QASWAMost automation projects start in the wrong place: with the new system. Ours start with the old one. The spreadsheet a team already trusts is the specification — every quirk in it is a business rule somebody once needed. Here is how that played out on Labor Mart, a labor and overtime platform we built for a multi-site senior living operator.
Where it started
Weekly labor and overtime performance for the whole portfolio ran on one 25 MB Excel workbook. It only refreshed when someone opened it. The logic that decided which community and which role a shift belonged to was hardcoded inside Power Query. Anyone with the file could see every community and every employee. And if the file broke, the process stopped.
None of that is unusual. It’s what a spreadsheet looks like after years of a smart person keeping a business running with the tools in front of them.
Is your spreadsheet a candidate?
- It only updates when someone opens it and presses refresh
- The rules live in formulas or queries that one person understands
- Everyone who has the file sees everything in it
- If the file breaks, the report stops — and so does the decision it feeds
- The numbers arrive after the week is over, when it’s too late to change them
What we did, in order
1. Treat the workbook as the specification
We recoded the workbook’s logic penny for penny — the community and role rules included — rather than redesigning it from a whiteboard. If the old file treated a shift a certain way, the new platform does too, until the people who own the rule decide otherwise.
2. Reuse the data you already have
The platform reads the payroll, staffing and occupancy feeds the operator already had. No new integrations, no rip-and-replace. It refreshes itself and is ready by 9:00 AM Central every day, without anyone opening a workbook.
3. Run both side by side until they match
Because the logic was recoded penny for penny, the platform’s figures could be reconciled against the workbook it replaces — and they were, before anyone relied on them. That is the whole point: nobody has to take the new numbers on faith, because they were checked against the old ones first.
4. Fix access while you’re there
The workbook shared everything with everyone. The platform runs entirely inside the operator’s own Azure tenant, with sign-in through their own identity provider, access limited to approved users, managed identities for every data connection and no stored passwords. Every budget or staffing edit is validated and logged.
5. Move the warning earlier, not just the report
A faster report of last week is still last week. Labor Mart flags overtime mid-week, while there are still two scheduling days left to prevent it, and a built-in analyst answers the common questions — who is at risk of overtime this week, why a community’s overtime spiked — from governed queries, showing the query under every answer.
Be precise about what a number means
We report that 56% of overtime hours fell in the groups Labor Mart flags by Wednesday. That is a backtest, measured by replaying past weeks. It describes how early the warning arrives — it is not a measured reduction in overtime, and we don’t present it as one.
If this sounds like your week
Tell us which report takes the longest. In a free 30-minute call we’ll show you a system we’ve already built for a comparable problem and tell you honestly whether yours is worth automating. Read the full Labor Mart case study, or see how we build AI and automation.