How Quartz built an AI-native operation for programme knowledge

Quartz Project Services is a quantity-surveying consultancy in the UK. Everything about their live programmes already lived in SharePoint: years of meeting minutes, drawings, cost reports and emails, somewhere around 3.5 terabytes of it. Having the information was never the issue. The slow part was getting to the right piece of it when a client asked a question or a monthly report was due.

Putting one report together could take a senior consultant about four hours: opening folder after folder, copying the right figures across, and checking they were working from the latest version each time. None of it was difficult work. It was just slow, and it kept landing on the most experienced people on the team, the ones whose time was worth the most.

We didn't ask Quartz to move their files or learn a new system. We built Leo, an assistant that sits on top of the SharePoint they already have, and let people ask it questions in Microsoft Teams using ordinary language, in the same place they already talk to each other during the day.

When someone asks Leo something, it gives the answer and shows exactly where it came from: the document, the page, and the section. A surveyor can ask where a particular planning condition sits on a permission and get the wording quoted back with a link to the source file. Because every answer points to a real document, people stopped digging through folders just to confirm something they already thought was true.

The monthly report now comes back as a Word document that already has the figures and references in it, ready to be checked and edited rather than built from scratch. The sources come attached to the draft, so reviewing it is genuinely a review, not a second search to prove where each number came from.

That one change took report preparation from around four hours down to about twenty minutes, which is roughly 80% of the time given back. The consultant still reads it, makes their adjustments and signs it off; they simply don't spend the first half of the morning assembling it. The time that used to go into finding files now goes into the programme itself.