Month-end reconciliation, 4 days to 60 seconds

What we found

A month-end reconciliation that took the CFO about 4 days by hand now runs in about 60 seconds. The client is Genesco Sports Enterprises. Every month, the finance team rebuilt job-profitability numbers by hand from Sage 50 exports. We built a web app with Claude Code, Anthropic's coding agent, that does the mechanical part of that close and hands the rest back to the finance team. It runs live at Genesco.

This is the record of that build. The figures here are Genesco's and they are public with their permission. They cover one company and one process, and your close will have its own shape. The arithmetic and the carry-forwards moved into software. The figures that depend on context stayed with the finance team.

The work by hand

Most of the old close was moving numbers between files. Files came out of Sage, and someone imported them, reconciled the totals across several spreadsheets, then re-keyed prior-month profit figures so the new month started from the right place. It was slow, and it was easy to get wrong.

Importing, reconciling and re-keying are mechanical work. The part that needs a finance brain came at the end, so the reviewer waited days to reach it and made the calls at the end of a long week. Those steps filled the close and added nothing to the decisions a finance team is paid to make.

The build

We built a web app that ingests the Sage exports, computes the job totals and pre-fills the figures the rules allow.

The app does the arithmetic on the exports. It computes the job totals and fills in every figure that follows a written rule, which here means the sums and the carry-forwards a person used to do with a keyboard.

The app then hands back a clean workbook for the judgment calls, so the finance team works in a familiar place on the short list of items that need a decision.

The reviewed workbook goes back in, and the app produces the final reconciled master. Someone uploads the exports, reviews the workbook, and downloads the master. Each run leaves a record of what went in and what came out, so the close can be audited afterwards.

It remembers each month's numbers so the next month starts pre-filled. Last month's agreed figures are already sitting in the new month, and nobody carries them across by hand.

How we built it with Claude

We built the app with Claude Code, Anthropic's coding agent. Claude Code wrote the code from our written description of Genesco's close: how the Sage exports arrive, which figures follow a rule, and which ones the finance team decides. The first working version was built in one day. Every change since then, including the adjustments the CFO asked for, also went through Claude Code, and we checked each one before it shipped.

The live app does its math with fixed rules and does not call a model. That was a choice. A reconciliation has to give the same answer every time it runs on the same files, and written rules do that. Claude built the software, and the software runs the close.

The judgment calls

The judgment calls stay with the finance team. We drew the line where the client drew it. A figure that follows a written rule belongs to the app, and anything that depends on context, on a conversation, or on a call about how a job should be treated stays with the team.

The app presents the cases that need a decision with the arithmetic already done. A reviewer who disagrees changes the number in the workbook, and the next pass takes that change as given.

The line follows a plain test. If you can write down every step and every criterion for a good outcome, software can own it. If a person regularly has to weigh something that was never written down, the person keeps it.

The steps we removed

Several steps from the old close no longer happen at all. We asked what each step was for, and much of it existed because a person was moving the numbers.

Re-keying prior-month figures existed because those numbers lived in last month's file and someone had to carry them over, and the app's memory between months ended that step. Reconciling totals across several spreadsheets existed because the exports arrived as separate files, and the app merges the exports now. Importing files existed because the export format did not match the workbook.

The review remains, and the finance team reaches it at the start of the close.

What changed

The 4 days became about 60 seconds, and the shape of the close changed with it. The work is now a structured upload, review and download flow, and it is auditable. The build shipped and runs live at Genesco, and we refined it on the CFO's feedback.

Time is the easiest part to measure. Other things changed as well. A whole category of error stopped, because transcription mistakes between files cannot happen when nobody types the figures across. The close no longer sits behind a reviewer's block of days, so it can be run again when a late number arrives.

Mark Tatum, Genesco's CFO, described it this way:

"Every month our team burned hours rebuilding job-profitability numbers from Sage by hand. Plainpath built us a tool that does the heavy lifting and leaves us just the judgment calls. It has saved us real time and headaches, and they were quick to adjust it when we asked."

Limits

These numbers belong to one company and one process. Genesco's close had clear inputs, clear outputs, a large amount of manual effort, and a rule set that could be written down. That is what made it a good first process. A close where judgment enters at every step compresses far less, and we say so before anyone quotes these figures at their own board.

The build is also unfinished. Someone on the client's side has to own it and keep asking for changes. At Genesco that person is the CFO, who used it, told us what to adjust, and got the adjustments. We wrote about builds nobody owns in why most AI pilots fail.

Where to start on your own close

Write down how the work runs today, in the order it actually happens, with who does each step and how long it takes. Mark the steps that exist because a person is doing them, and mark the point where someone makes a decision that no rule describes. Our guide to doing that is how to run your first AI audit in an afternoon. If you want a person to check that map and write a build plan against it, that is what we do in Discovery.

Frequently asked questions

What did Plainpath build for Genesco Sports Enterprises?

We built a web app with Claude Code that ingests the Sage exports, computes the job totals and pre-fills the figures the rules allow. It hands back a clean workbook for the judgment calls, then produces the final reconciled master. It remembers each month's numbers so the next month starts pre-filled.

How long does the month-end reconciliation take now?

A month-end reconciliation that took the CFO about 4 days by hand now runs in about 60 seconds. The work is a structured upload, review and download flow, and each run leaves an auditable record.

Does the software make the accounting decisions?

No. The judgment calls stay with the finance team. The app does the arithmetic and the carry-forwards that follow a written rule, then presents the items that need a decision. A reviewer who disagrees changes the figure, and the next pass takes that change as given.

Will our month-end close get the same result?

We cannot promise that. These figures come from one company and one process. Genesco's close had clear inputs and outputs, heavy manual effort, and rules that could be written down. A close where judgment enters at every step will not compress the same way.

What makes a process a good first build?

Clear inputs and outputs, a lot of manual effort, and a real gain when the work runs faster. It also helps if the process is not mission-critical, so there is room to learn safely on the first one.