real questions before launch. We log answers so issues can be found and fixed.
What does an AI feature cost to run?
It depends on the model and on usage. We estimate the cost per request during scoping and set up monitoring, so you see the real cost as usage grows.
Want this for your team?
Tell us what you need built. We'll reply with questions or a plan.