A repeatable operating system for every function of your business — installed one desk at a time, not a one-off build.
We sit with you and write down how the function actually runs today — inputs, decisions, hand-offs, what "good" looks like.
Email, calendar, CRM, forms, WhatsApp, Slack, your documents. Your data stays in your accounts.
Specialised agents with your tone, rules and templates. Approval rules for anything sensitive.
Agents work the queue. You approve, edit or reject from one inbox. Nothing risky goes out on its own.
We watch failures, tune prompts and workflows, and add capability monthly. You get a report and a roadmap.
"AI runs your company" is a claim we don't make. What we do promise: the boring, repetitive 80% gets done by agents, and the 20% that carries risk waits for a human.
Rules are configured per client and per desk. Loosen them as trust builds — or don't.
Not on day one. We install and operate one desk per business function — sales, support, content, research, the shop floor — with a human approving anything sensitive. Once a desk is reliable we add the next, until the functions you want covered are covered.
Two or three hours of your time in week one to map the process, access to the tools the desk will use, and someone who checks the approval queue daily (usually 10–20 minutes).
In your own accounts. Our system stores task records, approvals and outputs in a database we operate for you, exportable at any time. We can sign a DPA and discuss private deployment on Scale.
Sensitive actions never go out without approval, so mistakes are caught in the queue. Everything else is logged; we review failures weekly and fix the workflow, not just the one output.
Frontier models from the major providers, chosen per task. Routine steps run on cheaper models to keep usage low. We can restrict to a specific provider if your policy requires it.
Yes — 3-month minimum, then rolling monthly with 30 days' notice. You keep your prompts, workflows, data and exports.