Pilot small, decide on data
A safer road into AI for companies that refuse to buy on stories: one fixed-scope, fixed-length pilot. Before work starts we measure the baseline and register the success metrics; at the end, an evaluation report judges the pilot against exactly those metrics. If it misses, the exit terms agreed at the start close it cleanly — a pilot is a question, not a life sentence.
Who it fits, and what gets in the way
Owners and managers weighing AI adoption who need a conclusion their own organization can act on — with numbers attached.
Every vendor promises efficiency; the internal argument has run for months with nothing to decide on
A previous pilot ended in opinion against opinion: no baseline, no metrics, no verdict anyone trusts
The real fear is not pilot cost — it is the pilot that never ends, tuned forever, stopped by nobody
What you get
| Component | What is included | Deliverable |
|---|---|---|
| Scenario selection | One tightly bounded scenario picked from your workflow — only one | A pilot scope sheet |
| Baseline measurement | Current state measured first: volumes, handling time, error rates | A baseline report |
| Pre-registered metrics | Pass lines and judging method written down and frozen before build | A metrics sheet; the verdict uses it |
| Pilot run | Built to scope and run with a full trail; issues summarized weekly | Run log and weekly summaries |
| Evaluation & verdict | Judged against the registered metrics: passed, missed, or data insufficient — reported separately | An evaluation report fit for an internal decision |
| Tool & subscription inventory (optional) | A pass over the AI tools and subscriptions already on the books: what is used, what idles | An inventory list |
How it runs
Pick & measure
About a week: scenario chosen, baseline measured, metrics written
Build & wire in
Built to scope, connected to your current channels and data; internal dry run first
Pilot run
The agreed period, weekly summaries; metrics watched, never redefined
Verdict & next step
Evaluation report at close: expand, adjust, or stop — decided on the data, by you
Boundaries
Not included
- No promised pass: what is promised is the discipline — baseline, registered metrics, honest verdict
- No scope creep mid-pilot: added scenarios become new pilots, separately scoped
- No polished reports: a miss is reported as a miss, with causes classified
What we need from you
- One bounded scenario and one internal counterpart
- Access to the data and people the baseline measurement needs
How engagement and quoting work
Fixed scope, fixed length, fixed price — in the contract before work starts, priced per scenario by a person after discovery. Expansion after the verdict is your call; a miss closes under the agreed exit terms with data and reports handed over in full.
Questions on this page
How is this different from just buying the AI Staff service?
AI Staff is the long-term managed service. This is the proof step before it: smaller scope, fixed length, baseline and evaluation attached. Passing pilots commonly convert.
Can the report convince our leadership?
That is what it is built for: baseline, metrics, results and method, with the judging rules agreed before the pilot ran — not written at the end.
What happens on a miss?
The report classifies it: wrong hypothesis, insufficient data, or execution. Exit terms close the engagement; nothing rolls into a long contract on its own.
How long is a pilot?
Usually a few weeks, by scenario. The length sits in the contract and closing is automatic at term.
How do we start?
Describe the scenario in a few sentences on the contact page, or start from the pilot application page. We help draw the boundary first.
Related
AI Governance & Compliance Readiness
Pair the pilot with governance and an evidence trail
Pilot small, decide on data
Tell us where it hurts first. Scoping and a written quote come before anything starts; looking costs nothing.
Or ask Leo, the assistant at the bottom right, first — get the questions clear before you write to us.
