One scheduler. Two systems.
One fix that became a revenue line.

Challenge

Every event, typed twice.

At a major convention centre, a scheduler manually built each event roster in the payroll system. They re-entered it into a separate event system. The work was done diligently each week by someone who knew the job well. It was still slow, and the two systems inevitably drifted apart. When they disagreed, shifts went uncovered, or someone got paid for a shift they never worked.

Leadership knew AI and data could help — better decisions, less manual work, easier access to information. But they had real questions before they'd move: how do we govern this data internally? Which of it can a model actually see? And out of a dozen good ideas, which one do we do first?

Solution

Start where the pain already is.

Khashmere's forward-deployed engineers sat with IT, governance and privacy, and the business analysts who felt the pain first-hand. Together, they turned a list of ten AI & data priorities into a sequence — what to fix first, what could wait.

We configured Microsoft Fabric, stood up the groups that would own data governance going forward, and migrated the convention centre's Power BI reports onto the new foundation. Once the data had a single home, we built semantic models on top of it and connected those models to Copilot — so a question that used to mean a report request now gets answered in the flow of work.

Last, we closed the loop on the problem that started it all: an automated interface between the payroll and event systems, so a schedule entered once is correct everywhere. No more typing it twice.