Almost every conversation about records in this sector turns to AI. We put reporting first, because reporting is the part that is already late, already manual, and already being done twice.

The pattern is familiar. Someone pulls numbers out of one system, retypes them into a funder's spreadsheet, chases three colleagues for the parts they hold, and reconciles the total by hand the night before it is due. The information existed the whole time. It just was not shaped the way the report needed it.

The deployment ships with 28 report templates across eight categories. They are built to the Data Collection Instrument structure, so the shape of the output already matches what funders and agencies ask for, rather than being an internal format that has to be translated at the end.

These templates are in use today, not on a roadmap. That matters, because reporting is where the case for a records system gets made or lost. If the first quarter after deployment produces a report in less time, with less retyping and a source you can point at, the rest of the work becomes much easier to justify.

AI comes after this, not instead of it. A model asked to summarise a badly organised drive will produce a confident summary of a mess. Structure first, reporting second, inference last. That order is deliberate, and we would rather argue about it now than apologise for it later.