Testing before investment is an approach in which a company tries out technology on a limited scale before committing to full deployment. In artificial intelligence, this is especially important, because the benefit strongly depends on the quality of a company's own data and on the specific process, so a solution that works elsewhere may not work for you. A usable pilot has three characteristics: a success criterion defined in advance and expressed as a number, a scope limited to one process or team, and a clear deadline after which a decision is made. Without a criterion, the pilot turns into endless trialling that nobody brings to an end. Measure time saved, the error rate and user satisfaction against the original state, which you should record before launch. Some pilots can be carried out through support programmes offering preferential access to technology.
See also: European Digital Innovation Hub (EDIH), Feasibility study, Register of AI systems.
