Expense Categorization, Automated — No New Software

Snapshot

• Client: Founder-led coaching business, lean finance function
• Problem: Expense categorization was manual and month-end was a scramble
• Solution: Vendor-to-category memory plus automatic monthly rollups
• Stack: Airtable, Zapier
• Build time: 4 days

The Problem

This is the least glamorous case study we've published and it might be the one most businesses need. It's also the one where expense categorization automation didn't require buying a platform — the two fixes below run on a table that already existed.

Expenses were being tracked. Sort of. Transactions landed in a table with a vendor, an amount, and a date, and then somebody had to sit down and decide what each one was. Software or contractor? Marketing or tooling? Is this the same "Google" as last month's "Google"?

Because it was manual, it got deferred. Because it got deferred, it piled up. And because it piled up, month-end became an archaeology dig — someone reconstructing three weeks of context from memory and bank statements, usually under time pressure, usually getting some of it wrong.

Two symptoms told us this was worth fixing:

• The current month's numbers were never available during the current month. The founder was steering with a rear-view mirror that lagged by weeks.
• The same vendor got categorized differently depending on who did it and how tired they were, which made year-on-year comparison meaningless.

Expense Categorization Automation Without a New Platform

No new tools. No accounting migration. Two small automations on top of the table that already existed. That's the whole pitch, and it's also the honest reason platforms like Navan's AI-GL-coding approach or Ramp's card-plus-tracking bundle aren't the right fit for every business — they solve this by selling you infrastructure you may not need yet.

One: the system remembers how you categorized a vendor and does it for you next time.

Two: every expense links itself to its monthly summary record on the way in, so the rollup is never stale.

That's it. Neither is clever. Together they change the entire rhythm of the finance month, because the work stops arriving in one lump at the end.

The Build: Vendor Dictionary, Flag-or-Guess, Auto-Link

A vendor dictionary that learns

The first time a vendor appears, a human assigns the category. That decision is stored. Every subsequent transaction from that vendor files itself — same category, same tax treatment, no thinking required. The dictionary is a small table that quietly becomes the most valuable asset in the finance stack.

Flag, don't guess

When a vendor isn't recognized, the automation does not pick the closest-looking category. It leaves the field empty and flags the record for review. An expense marked "needs a decision" is honest. An expense auto-filed into the wrong bucket is a silent error that surfaces at tax time.

Automatic month linking

Each expense resolves its month from the transaction date and links to the matching monthly summary record, creating that record if it doesn't exist yet. No manual grouping, no pivot rebuild, no formula that breaks when the year rolls over.

A rollup that's always live

Because linking happens at intake, category totals and month-on-month comparisons are current at all times. Opening the base on the 12th shows you where the month actually stands.

What Month-End Looks Like Now

Month-end got boring

The close turned into a review of flagged exceptions rather than a categorization marathon. Most months, the books were effectively closed the day the month ended.

Categorization became consistent

The same vendor gets the same treatment every time regardless of who's looking at it, which is what makes period-over-period comparison mean anything.

The founder got current numbers

Spend by category, this month, right now. That's a different decision-making position than finding out six weeks later.

The system got smarter without maintenance

Every new vendor teaches it once. The share of transactions needing human input drops on its own, and nobody has to maintain rules.

Why This Matters

Founders obsess over revenue systems and tolerate finance ops that would embarrass a bookkeeper. But margin is made on both sides of the line, and you cannot manage spend you can only see in arrears.

The wider lesson is about scale of solution. This didn't need a platform migration or a finance hire. It needed two automations sitting on a table that was already there.

Worth asking:

• Can you see this month's spend by category today, without asking anyone?
• How much of your month-end is categorizing things versus reviewing them?
• Does your system tell you when it doesn't know something, or does it guess?