Taking monthly share-program reports out of Excel and putting them inside the product

Every month, finance teams rebuilt the same spreadsheet by hand

Every month, finance teams at our client companies had to file a report on their share programs. The numbers lived in the product. The reporting work didn't. People would export the data, paste it into Excel, apply the rules by hand, and send the file out. One missed value held up the whole month. Customer Success was on call to rebuild the spreadsheet every time something changed.

“If a screen is asking a finance team to do mechanical work the system could do, the design isn't finished.”

Four decisions that moved the work out of Excel

Decision

A guided report flow built into the product

Finance picks the period and the system fills the rest — values, rules, breakdowns. The same surface handles every report type the team has to file.

Built unprompted

Reusable grouping — set it up once, every future report inherits it

Each client groups their report differently. The old workflow handled this in Excel. I designed a grouping pattern that lives in the product, so the same setup carries across every month. Removed about 1 000 hours/year of manual Excel work.

Strategic

Currency rules — configurable inside the product, not patched in Excel

Custom currency rates and country-specific overrides used to be fixed by hand. I designed a permission-aware setup with validation and safe defaults, so finance can change a rate without opening a spreadsheet.

Problem reframed

Fair value valuation — guided in the product, not outsourced to an external expert

Working out what shares are worth on paper used to mean hiring an outside expert. I designed a guided flow with inline explanations, visible formulas, and history/comparison views — so company admins can run it themselves and trust the result.

AI didn't replace the finance experts. It shortened the distance to them.

• Prototyping in v0. I used v0 to sketch a new currency-rate setup directly in code, not static frames — and walked into the engineering review with a working flow. See the v0 conversation ↗
• First-pass tooltip and validation copy. AI drafted the wording in finance-correct language. A finance expert and I edited every line before it shipped.
• Stress-testing my own thinking. Before scheduling the real review, I'd ask AI what an auditor would object to in a flow. Two rounds out of three were noise. The third caught a real hole.

What I rejected — AI-drawn UI explorations. In a domain where a misnamed field can break a report, the cost of a wrong label was bigger than the time AI saved drawing it.

I shipped cards where the audience needed a table

The reports list could have used a third-party table — sortable columns, dense rows, the format finance admins live in. I replaced it with cards instead. Cards matched the rest of the product, and consistency felt like the responsible choice.

I was optimising for the wrong thing. The audience here is admin super-users who work in tables all day. They needed capability over consistency — sort columns, scan rows, act. The right call would have been to keep the table, accept the design-system deviation, and document the exception. That's the decision I'd make now.

Research that fed every decision