What a Proper Copilot Pilot Looks Like

Last one in this series, and the most practical one.

Four days of licensing costs, CMMS gaps, what agents can do, and where your data goes. So now we arrive at the most important question. "OK, so what does actually doing something about this look like in our organisation?"

Here is what I would do. I would not start with a platform conversation. I would start by asking you to walk me through two or three processes that quietly eat time every week. The ones people have just accepted as part of the job.

A classic example is timesheet extraction for payroll. Someone exports data from a field management system, pastes it into a spreadsheet, cross-references it against scheduled hours, flags discrepancies, emails a supervisor, waits for a response, then manually updates before submitting to payroll, where it then gets entered into a payroll system, either manually or with an import that still requires a lot of manual intervention. That whole sequence, done weekly or fortnightly, across multiple sites, is a candidate for significant automation. Not all of it, but most of it.

Another one I see regularly is bank reconciliation prep. Pulling transactions, matching against purchase orders, chasing down unmatched items across email threads. Hours of work that follows the same logic every single time.

Those are the workflows I would scope a pilot around. Six weeks, 5 to 12 users who actually touch those processes, and a clear measurement of time spent before we start versus after we finish.

The licence cost for a pilot cohort of that size is modest and the technology outlay is not the risk. The risk is building something that does not reflect how the work actually happens on the ground, which is why the design stage matters as much as the build.

What I would want to hand you at the end of six weeks is a clear number — hours saved per week per workflow — a view on what broke or needed adjusting, and a recommendation on whether to expand.