Using AI to build a simple weekly cash flow forecast from…

Most small business owners find out about a cash shortfall the week it happens. This post walks you through a repeatable workflow that gives you a simple cash flow forecast for your small business every Monday morning, built from two reports you already have in your accounting software and an AI you already use. The workflow works because it turns raw invoice and bill data into a structured forecast without formulas, accountants, or new software.

What you need before you start

ChatGPT{:target="_blank"} (GPT-4o) or Claude{:target="_blank"} (3.7 Sonnet) or Google Gemini{:target="_blank"} (2.0 Flash or 2.5 Pro): Any of these accepts pasted CSV data or uploaded files and can generate a multi-week cash forecast. Free tiers work for this; paid tiers handle larger file uploads more reliably.

QuickBooks Online{:target="_blank"}, Xero{:target="_blank"}, or Wave{:target="_blank"}: Your accounting software is the data source. You need access to the reports section. Wave is free. QuickBooks and Xero have paid plans — check their current pricing before signing up.

Microsoft Excel{:target="_blank"} or Google Sheets{:target="_blank"}: Used for a quick 5-minute data clean-up before you paste into the AI. Either works.

Time required: 30–45 minutes the first time you set it up; 10–15 minutes every week after that.

Skill level: No coding required. You need to be comfortable exporting a report and copying text.

Export the two reports that contain your cash picture

You need two numbers: money coming in and money going out. Those live in your AR and AP reports.

If you use QuickBooks Online:

Open QuickBooks Online and click Reports in the left menu.
Search for Accounts Receivable Aging Summary and open it. Set the date range to cover the next 30 days. You will see a list of open invoices grouped by how overdue they are.
Click Export in the top right and select Export to CSV. Save the file as AR_[date].csv.
Return to Reports and search for Unpaid Bills. Click Export to CSV and save it as AP_[date].csv. This report lists all outstanding bills with their due dates.

If you use Xero:

Open Xero and navigate to Reports > Accounting.
Select Aged Receivables Summary. Download it as CSV. Save as AR_[date].csv.
Select Aged Payables Summary. Download it as CSV. Save as AP_[date].csv.

If you use Wave:

Open Wave and click Reports in the left sidebar.
Select Aged Receivables. Export to CSV.
Select Aged Payables. Export to CSV.

These two exports take under 2 minutes once you know where to find them.

Clean up the data before you paste it into AI

This step protects your clients and keeps the AI focused on numbers, not names.

Open your AR_[date].csv in Excel or Google Sheets. You will see client names in one column and invoice amounts in another.
Replace each client name with a generic label: Client A, Client B, and so on. This takes 2–3 minutes for most small businesses.
Delete any columns you do not need for forecasting — phone numbers, addresses, internal notes. Keep: client label, invoice amount, due date, and days overdue. You should end up with a clean 4-column sheet.
Repeat steps 2–3 for your AP_[date].csv. Replace vendor names if you prefer, though vendor privacy is less of a concern than client confidentiality.
Copy the contents of both cleaned sheets into a single plain-text block, or upload both files to your AI chat.

Removing client names takes under 5 minutes and means you never paste confidential business relationships into an external tool. The numbers are what the AI needs anyway.

Run the AI cash flow forecast prompt

Paste this prompt directly after your data. Adjust the bank balance and fixed costs to match your actual situation.

Paste this prompt:

I'm going to give you two data sets from my accounting software. The first is my Accounts Receivable — open invoices with due dates and amounts. The second is my Accounts Payable — bills I owe with due dates and amounts.

My current bank balance is: [insert your balance]

My recurring fixed weekly costs not listed in the AP data are:

Rent: [amount, due date]

Payroll: [amount, weekly or bi-weekly]

Software subscriptions: [amount]

Any other fixed costs: [amount]

Using this data, build me a 4-week rolling cash flow forecast. Show me:

A week-by-week table with: Cash In, Cash Out, Net Cash Movement, and Ending Bank Balance.

A plain-English risk summary of no more than 5 sentences. Flag any week where my ending balance drops below [insert your minimum comfortable balance, e.g. $5,000].

A list of any invoices more than 14 days overdue that I should follow up on this week.

Treat invoices as likely to be paid within 7 days after their due date unless they are already overdue. Treat all bills as paying on their due date. Do not assume any income beyond what I have provided.

That last instruction matters. Without it, the AI will sometimes get optimistic.

Paste your cleaned AR and AP data above this prompt, then paste the prompt itself. The AI will generate a table and a written summary.
Review the Ending Bank Balance row for each week. Any week that drops below your stated minimum is a potential shortfall — it should be flagged in the risk summary.
Save the output. Screenshot it or paste it into a weekly doc so you have a record to compare against next week.

