AI engineeringSoftware engineering
AI that does real work in your systems, with a person in charge
We build agents that work through permissioned tools, private AI on hardware you choose, and the software around them.
- A request comes inAn email, a WhatsApp message, a document or a form.
- The agent prepares itIt reads the request, checks your systems and drafts the work within its rules.
- A person approvesSomeone on your team checks the work and signs it off. Every step is logged.
Built withModels, frameworks and platforms we build with.
- OpenAI
- Anthropic Claude
- Google Gemini
- Meta Llama
- Mistral AI
- WhatsApp Business
- Flutter
Names and logos belong to their owners. Listed to show the tools we use, not a partnership.
What we are building
We are building our own AI products, of the kind described above: agents that do real work in the systems people already use, with a person approving what matters. We will name them when they are ready for you to use, not before.
StatusIn development
About our productsWhat they are built around
- Your own systems They work in the tools your team already uses, through permissioned access.
- A person approves Anything that matters waits for someone on your side to sign it off.
- Private by design Your data stays where you choose, and every step is recorded.
Get updates when there is something to show.
Your data stays where you choose, and a person signs off
Every change passes a senior review and tests that can fail
Our AI agent writes the code. A person reviews every change before it is merged. Tests and a build gate decide when the work is done, whatever the agent reports.
How we build- Senior review of every change
- Tests on every change and a build gate before anything ships
- An independent reviewer that tries to prove the work wrong
- A person on your side signs off before it goes live
We build to the data protection rules and security controls your organisation works under. We hold no security certifications and claim none. The Security page sets out what goes where.
We also take on selected client projects
Alongside our own products, we build for clients where the work fits what we know: AI agents and private AI, and the software around them. Each offer has its own page.
Work with usAI that works
All AI offers- AI harness for your teamsA safe AI workspace for your staff, connected to your systems
- Agents for operationsReal work in your systems, approved by a person
- Connecting agents to your systemsPermissioned connections to your ERP, CRM and APIs
- Document AI and knowledge assistantsAnswers from your documents, with the source for each
- Assistants on web and WhatsAppAssistants that answer and act on your channels
- Agent evaluation and QATesting, monitoring and cost control for live agents
The questions buyers ask first, on data, approvals, cost and ownership, are answered on one page.Common questions
Every team has repeat work an agent can prepare
If a task repeats and follows rules, an agent like ours can prepare it and a person can approve it. Each card names the system the agent works in and the role that approves. These are examples of the work, not client stories or product features.
All use casesStart with one task
Name one task your team repeats and the systems it touches. We reply with what a first agent for it would look like, and who on your side would approve its work.