AI pilots and proof of value
An AI pilot is a controlled deployment in a live building with real users, measured against criteria agreed in advance, used to decide whether a technology is worth scaling across a portfolio.
Does this work in our building, with our people?
What we do
01 / 05A demo proves the product works. A pilot proves it works here.
We run controlled pilots in live environments with real users, measured against criteria agreed before starting. Criteria set afterwards are not measurement, they are marketing.
Nothing scales until it has survived contact with the actual estate. The technology is rarely what fails. Adoption, integration and operational fit are what fail, and none of those show up in a vendor demonstration.
Sometimes the honest answer is stop, and we will say so. A pilot that ends in a clear no has done its job and saved a portfolio rollout.
- Pilot design and success criteria, agreed before deployment
- Live deployment in a real environment, with real users
- Measured results against the criteria, not against the pitch
- A scale or stop recommendation, stated plainly
Workplace, IT and innovation leaders. Operators looking for evidence before a portfolio commitment.
How it works
02 / 05- Weeks 1 to 3
Design and criteria
What we are testing, where, with whom, and what result would justify scaling. Signed off before anything is installed.
- Weeks 3 to 5
Deploy
Into a live environment. Real floor, real people, real operational conditions, integrated into the systems it will have to live with.
- Weeks 5 to 17
Run and measure
Long enough to pass through a full operational cycle and past the novelty period. Measured continuously, not surveyed at the end.
- Weeks 17 to 19
Scale or stop
The evidence, the recommendation, and what a portfolio rollout would cost and require. Including the case for stopping.
Durations are indicative and depend on the size of the estate and how quickly we get access.
Questions we get asked
03 / 05How long should a pilot run?
Long enough to get past the novelty period, which usually means at least eight to twelve weeks of live running. Anything shorter measures curiosity. Behaviour change and operational friction only appear once the technology has stopped being new.
Who pays for the technology during a pilot?
It varies, and it is worth negotiating. Many vendors will fund or discount a pilot in a reference environment. We help you structure that without giving away the commercial position for the rollout that follows.
What happens if the pilot fails?
You have bought a fast, cheap answer instead of a slow, expensive one. We report it plainly and say why. A pilot that fails for a fixable reason gets a route forward. A pilot that fails because the premise was wrong gets a stop.
Can we pilot on a live floor, or do we need a test space?
A live floor, wherever it is safe to do so. Innovation labs and test suites produce results that do not survive the move to a working environment, because the people in them have volunteered and the conditions are not real.
How do you measure a pilot?
Against the criteria agreed at the start, using instrumentation defined at the same time. Typically a mix of system data, observed behaviour and structured user feedback. The measurement plan is a deliverable before deployment, not an afterthought.
Ready to press start?
05 / 05Tell us about the space, the asset, or the problem.
We will tell you honestly whether we are the right people for it.