Scope a pilot
A pilot is how most of our customers start: one use case, your data, your models, running inside your perimeter, with the exit criteria written down before anything is installed. Six weeks is the default; the scope is fixed and priced up front after a short briefing — €25,000 to €90,000 depending on how much has to be true at once, and credited against your first year of platform — so procurement gets a single line item and you get a date.
What a pilot covers
- A couple of your data sources ingested. Real sources land in the lakehouse on open table formats, inside your perimeter — not a sample extract on someone's laptop.
- A couple of your processes mapped, and a first ontology. Before anything is automated we describe two or three of the processes in scope and build the ontology that carries them — the model everything else in the pilot reads from, and the thing you keep whatever the verdict.
- One use case, chosen for evidence. An agentic workflow or an analytics question that matters to a sponsor and can be judged in six weeks — not a platform tour.
- Your deployment model. Air-gapped, on-premises, sovereign cloud or confidential compute — the pilot runs where production would run, so nothing has to be re-proven later.
- Your models and your data. Open-weight models you already trust, or your existing provider, connected to the sources the use case actually needs — governed, logged, and reversible.
- A written verdict. Week six ends with a document: what worked, what did not, the measured result against the criteria, and what production would take.
How it runs
- Briefing (30 minutes). We agree the use case, the deployment model and the exit criteria. If a pilot is the wrong size, we say so — an Organisational AI Day is often the better first step.
- Scope and price (one week). A one-page scope, a fixed price and the people on both sides. Signed, not negotiated at length.
- Six weeks, weekly checkpoints. Our architects work with your team inside your environment. You see progress every week, not at the end.
- Verdict and decision. The written result and a recommendation. Continue to production, adjust, or stop — with the evidence to justify any of the three.
Who this is for
Public bodies, defence and security organisations, and regulated enterprises that cannot send data to a hosted service to find out whether organisational AI works for them. If you can prove it in a vendor's cloud, you do not need us; if you can't, this is how you prove it inside your own walls.
Prefer to write? Email hello [at] scrydon.com and we will get back to you.
Prefer a lighter first step? Organisational AI in 30 minutes, an Organisational AI Day or Lakehouse to Ontology in a Day.