01

Custom AI products

We design and build complete AI products, from the first working prototype to software that is ready for real users. We handle the product thinking as well as the engineering, so what we build is useful as well as clever.

  • Product and interaction design
  • Web applications and APIs
  • A shared, written definition of what good output looks like
  • Launch planning and support
02

AI agents and assistants

Assistants that answer questions, draft text, sort incoming work and take actions inside existing tools. We design them with explicit limits: what they can see, what they can do, and when they hand over to a person.

  • Conversational assistants that keep track of context
  • Assistants that complete tasks in the tools a team already uses
  • Review and approval steps for sensitive actions
  • Sensible behaviour when the assistant is unsure
03

Integration and infrastructure

An AI product is only as useful as its connection to the rest of the business. We build the parts that make it dependable: links to your data and systems, and foundations that cope when demand rises.

  • Connections to existing data, systems and internal tools
  • Foundations sized for the busiest day, not the average one
  • Graceful handling of failures and slow-downs
  • Attention to speed and running cost from the start
04

Ongoing operation

Launch is the beginning. Usage grows and needs change. We can stay on to run what we build, or hand it over cleanly to your team with what they need to run it themselves.

  • Monitoring of quality, speed, reliability and cost
  • Regular review, with fixes where needed
  • Careful release of improvements
  • Documentation and handover to an in-house team

The kinds of things we build.

These are illustrations of typical work, not descriptions of specific past projects.

  • A support assistantAnswers routine questions from a company's own documentation and passes harder ones to a person, with the conversation so far.
  • A drafting toolWrites first versions of replies, summaries or reports for staff to check, edit and send.
  • A sorting pipelineReads incoming messages or documents, labels them and routes each one to the right queue or team.
  • A high-volume conversational serviceHolds many conversations at the same time, each with its own context, within a fixed budget for response time and cost.
  • An internal question-answering toolLets staff ask plain-language questions and get answers drawn from the company's own records, with sources shown.
  • A document workflowPicks out the key details from incoming documents and fills in the forms and records that depend on them.

Not sure which of these you need?

Describe the problem in a few sentences and we will suggest where to start.