Export your profit and loss with monthly columns and last year alongside, strip out names, and ask an AI assistant to explain each section in plain English, calculate gross margin and overheads as a share of sales, and list the biggest changes with possible causes. Then check its sums in a spreadsheet and take its open questions to your accountant.
Most people start with "is my business doing well?" and get polite, generic reassurance. The answers get useful when you give the AI context about how you make money and ask about specific lines. And remember what it can't see: cash in the bank, customer deposits held, loan repayments and big purchases don't appear on a P&L, so a business that's profitable but short of cash looks fine unless you give it more than this one report.
Step 1: Export a version of the report the AI can read properly
The export decides how good the explanation is. Spend five minutes on it.
- Monthly columns for the last 12 months, plus the same 12 months last year or a year-on-year comparison. A single annual total hides seasonality, which is often the whole story.
- Accrual basis if you can. Accrual means income and costs are recorded in the month they relate to, not when cash moves. Most cloud accounting software offers both; the report settings will show which you're looking at.
- Spreadsheet format rather than PDF. A CSV or Excel export keeps the numbers as numbers. A PDF often works, but columns sometimes get misread.
- No names. If your line descriptions include customer, supplier or staff names, rename them ("Wages: operations team") before uploading.
An illustrative layout that works well:
Account | Oct | Nov | ... | Sep | Total | Last year total
Tour sales | | | | | |
Cost of sales: hotels | | | | | |
Cost of sales: coaches | | | | | |
Cost of sales: guides | | | | | |
Gross profit | | | | | |
Wages: office team | | | | | |
Marketing | | | | | |
...
Net profit | | | | | |Step 2: Tell the AI how your business makes money
Without context the AI treats every business as the same. A few sentences first change the quality of everything after. A filled-in example for an illustrative small tour operator:
Context before you read the attached P&L:
- We're a small tour operator running group trips of 4-10 days.
- Customers pay a deposit when they book (often 6-9 months ahead) and the
balance 8 weeks before departure. Our accountant records the income in the
month the trip departs, so deposits don't show on this report.
- Cost of sales is hotels, coaches, guides and entrance tickets, all per trip.
- Most trips run May to September; October to March is quiet.
- Accrual basis. Figures in USD.
I'm not an accountant. Explain things plainly and tell me when you're unsure.The line about deposits is the one that stops the AI telling you that winter losses mean the business is in trouble. The line "tell me when you're unsure" gives it permission to say so, which reduces confident guessing.
Step 3: Ask for a line-by-line translation
Ask for each section in turn rather than the whole report at once. A prompt that works:
Go through the P&L section by section: income, cost of sales, gross profit,
overheads, net profit. For each line: say what it means in one plain sentence,
give the 12-month total and last year's, and say whether the change looks
normal for a business like ours or worth a question. Don't calculate ratios
yet. Mark anything you're inferring rather than reading from the report.An illustrative extract of what comes back:
Cost of sales: coaches. What you paid coach companies to move groups. $182,000 this year against $154,000 last year, up about 18%. Tour sales rose about 14%, so coach costs grew faster than sales. Worth a question: has the price per coach day gone up, or did trips use more coach days? (Inferred: I can't see the number of trips or coach days.)
That's a useful answer because it separates what the report says from what the AI is guessing. If yours doesn't, ask it to.
The lines that confuse owners most, and the question worth asking about each:
| Line | What it means | Question to ask the AI |
|---|---|---|
| Cost of sales (direct costs) | Costs that rise and fall with each sale or job | Which of these grew faster than sales, and by how much? |
| Gross profit | Sales minus direct costs: what's left to pay overheads | What share of sales is this each month, and is it drifting? |
| Overheads (operating expenses) | Costs you pay whether you sell or not | Which overheads are fixed and which could be cut quickly? |
| Depreciation | The cost of equipment spread over its useful life, not cash spent this year | How much of the year's costs are non-cash? |
| Interest | The cost of borrowing (the loan repayment itself is not on the P&L) | What would profit be before interest? |
| Other income | Income outside your main trade: grants, insurance payouts, interest | What's profit without the one-off items? |
Step 4: Get the ratios, then check the maths yourself
Three ratios tell most small businesses what they need:
- Gross margin = gross profit ÷ sales. How much of each dollar of sales is left after direct costs.
- Overheads as a share of sales = total overheads ÷ sales. How much of each dollar the running costs take.
- Net margin = net profit ÷ sales. What's left at the end.
Ask the AI to calculate them by month and for the year, showing the figures it used. Then check at least the annual ones in a spreadsheet with a formula like =B8/B2. AI assistants are good at explaining and noticeably less reliable at arithmetic across a wide table, and the mistakes look plausible.
A realistic example of how that goes wrong. Asked for gross margin, an assistant replied "Gross margin is 39.5%". The sum it had done was gross profit ($340,000) divided by cost of sales ($860,000), not by sales ($1,200,000). The correct figure is 28.3%. Nothing about the answer looked wrong until the owner checked it. Asking for "the figures you used" in the prompt makes this kind of error visible straight away.
