Claude for Finance: How SMEs Can Finally See Where Their Margins Are Leaking

Why “we’re profitable” isn’t a number

In 30 years of building customer and data strategy for banks and agencies, I learned one thing that never changes with the tools: the businesses in trouble are rarely the ones losing money on paper. They’re the ones who can’t tell you, channel by channel, product by product, where the money is actually being made or lost.

AI doesn’t fix that on its own. A chatbot that writes your invoice reminders is convenient. It is not the same as knowing your true cost per acquisition, or which service line is quietly subsidising another. That’s not a tools problem — it’s a clarity problem. And clarity is exactly where AI earns its keep, if you point it at the right questions.

This is Economics Engine work: fixing unclear unit economics, high hidden costs, and squeezed margins — one of the 10 AI Profit Engines behind the AI Customer Profit Engine system.

The SEE → QUESTION → DECIDE framework

Three stages, in order. Skipping a stage is why most “AI for finance” advice doesn’t stick.

• SEE — get your real numbers into one place Claude can read
• QUESTION — use Claude to interrogate the numbers, not just summarise them
• DECIDE — turn one answer into one action, this week

Step 1 — SEE: get your real numbers in one place

You don’t need new software. You need your existing numbers somewhere Claude can actually work with them: a simple spreadsheet export of revenue and cost by channel, by product line, or by client — whatever split matters most in your business.

The minimum viable version: month, channel/product, revenue, direct cost, and any marketing spend attached to it. That’s it. Most SME accounting exports already have this; you’re just pulling it out of the accounting software and into one flat sheet.

Step 2 — QUESTION: let Claude find the leak

This is where most people undersell what Claude can do — they ask it to “summarise the spreadsheet” and get a summary they already knew. Ask sharper questions instead:

“Here is my revenue and cost by channel for the last 6 months [paste or attach]. Calculate gross margin per channel. Which channel has the lowest margin, and has that margin been shrinking or growing month over month? Show your working.”

“Based on this data, if I had to cut spend on one channel to protect cash flow, which one, and what would I actually lose — leads, revenue, or just noise?”

“Compare my cost per lead against my cost per acquisition for each channel. Where is the gap between the two largest, and what does that usually mean?”

The difference between a summary and a diagnosis is that a diagnosis tells you what to stop doing.

Step 3 — DECIDE: turn the answer into one action

Clarity that doesn’t lead to a decision is just an interesting chart. Once Claude has surfaced the leak, pick exactly one action for this month: pause a channel, renegotiate a cost, change a price, or reallocate a marketing budget. Not five actions — one. Measure it. Then run the same three steps again next month.

A before-and-after example

Before: “We’re growing — revenue is up 18% this quarter.” Spend continues across all six channels roughly evenly, because nobody has broken out margin by channel.

After running SEE → QUESTION → DECIDE: Revenue is still up 18%, but one channel is responsible for 40% of spend and under 5% of net margin after real costs are counted. That budget gets reallocated to the two channels actually driving profit. Same revenue, materially better cash position — without spending a ringgit more on marketing.

FAQ

Your next step

Jane Chew | AI Strategy Coach, Founder of DigitalAI Business Club | DigitalAIBusinessClub.com