How the pilot runs
A pilot is a time-boxed confirmation of a solution’s value on a single, selected process. The process is identified by the AI readiness audit, and the reference point is the baseline measured during that audit.
The pilot runs on real data, in the real process, with the people who do this work every day. From day one, it follows the same governance rules as a production solution.
The decision on production deployment rests on the measured result.
Duration
Six to eight weeks allows the solution to be tested on a real volume of work and the result to be measured reliably. At the same time, it gives the organisation a prompt decision on the way forward.
Measurable result
We relate every result to a specific unit of analysis: a process, a role or a team. We measure it before and after the pilot, using metrics agreed before the start:
- process cycle time,
- effort per unit of output,
- quality, measured by the number of errors and escalations,
- throughput with the same headcount,
- time to decision,
- unit cost per operation.
We set the result against the cost of development, model operation and maintenance. Solutions that pay back within the agreed time horizon move into production.
Production deployment
Production deployment is the full delivery of a solution that has proven its value in the pilot. It includes documentation, user training and adding the solution to the organisation’s service catalogue, with a service definition, support level, incident reporting path and change management process.
A production solution operates as a digital product: it has an owner, a change backlog, versioning, usage metrics and a planned end of life. We describe the principles behind this on the AI governance and oversight page.
Handover of ownership
Every deployed solution has an owner within the organisation. Handover of ownership means that this person understands how the solution works, the data it uses, how its result is measured and the procedure to follow in the event of a failure.
We create documentation and operating instructions during deployment, so that the organisation’s team can take over the solution in full. We design for maintenance from day one.
A consistent way of working
The implementation stage leads to an organisation that works with artificial intelligence as a coherent whole. Individual solutions form a consistent way of working, and each subsequent deployment builds on the same foundation of governance, data and architecture, which makes it run more smoothly.