Delivery
Nobody writes a specification. Someone records themselves working.
The hardest part of automating a task is describing it accurately. So we do not ask you to. A member of your team does the job once, at normal speed, on video — and that recording becomes the specification, the test and the training material.
Why it works this way
Two ways of finding the same thing on screen
Automation usually breaks for one of two opposite reasons: it is too rigid and any change breaks it, or it is too loose and it clicks the wrong thing. We run both approaches together, because they fail in opposite directions.
The visual layer
Comes from the recording. Anyone can produce it — no technical knowledge needed. It tolerates a screen being restyled, and says "I think that is the button."
Tolerant
The structural layer
Comes from inspecting the page. Exact or nothing. It breaks when anything changes, and says "that is the button."
Precise
The visual layer proposes. The structural layer confirms. If they agree, the bot acts. If they disagree, it pauses, captures the screen and asks a person. If both fail, it halts. It never guesses its way through.
The platform
Every engagement runs on the same instrumented system
We do not hand you a bot and leave. Everything we run sits inside a delivery platform that scopes access by client, records every action, and monitors whether the work is still happening.
Scoped by client
Each client's data is separated at the platform level. Our own staff crossing into a client's records is logged and shown, not silently permitted.
Three separate audit trails
System events, configuration changes and clinical records are held apart. Different retention, different access, different legal weight.
Missed-work alerting
The platform knows what should have run and when. A sweep that stops happening raises an alert — absence is treated as a fault.
Change control
Configuration is edited, approved and deployed by three separate permissions. The person who writes it is not the person who ships it.
Clinical safety
A hazard log is produced for every deployment and signed by your Clinical Safety Officer before anything goes live.
Hosted by us
Runs on our infrastructure. Your IT team installs nothing, patches nothing and maintains nothing.
What we commit to
The promises our own team is authorised to make
We keep an internal list of what may and may not be promised to a client. Everything on this website comes from the first list.
| We commit to | What that means in practice |
|---|---|
| Five to six weeks to user acceptance | From signed engagement letter to a working bot in acceptance testing, for a workflow that fits a pattern we have built before. |
| No coding required from your staff | The bot is taught by recording someone doing the task. Your team needs no technical skill to produce the source material. |
| Nothing installed on your infrastructure | Hosted and maintained by us. |
| Full logging of every action | Timestamp, item, route taken, outcome and retry count — for everything the bot does. |
| The bot watches itself | An unexpected drop in throughput raises an alert rather than waiting to be noticed. |
| Your Clinical Safety Officer signs first | Hazard log completed and signed off before go live, every time. |
| Acceptance criteria set at the start | Agreed by you in week zero, tested against in week six. |
What we do not promise. Timelines shorter than five weeks. Volume or speed comparisons against your current manual process before we have seen it. Integration into a specific supplier's internal interfaces before we have confirmed it is possible. Anything on the roadmap. If we cannot commit to it, we say so in the first meeting.
Afterwards
Each engagement makes the next one cheaper
Document shapes, letterheads and validation rules learned on one engagement carry over to the next. A trust joining later starts with the benefit of everything built before it — which is a structural advantage a day-rate supplier cannot offer, because sharing the learning would reduce their own billing.
See it running before you commit to anything
We would rather show you the system refusing to do something than talk about how well it works.