AI that works and demonstrably holds up. That is, in one sentence, what Evident stands for.
Anyone can switch on a chatbot. Trusting it takes more: agreements on which data may go in, who may do what and what the AI must not do. Without those agreements it works until it goes wrong.
That is why Evident works in four steps, and after every step you have something in hand.
Understand
First the work is mapped: where the hours go and which data it touches, and it is agreed which data may go in. You get a process snapshot with the hours per week: the baseline.
Set up
Then AI is set up on one or two processes, with the tools you already have, together with the people who do the work. Along the way it is decided who may do what and what the AI must not do. You get a working setup with a short manual, a one page agreement framework and the access model.
Embed
Next your team learns to work with it and those choices are embedded in the organisation. You get work instructions and an owner per process.
Hand over
Finally the hours are measured again against the baseline, and your team carries on by itself. You get a handover file with the measurement.
"Works" is about the hours that come back. "Demonstrably holds up" is about the agreements, and about the measurement that shows what it delivered. Evident does not promise a percentage of time saved up front; the measurement shows what it is.
Together this is the Evident AI Kickstart. The proposal sets four to six weeks for it, for one or two processes. What is not included is stated as well: licences, IT management and legal advice.