Terms

Marketing material. Nothing on it is an offer capable of acceptance, and visiting it creates no relationship between us. Prices shown are our current published rates and can change; the price that applies to you is the one written into your agreement before work begins. Capability on this site is labelled live, preview or design partner, and those labels are part of how we describe what you are buying rather than decoration.

Three rungs, in order. Diagnose: a time-boxed read of your own operations that ends in numbers and a decision, credited against a build if you continue. Build: one workflow or one complete system, live in your own workspace, with its approval rules and its audit trail configured with your team. Operate: we run it — monitoring, tuning, cost and model routing, and a governance report, with a named operator rather than a ticket queue.

Current fees and what each rung includes are on the pricing page, and what is in scope for you is fixed in writing before the work starts. We do not bill for work you have not agreed to.

Scope, dates and price are written down first. From you we need three things: access to the systems in scope, a named person who can actually approve decisions, and the content or answers you want an agent to be able to use. Engagements that struggle almost always struggle on the second of those.

New agents run in shadow before they touch anyone: measured against your own history, with the scorecard shown to you, and go-live happens when the scorecard supports it rather than when the calendar does. That commitment goes into the agreement, which is the only place a commitment means anything.

Two commitments are on the table today. The first is the shadow-first rule above. The second is a sixty-day exit on ongoing operations: if it is not working, you leave with your data and your configuration, without an argument about the remaining term.

Two further guarantees are written and deliberately not offered yet — one on the audit ledger, one on spending ceilings — because workspace-level ceilings are in preview and the downloadable evidence bundle is not built. A guarantee the platform cannot enforce is worth less than nothing, and it is precisely the kind of promise that regulators have been fining vendors over. We will offer both when the product can stand behind them.

Your content and your business data stay yours. You can ask for an export at any point, and we design deliberately for you being able to leave: your front end is on your own domain, the model providers underneath are swappable, and a private deployment is a delivery engagement we will quote rather than a lock we will defend. Retention and deletion are set in the agreement — see the privacy summary for the website side and the security page for the product side.

• No uptime figure and no service level, today. We do not publish one because we do not yet measure availability in a way we would defend in a contract. What we will commit to is written into the agreement.
• No warranty that an AI system will be right. It will sometimes be wrong. Risk tiers, human approval on anything irreversible and an audit trail are how that is managed — they are a design, not a guarantee of correctness, and anyone selling you the second thing is selling you something that does not exist.
• No transfer of your accountability. Where a regulator holds you responsible for a decision, you remain responsible for it, including for what your people approve. The product is built to make that approval real, recorded and defensible rather than to remove it.
• No promised business outcome. The diagnosis exists so that you can find out what is actually automatable on your own data before committing to a build. If the numbers are not there, we would rather tell you at the end of two weeks.

Do not use the platform for anything unlawful. Do not put categories of data into a workspace that your agreement does not cover — protected health information, for example, before a business associate agreement exists. Do not attack the platform or attempt to reach another organisation’s data.

And if you find a way to make an agent take an action a person should have approved, tell us rather than using it. That is not a threat; it is an invitation, and the disclosure route is on the security page with what we commit to in return.

The limits of what either side is liable for are agreed in the contract, proportionate to what you pay and what the work touches. We are not going to print a number on a marketing page: a liability figure that has not been negotiated against a specific scope is decoration, wherever you see it.

This page is dated and reviewed at every release. Terms inside a live engagement change by agreement between us, not by us editing a web page and calling it notice.