These are tools built for real month-end, reporting and planning problems, not demos. Each one replaced a manual step, a fragile spreadsheet or an unchecked number with something that runs every month and can be traced back to its source.
Power Platform · Power Query · Power BI · Copilot Studio · Claude · Python · HTML/JS · Excel
Pick a stage of the finance cycle to filter the register. Open any entry for the problem, what was built and the result. Company names, people and figures are generalised or left out.
The build was the easy part. The harder work was making AI reliable, getting people to use it, and keeping it under control.
AI gives slightly different answers each time; month-end cannot. Every check that decides something runs as a fixed rule kept in its own place, and the same rule gives the same result every month. AI reads the documents, applies nothing on its own judgement, and explains what the rules flagged.
The roll-forward that AP can actually reconcile, the formatting that matches the rest of the workbook, the app a cardholder can finish on a phone. Adoption came from shaping each tool around the person who uses it, then fixing what they found in the first live month.
Business processes move through fixed states. AI can help at each step, but it never moves money, posts a journal or changes a supplier's bank details. Every number traces back to its source and its approver.