Step 5: Ask why the numbers moved
Once the ratios are right, ask for the story behind the biggest changes:
List the five biggest changes against last year, in dollars. For each,
give the size of the change, two or three possible causes for a business
like ours, and what I could check to find out which is right. Label every
cause "possible" unless the report itself shows it.Illustrative extract:
Marketing up $19,000 (from $36,000 to $55,000). Possible causes: more paid advertising; a brochure or trade show; a new agency retainer. Check: the marketing account's transactions for the three largest months, and whether bookings rose in the months after.
The "what to check" part is the valuable bit. It turns the AI from something that tells you answers into something that tells you where to look, which is safer and usually faster.
Sharper questions than "how am I doing?"
Once the basics are clear, specific questions get specific answers. These work for most businesses; swap in your own line names.
- "Which three costs grew faster than sales, and by how many percentage points?" Costs growing faster than income is the most common reason profit falls while sales rise.
- "If sales were 10% lower next year and costs stayed the same, what would net profit be?" A quick sense of how fragile the profit is.
- "Which months made a loss, and was that true last year as well?" Separates seasonality from a new problem.
- "What's the average gross profit per month in the busy season against the quiet season?" Useful for deciding how much cash to hold back in good months.
- "Which overheads have gone up every quarter for a year?" Creeping subscriptions and contracts show up here.
- "What would gross margin be if [largest direct cost] went up 5%?" Prepares you for a supplier price rise before it lands.
- "Which lines would a lender or buyer ask about first?" A good rehearsal before a loan application or a sale conversation.
- "What can't you tell from this report?" The most underrated question. A good answer lists cash, deposits, stock, debts and owner's pay, and reminds you what else to look at.
Ask them one or two at a time. A list of eight in a single message gets eight shallow answers.
A tour operator's year, read through AI
Pulling the steps together for the illustrative tour operator above:
| This year | Last year | Change | |
|---|---|---|---|
| Tour sales | $1,200,000 | $1,050,000 | +14.3% |
| Cost of sales | $860,000 | $735,000 | +17.0% |
| Gross profit | $340,000 | $315,000 | +7.9% |
| Gross margin | 28.3% | 30.0% | -1.7 points |
| Overheads | $290,000 | $255,000 | +13.7% |
| Net profit | $50,000 | $60,000 | -16.7% |
| Net margin | 4.2% | 5.7% | -1.5 points |
What the AI surfaced, once it had context: sales grew, but profit fell because direct costs grew faster than sales (coaches especially) and overheads rose with a bigger marketing spend. Without context, its first reading had been that the business "lost money in seven months of the year, which is a concern". With the context about departures, it correctly described that as seasonality, and pointed out that the winter losses were about the same size as last year's.
What the owner then checked: the coach account showed a new supplier with a higher day rate and fuel surcharges added mid-season. The marketing rise was a trade show that produced a strong block of bookings, which will show up in next year's sales. The question for the accountant was whether the price increase for next season (planned at 6%) would restore the gross margin, which the AI helped model but the owner confirmed with actual quotes.
What the AI couldn't see: $410,000 of customer deposits held for next year's trips. They're a liability on the balance sheet, not income, and they're the reason the bank balance looks healthy in a winter that the P&L shows as loss-making. That gap is the next section.
What a P&L can't tell you, as a storage facility found
An illustrative self-storage owner asked an assistant to read the year's P&L: sales $420,000, net profit $96,000. The assistant's summary said the business was "highly profitable, so cash should be comfortable". The owner's bank balance had fallen by $120,000 over the year.
Both were true. During the year the owner spent $180,000 building a new block of units, which is recorded as an asset rather than a cost, so it doesn't appear in the P&L beyond a slice of depreciation. Loan repayments of $36,000 were capital, not expense. The P&L measured profit; it said nothing about where the cash went.
The fix is simple: when the question involves cash, give the AI the balance sheet and a cash-flow statement as well, or at least the bank balance at the start and end of the year and a list of large capital payments. Its explanation then changes to something like "profit of $96,000 plus depreciation, less $180,000 of building work and $36,000 of loan capital, roughly explains the fall in cash", which is what the owner needed to hear. Cash-flow forecasting in plain English covers the forward-looking version of that question.
Adjust for your own pay before judging the profit
How much the owner takes out, and where it appears, changes what "profit" means. In some business structures the owner's pay is a salary on the P&L like anyone else's. In others, such as many sole traders and partnerships, the owner is paid from the profit, so the net profit line is before the owner has been paid at all. A "profit" of $60,000 in that case may just be a modest wage.
Tell the AI which applies, and ask it to show the figure both ways. For an illustrative two-partner letting business showing net profit of $118,000 with no partner salaries on the report, the useful restatement is: "After paying each partner a market wage of $45,000, the business earns $28,000, a net margin of about 5% on sales of $560,000." That's a very different conversation about prices and costs from the headline figure. Your accountant can tell you which treatment your structure uses if you're unsure.
