How it works
Three steps. You are in the room for all of them.
You bring the problem
Describe it in the form below, in your own words. Anonymise it if you want: change the names, round the numbers, leave out the client. We reply within a working day and set up a call.
We make it with Sheaf
We build an agentic solution for your problem on OpenShell with Sheaf signalling: the agents that do the work, the rules they work under, and the points where a person has to say yes before anything happens.
We break it to prove it
We feed it conflicting facts, bad data and instructions that pull against each other. Every time the solution is about to act on reasoning that does not hold together, Sheaf stops it and asks a person. You watch every attempt, and you see what held.
Bring us something
The harder it is, the better the test. We are looking for problems that are:
- complex
- messy
- high-risk
- full of conflicting information
- dependent on human judgement
What you walk away with
Not a slide deck. A working thing, and the record of how it was tested.
- A working solutionBuilt for your problem, on OpenShell with Sheaf signalling, ready for deployment.
- The break recordEvery way we tried to break it, what it stopped, what it asked a person about, and what held.
- Something to show your boardA demonstration on your own problem, not a general one, of agents doing real work safely.
- A clear next stepWhat it would take to run this for real in your business, and what it would cost.
Thank you. We have your problem.
You will hear from us within a working day, from hello@sheaf.one, to set up the first call.
If it is urgent, write to that address directly.
Prefer email? Write to hello@sheaf.one with the problem in a few lines.