The Very Big Spreadsheet

The Compliance team kept a list of business risks for the US Oncology Business Unit. The list lived in a spreadsheet. Updates came in by email. Every quarter, someone built the leadership slides by hand.

It was hard to see who owned each risk. It was hard to see what had changed. And when auditors asked for proof, everyone went digging.

I didn’t start by drawing screens. I started by learning how the work really happened. In my first weeks I took 58 screenshots and drew 17 flow diagrams.

I also made copies of the spreadsheets filled with made-up data. That way the team could design with real fields and no real records.

On 22 September 2025 I wrote a design document. It listed six goals and four groups of people. For each group I wrote what they needed, what hurt, and what success would look like.

With the goals written down, I used Figma Make, an AI design tool, to build clickable prototypes. I gave it my goals and my four groups of people. I didn’t take its first try. I asked it to grade its own work, told it the pieces needed serious rework, and had it rebuild them on a shared design system.

I also helped design a team idea, an AI assistant that could answer questions about risks. But on 16 October 2025 we made a call. Getting data in comes first. The chatbot could wait. A risk register nobody can fill in doesn’t need a chatbot yet.

A prototype shows an idea. A builder needs rules. On 14 November 2025 I wrote version 1.0 of the requirements document. I was the only author, and I kept it up to date for five months.

It said who can do what, what each screen must do, and what an auditor would check. Two of my favorite rules:

Before ECS started building, I made one reference pack. It put four things side by side: the old spreadsheets, the Power BI demo, my Figma screens, and ECS’s first wireframes.

Now we could point at real buttons instead of guessing. The business owner took a red pen to ECS’s wireframe. Click Q1, she wrote, and show that quarter’s progress, without going into edit mode. ECS confirmed the change, and she approved the wireframes with one fix.

In January the business asked for 5 screen changes. Before any of them reached the build, I traced each one back to the rule book or to the person who asked. None of them changed how the data was stored, so the build date could hold.

Then came testing. I planned when it starts, when it’s done, and how each role gets tested. The business owner found the problems and ECS fixed them. On 19 February I sent everyone 8 clear to-dos.

We also moved one rule, the one that flags late work, to Phase 2. A steady first release mattered more. The app still launched later than the first plan, because we kept adding requirements.

ECS built the risk register in Power Apps, and it went live in March 2026. That was about four and a half months after my first draft. To be clear, I wrote the rules. ECS built the app.

Then real use taught us something. The business owner needed each quarter’s progress in her committee report, and the app couldn’t include it. I sketched a way to pick which fields go in the report, and wrote year and quarter filters into the rules. She said picking the fields was exactly what she needed.

For Phase 2 I wrote a plain-language deck. The rule book was long, and the team was spread across time zones. So I used short words and one idea per slide. If we agreed on it there, nobody could say later they didn’t know.

On 21 April my deck said a task was late if nobody touched it for a set number of days. Two days later, the business owner stopped me. The work might be done, she said. Someone just hadn’t written it down yet.

She was right. Calling finished work “late” would leave a bad mark in the audit history. So I rewrote the rule around missing updates instead.

In early May I walked a teammate through the Phase 2 rule book and started handing the register over. Then I moved on to the next big job: an AI Workbench prototype, one shared app for five workflows and four teams. That’s a different book.

Here is the honest page. There is no record of how many people used the app, or how much time it saved. We pushed the dashboard for tracking results by role to Phase 3. I asked ECS for a weekly view of requested, active and done work. By May, ECS still hadn’t started using it.

Next time, I’d set up the numbers on day one. How many owners update on time? How long does the quarterly report take? How many updates go missing? A story is nice. A number is proof.