Capital Planner
Ask capital-appropriation questions in plain language — remaining budget, spend, ownership — instead of hunting through a spreadsheet.
The story, in brief.
Finance team members needed answers to capital-appropriation questions — remaining budget, total spend, ownership — without opening a spreadsheet and scrolling. Capital Planner, built by Tim Maka, answers them in plain language, and is now being re-architected onto Dataverse for reliability at scale.
How it’s built.
The production build: a natural-language query agent grounded in a single SharePoint-hosted Excel file of Capital Appropriation Request (CAR) data, maintained by Tim Maka. Finance team members ask questions like “How much is left on CAR123456?” or “How many CARs does Michael Pietrangelo own?” and get an answer without a manual spreadsheet lookup.
Steps through 2 generations — from PROD to DEV POC.
What AI actually does here.
Spreadsheet → conversation
Turns the everyday task of finding a CAR balance, spend, or owner into a plain-language question — the friction removed for a whole Finance team.
Grounded in the source of truth
Answers come from the authoritative CAR file maintained by Finance, so the agent’s reply matches the system of record rather than a stale copy.
A better data layer via MCP
The DEV POC moves the data into structured Dataverse tables queried through the Dataverse MCP Server — a more reliable, scalable foundation, proven with accuracy and performance checks before promotion.