One workflow.
Proven on your own work.
Let AI prove its value on your own work before you commit to it. Somewhere in your week there is one job that eats more than its share. We rebuild it around AI and test it beside the way you do it today. Then your people run it.
Fixed price, agreed before we start. You pay when the test proves it. Rather start with training your team? See the sessions.Value starts with one job.
The next ones build on it.
The pilot stands on its own: one workflow, a working system and a measured result. If you go further, the next job doesn't start from zero. The rules, the setup and the know-how are already in place.
A narrow, fixed-price pilot with a gate you control
One job, found together or brought by you. We rebuild it around AI on your own seats and your own rules, run it alongside the current process, and measure. The pilot ends with a working system and a number.
Start hereWorkflow discovery and prioritization
With one win in hand, we map the rest of the work: which jobs have the same shape, which pay back first, which are better left alone. You get a short, ordered list, not a transformation roadmap.
If you want moreTools rolled out, people trained, habit built
The sessions, run on your real work. Training surfaces the next workflows and discovery gives training its targets, so the two usually run together. By now your team is teaching each other.
When you're readyEach step stands on its own, and stopping after any of them is a fine place to land. You own everything at every stop: the rules, the system, the results.
What the pilot
looks like from inside.
Scope and the test
We pick the workflow, or find it together if you arrive without one, write down how it works today, and agree the measure that will decide whether the pilot passed. Before anything is built.
The test, in writingBuild on your seats
The system is built on the tools your business already pays for, with your rules written into it. Nothing lives on our side.
Your rules file, on your seatsRun it beside the old way
The new process shadows the current one. We compare, find the gaps, tighten the rules. Your most experienced person is the judge.
A side-by-side logHandover
One-on-one time with the people who will run it, the rules handed over in plain language, and a short note on what to try next. We stay on call after handover while the new way beds in; that support is part of the pilot, not an extra.
The handover noteFixed price for a defined scope, agreed before we start.
Half when the test proves it. Half at handover.
The gate is yours.
Support after handover is included.
The test is agreed in week one, in writing, in your terms: minutes per report, errors caught, hours back. If the numbers say the pilot passed, you pay the first half. Not before.
Good first pilots look alike.
They come back every week, they eat hours, and someone wants them fixed. Check yours against the list below. If it reads more like the second list, it still wants AI, just through your people rather than a pilot.
- Happens every week, the same way, and eats hours
- Needs judgment, not just typing, and the judgment can be written down
- Has a before: a time, an error rate, a backlog you can measure
- Runs on documents and data you can share, or can stand in for
- Is owned by someone who wants it fixed
- Happens once, or differently every time
- Has no measure anyone would agree on
- Hinges on a relationship, not a process
- Depends on data nobody is allowed to move
- Is really a people problem, and people can be trained
Start a conversation
Bring the job.
Twenty minutes.
Tell us the job that eats the week, or tell us you're not sure where to start. We'll tell you whether there's a good first pilot in it, what the test would be, and what it would cost.
Two sentences about the job is plenty. So is "I don't know yet."What to expect 20 min · No slides
- How the job works today, and who does it
- An honest read on whether AI fits it, or doesn't
- What the test would be, in your terms
- A fixed price, or a reason to start somewhere else