Three checks before you act on anything it said
- The ratios agree with your spreadsheet. Recalculate the annual gross margin, overhead share and net margin with your own formulas. If any differ, go back through the AI's figures before reading its explanations.
- The top two explanations match the transactions. Open the ledger for the two biggest changes and look at the largest transactions behind them. If the AI said "more coach days" and the transactions show a price rise, trust the transactions.
- Your accountant recognises the picture. Send them the one-page note once a quarter. If they say "that's not how we record that", you've found a gap in your context prompt, and fixing it improves every month after.
Twenty minutes on these checks is what separates a useful reading from a plausible one.
Turning the conversation into a one-page monthly note
Once the prompts work, reuse them every month and keep the answer to one page. Illustrative note for the tour operator's August:
AUGUST - ONE-PAGE NOTE
Sales $236,000 (Aug last year $214,000, +10%)
Gross margin 27.1% (last Aug 29.4%) - coach costs again; new supplier rate
Overheads $24,800 (in line with budget)
Net profit $39,200 (last Aug $41,900)
Three things to look at:
1. Coach cost per trip day: $612 vs $540 last year. Ask second supplier to quote.
2. Two trips ran with 6 of 10 places filled. Minimum numbers policy?
3. Card fees up $1,900: more customers paying balances by card.
Questions for accountant: none this month.That note takes about 15 minutes to produce with a saved prompt and a fresh export. If you want it produced without the export step, automating monthly management reports covers the options, and explaining budget-versus-actual variances adds a budget column to the same routine.
Which tool to use, and what to keep out of it
Any of the main chat assistants (ChatGPT, Claude, Gemini) can do this well with an uploaded export. Use a business plan, where content isn't used for model training by default, or switch off the model-training setting in a personal plan's privacy settings. If you work in Excel and your plan includes Copilot in the Office apps, it can explain and chart the exported report without the file leaving your Microsoft 365 account.
The AI built into your accounting software is the other route. Xero's JAX and QuickBooks' Intuit AI features can answer questions about your figures from the live ledger, which saves the export and means the numbers are always current. The trade-off is less control over the prompt, so it's worth trying both on the same question and comparing answers once.
Keep out: staff names next to salaries, customer names, bank and tax reference numbers. None of them improve the explanation. And keep your accountant in the loop for anything that affects tax or the year-end accounts. The AI helps you understand the report; it doesn't decide how items should be recorded. Once the P&L reads clearly, finding AI savings line by line in your P&L is a practical next use of the same export.
Questions about reading a P&L with AI
Is it safe to upload my profit and loss to ChatGPT or Claude?
A P&L is usually less sensitive than payroll or bank data, but it's still confidential. Use a business plan, where content isn't used for training by default, or switch off the model-training setting on a personal plan. Remove customer, supplier and staff names from line descriptions before uploading, and never include bank account or tax reference numbers.
Should I give the AI a cash-basis or accrual-basis report?
Accrual basis, if your software offers it, because it matches income and costs to the period they relate to, so margins mean something. A cash-basis report shows when money moved, which is useful for cash planning but misleading for profit. Tell the AI which basis the report uses; the explanation of a lumpy month depends on it.
Can AI tell me whether my prices are too low?
It can show you what your gross margin is, how it's moved and what margin you'd need to cover overheads and a target profit. Whether your prices are too low also depends on competitors, demand and your costs per job, which the P&L doesn't show. Use the AI's margin maths as a starting point, then look at pricing product by product or job by job.
Further reads
- How to Track Profit per Project in a Service Business With AI — Go below the P&L to profit on each job or project.
- Break-Even Analysis With AI: A Worked Example for a New Product — Use the same margins to work out a break-even point.
- How to Get a Weekly Business Summary Emailed to You by AI — Get a short summary of the numbers every week.
- How to Build a 13-Week Cash Flow Forecast With AI Help — The cash view that the P&L leaves out.
- How to Prepare a Business Loan Application With AI Help — Explain your numbers to a lender with the same method.
- Can AI Do My Bookkeeping? What Still Needs an Accountant — Make sure the books behind the P&L are right first.
- How to Check Your Margins Product by Product With AI — Work out what each product really earns after discounts, shipping, fees and returns, with AI doing the sums in code and you checking three products by hand.
- How to Speed Up Month-End Close in a Small Business With AI — A day-by-day close calendar, four AI prompts with sample outputs, and a farm shop that went from a nine-day close to four without skipping checks.
- Can AI Do My Business Taxes? What It Can Safely Prepare — What AI can safely prepare for your business tax return, where its tax answers go wrong, and a year-end pack you can build before the accountant sees it.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI, Anthropic and Google business plan data-use terms (checked September 2026); Xero and QuickBooks help on their built-in AI assistants.