How to read your AI cash flow forecast

The table does one job: it shows you which week your balance gets thin. A week where Cash Out exceeds Cash In is not automatically a problem — if your ending balance stays healthy, you are fine. The problem is when the ending balance drops low at the same time a large bill is due.

The most common cash crunch pattern is a cluster of big bills — rent, payroll, quarterly taxes — landing in a week where major client invoices are already overdue. A monthly P&L hides this completely. A weekly cash view shows it 3–4 weeks early.

When you see a shortfall flagged:

Chase the invoice first. The AI will list the overdue ones. Call or email the same day.
Delay a discretionary purchase if the shortfall is small and temporary.
Talk to your bank about your credit line before the shortfall arrives, not after. At current small business line-of-credit rates of 8–12% APR, early outreach gives you options. Reactive borrowing costs more.

Make it a weekly habit: 15 minutes on Monday morning

After the first setup, the weekly run shrinks dramatically. Here is the full recurring sequence:

Export AR and AP reports from your accounting software (2 minutes).
Replace any new client names with labels in your clean-up sheet (2 minutes).
Open your AI chat, paste your saved prompt template, update your bank balance and any new fixed costs, paste the data (5 minutes).
Read the risk summary and flag anything that needs action today (3–5 minutes).
Send a follow-up email or call on any invoice the AI flagged as overdue (time varies).

Keep your base prompt in a text file or a notes app. Each week you update three things: your bank balance, the data paste, and any new fixed costs. Everything else stays the same.

When something goes wrong

The AI produces a table but the numbers look wrong. The most likely cause is misaligned columns — the AI read the date column as an amount. Open your CSV, confirm the column order is: label, amount, due date. Re-paste with a note in the prompt that says "Column 1 is client label, Column 2 is amount owed, Column 3 is due date."

The AI ignores some of your fixed costs. You listed them in paragraph form and the AI skimmed them. Reformat them as a bulleted list inside the prompt, exactly as shown in the template above. Bullet points process more reliably than prose in financial prompts.

The forecast looks fine but you still ran short. The AI only knows what you gave it. A client who paid a partial amount, an auto-renewing subscription you forgot, or a bill that came in after your export — none of those appear unless you add them. Treat this output as directional, not accounting-grade. For businesses with complex payment arrangements or irregular income, a monthly accountant review remains worthwhile.

What to do next

Run this workflow this Monday with real data from your accounting software. The first run takes longer; that is the setup cost. By the second week, you will have a template and a habit.

If you want to extend this into a full client invoicing follow-up system, see how to build an AI-assisted accounts receivable follow-up sequence for overdue invoice emails.

FAQ

Can I forecast cash flow without an accountant using just my invoice data? Yes. Export your invoice list and bill list from your accounting software or spreadsheet in the same format: label, amount, due date. Paste them into the AI prompt exactly as described. The AI does not care whether the data came from QuickBooks or a manual spreadsheet — it cares about structure. Make sure your due dates are in a consistent format (MM/DD/YYYY works well).

Is it safe to paste financial data into ChatGPT or Claude? Neither OpenAI{:target="_blank"} nor Anthropic{:target="_blank"} uses your conversation data to train their models by default when you use paid tiers. Check the current privacy settings in your account to confirm. The client name replacement step in this workflow removes the most sensitive identifiers before anything leaves your machine.

How accurate is an AI cash flow forecast compared to accounting software forecasting tools? AI-generated forecasts are directional. They tell you which week looks risky based on the data you provided. They do not account for partial payments, returned checks, or invoices paid outside your normal terms unless you explicitly include that context in the prompt. Accounting software forecasting tools like QuickBooks Cash Flow Planner use live data but require higher subscription tiers and still rely on your data being clean. The AI approach is more flexible; the accounting software approach is more automated.

What if I have seasonal income and my cash flow varies a lot week to week? Add a note to your prompt: "My business is seasonal. The following weeks historically have lower income: [list them]." The AI will factor that into the risk summary. You can also add a row to your prompt with expected income from recurring retainers or contracts even if they are not yet invoiced — label it "Expected recurring income" and include the amount and timing.

Do I need to do this weekly, or can I do it monthly? Monthly is better than nothing. Weekly is meaningfully better than monthly. The most common shortfall pattern — large bills coinciding with overdue invoices — happens at the week level, not the month level. A monthly view can show a healthy average while hiding a brutal two-week stretch in the middle. The 10–15 minute time investment is the only reason to skip a week.

Using AI to build a simple weekly cash flow forecast from your invoices and bills so you see a cash shortfall before it hits