6 min read
In 2024, 18% of small businesses using AI in one closely watched survey applied it to financial analysis and business planning. In 2026, that figure is 42%. Over the same two years, the share using AI for written marketing content fell from 74% to 65%. Small business AI is moving away from the blog post and toward the balance sheet, and that shift changes what a solo owner should learn next.
Those numbers come from a Goldman Sachs report on 2,473 alumni of its 10,000 Small Businesses program, surveyed between August 10 and 24, 2026, and summarized here. Treat them as directional. These are owners who completed a business education program, so they are not a mirror of every small firm. But the direction is the useful part, and a far larger dataset points the same way.
What the money shift really tells you
Marketing was the natural first use of AI because mistakes were cheap. A weak caption costs you nothing. So owners started there, got comfortable, and are now moving to the work that decides whether the business survives: pricing, cash, margins and planning.
That is the right order. It is also where the risks change character. A weak caption is embarrassing. A wrong cash forecast can mean a missed tax payment.
The Goldman report also shows how shallow most use still is. 86% of surveyed businesses use AI, but only 15% have fully built it into core operations, and 74% of AI users have no formal policy governing its use. Ireland’s Small Firms Association found a similar gap, saying 92% of small businesses use generic AI tools while only 30% are customising the technology. So the opportunity for a solo owner is not more tools. It is one financial routine done properly, every month.
Results so far are encouraging, with a caveat about who is reporting them. Intuit’s 2026 AI Impact Report, based on more than 34,000 owners, found 78% of US businesses say AI improved productivity and 43% say it increased revenue, with only 2% reporting a revenue decline. Intuit sells QuickBooks and these are self-reported answers, so read them as a sign that owners feel the benefit, not as proof of it. The same report also found that the main barriers are privacy, fear of errors and uncertainty about what AI can do, rather than cost, and that adoption is lowest where human judgment matters most.
The three-layer model for AI in your finances
Here is the framework I recommend. Sort every money task into one of three layers, and only move down a layer when the one above is routine.
Layer 1: Explain (start here)
AI reads what already happened and tells you what it means. You give it a profit and loss export, and you ask which costs rose fastest and why that might be. Nothing changes in your books. The worst outcome is a bad explanation, which you can check against the numbers in front of you. This is where every solo owner should start, because it builds financial attention, not just speed.
Good first questions:
- What were my three largest cost increases this quarter compared with last quarter?
- Which customers or products made up the top 80% of revenue?
- Which recurring subscriptions have I paid for in the last six months?
Layer 2: Scenario
Now AI helps you think forward. What happens to margin if a supplier raises prices by 8%? How many extra clients at my current rate would cover a new hire? How much cash do I need to get through my slow months?
The maths here must be visible. Ask AI to build the scenario in a spreadsheet with its formulas showing, rather than stating an answer in a chat, so you can change an assumption and check the result yourself. Our review of the Google Sheets AI function covers what a spreadsheet can do well here and where its limits are.
Layer 3: Draft for approval
Only here does AI prepare something that will affect real money: a chase email for an unpaid invoice, a bill categorization, a payroll run. It prepares and you approve, every time. This is the layer the software industry is racing toward, and the one where a solo owner has no second person to catch a mistake.
Our view on this has not changed since we looked at QuickBooks AI drafting payroll: let it do the typing, never the trusting. For a low-risk way into this layer, the late-invoice reminder system is a good first project, because nothing leaves your outbox without your eyes on it.
Where to stop, and why that line matters
A strategy is as much about what you refuse to do. Keep these out of AI’s hands, whatever the demo promises:
- Tax positions. AI can organize your records and draft questions for your accountant. It should not decide how something is treated.
- Anything that moves money without you. Payments, transfers and payroll submission need a human click.
- Figures you cannot verify. If you cannot trace a number back to a source document, it does not go into a decision.
- Sensitive data in the wrong plan. Before you paste bank exports or client names anywhere, check what your plan does with your data, and remove details the task does not need.
Write those rules down. With 74% of AI-using businesses in the Goldman survey running without a formal policy, a single page puts you ahead of most of them, and our guide to writing an AI policy on one page gives you the template.
Your 90-day plan
You do not need to reorganize the business. You need a recurring habit that compounds.
- Month one, Explain. Book a 45-minute monthly money review. Export your numbers, ask the three questions above, and write down one thing you learned.
- Month two, Scenario. Build one spreadsheet model for your biggest open question, usually pricing or hiring. Keep the formulas visible, and test one result by hand.
- Month three, Draft. Pick one money task you do repeatedly and let AI prepare it for your approval. Track how many drafts you needed to fix. If you are correcting most of them, go back a layer.
The prize is not saved time, although you will save some. The prize is that a one-person business finally gets a monthly finance review, something most solo owners have never had. Goldman’s numbers suggest more of your peers are starting to get one.
Marketing got you comfortable with AI. Money is where it starts to pay. Schedule the first 45-minute review this week, and put it on the calendar as a recurring meeting with yourself.
