Easy + Rigorous Forecasting

It's Thursday afternoon and you're reviewing your finances when that sinking feeling hits – there's not enough cash to cover next month’s payroll and your biggest customer just emailed asking for a payment extension 😬.

Cash flow surprises are the business equivalent of surprise house guests – stressful, disruptive, and arriving at the absolute worst time. Fortunately, generative AI can help you see these "guests" coming weeks in advance and guide you to prepare accordingly.

Here’s how.

Step 1:
Gather Your Financial Data

To create accurate forecasts, you’ll need to assemble these data, which likely means exporting files from your accounting software (QuickBooks, Xero, FreshBooks, etc):

• Bank statements (3+ months)
• Past cash flow statements
• Accounts receivable aging reports
• Customer payment records
• Seasonal trend data (if available)
  
  

Get as much as possible in spreadsheet format; .PDFs should work in a pinch but tabular data will make this more effective. If you’re using Claude, then export data in a .CSV as it struggles with .XLSX. 

The quality of your forecasts will directly reflect the quality of your data. Take some time to clean up your records, reconcile accounts, and ensure everything is current before proceeding. Consider this exercise a forcing function.

Note that generative AI models universally struggle with human-crafted financial model spreadsheets given their formatting and the playoff between formulas and values. But all of the frontier models will do great with standardized exports from pretty much any accounting software.  

Step 2:
Build Your “Cash Flow Advisor” Jig

Before starting, make sure that you’ve opted out of including your data in training in whichever AI model you’ll be using. This is automatic if you’re paying for Claude or Gemini via Workspace, in ChatGPT you’ll need to turn this off in Settings -> Data Controls -> Improve the Model for Everyone.

Now it’s time to build a "Cash Flow Advisor" jig—a customized version of your preferred AI model, with specific knowledge of your business's financial patterns. In your custom GPT (ChatGPT), Project (Claude or ChatGPT), or Gem (Gemini), start by writing custom instructions that go something like this:

~~~

Task: You are an AI financial advisor specializing in cash flow forecasting. Analyze our financial data to identify key insights and provide actionable recommendations.

Context: Our goal is to improve our cash position by understanding cash flow patterns, customer payment behavior, and cost structures. Your analysis should focus on practical financial strategies.

Analysis Areas:

• Differentiate normal vs. abnormal cash flow patterns based on historical data.
• Identify customer payment trends, categorizing clients by reliability and risk.
• Assess seasonal fluctuations in revenue and expenses.
• Track fixed cost trends and patterns in variable expenses.
• Detect early warning signs of potential cash flow issues.
  
  

Format:

• Findings: Present specific data-driven insights.
• Actionable Insights: Recommend practical steps to optimize cash flow.
• Confidence Level: Explain forecast accuracy and key factors that could impact outcomes.
  
  

Tone: Professional, data-driven, and proactive, with a focus on clarity and actionable financial advice.

Persona: Act as an expert financial strategist with deep experience in forecasting, business finance, and risk management.

~~~

Use this as a jumping off point; go right up to the character limit and provide as much context about your business as possible. If you’re not sure where to start, go with simpler instructions that you’ll revisit as you learn more.

Here’s what I wrote about prompting last week:

There's a fair bit of snake oil out there about prompting; if you're not using the API then going deep on prompt engineering likely isn't worth your time. Generally, being good at prompting is just like being a good manager: give clear instructions, provide examples of what good looks like, provide appropriate context, and give feedback along the way. Using delimiters helps; if you don't know what those are or if you're stuck, this guide should help.

Upload your financial data to the Project / Gem / GPT’s knowledge, including:

• Past cash flow statements
• Bank transaction histories
• Accounts receivable aging reports
• Customer payment records
• Seasonal trend data (if available)

Don't worry if your financial records aren't perfectly organized – part of your AI's job is to make sense of messy data. Just provide what you have and be clear about any limitations or gaps. It’s also a good idea to make sure you’re using plain language file naming conventions (or including a breakdown in the custom instructions). 

Step 3:
Create a Forecast

Now it’s time to generate your first cash flow forecast. Start with a 90-day horizon, which gives you enough time to spot trends without getting too speculative.

Try this prompt in a new conversation in your Cash Flow Advisor jig:

~~~

Task: Analyze our financial data and generate a 13-week cash flow forecast, detailing our expected weekly cash position.

Context: Our objective is to maintain financial stability by forecasting cash inflows, outflows, and runway while identifying potential risks or liquidity concerns. Your analysis should be data-driven and actionable, focusing on key insights that can help improve cash flow management.

Forecast Structure (for each week):

• Starting Cash Balance
• Expected Cash Inflows (broken down by major categories, e.g., revenue, customer payments, financing)
• Expected Cash Outflows (broken down by major categories, e.g., payroll, rent, supplier payments)
• Ending Cash Balance
• Runway (weeks of operations at the current burn rate)
  
  

Key Areas of Focus:

• Risk Alerts: Flag any weeks where our cash balance falls below [your minimum comfort level] or experiences unusual fluctuations.
• Trend Analysis: Explain what’s driving the changes, including seasonal patterns, payment timing, or unexpected costs.
• Actionable Insights: Provide specific recommendations to optimize cash flow and mitigate shortfalls.

~~~

What makes AI forecasting different from traditional spreadsheet methods is pattern recognition. Your Cash Flow Advisor might identify that:

• Certain customers consistently pay on the 20th regardless of invoice terms
• Your cash flow tightens predictably three days before payroll
• Seasonal utility costs spike earlier than you realized
• Vendor payments cluster in ways that create temporary cash squeezes

These insights go beyond simple math to reveal the underlying rhythm of your business's cash flow – patterns that you might never otherwise observe. Ask questions based on what you learn; being able to go back-and-forth in a conversation with your data is what makes this approach so powerful.

Step 4:
Explore Scenarios

The real magic happens when you start exploring scenarios that would normally require rebuilding entire spreadsheet models. With your Cash Flow Advisor, you can run these simulations in seconds.

Here's a few ideas of scenario prompts to try:

~~~

How would our cash flow forecast change if our average customer payment time extended by 15 days? What preventative actions could we take now?

~~~

We're considering hiring two new employees at $5,000/month each starting next month. How would this impact our cash runway? What revenue increase would we need to maintain our current financial position?

~~~

We need to purchase $30,000 of new equipment in the next quarter. Based on our cash flow patterns, when would be the optimal time to make this purchase with minimal disruption?

~~~

What would happen to our cash position if our largest client delayed payment by 30 days or stopped ordering entirely? Do we have sufficient reserves?

~~~

Use these as a jumping off-point; the possibilities are endless.

Step 5:
Make it Visual

Numbers in spreadsheets rarely tell compelling stories. Visual representations of your cash flow transform abstract figures into intuitive patterns that anyone can understand – especially important when communicating with colleagues who may not be financially oriented.

Ask your AI to create visualizations with a prompt like this:

~~~

Based on our cash flow forecast, please create:

1. A cash runway timeline showing projected daily cash balance for the next 90 days
2. A weekly comparison chart of expected inflows vs. outflows
3. A heat map highlighting high-risk periods where cash might drop below our safety threshold
4. A dashboard summary showing key metrics: minimum cash position, average weekly burn rate, and days to potential cash shortfalls

~~~

As you see visualizations of scenarios, they’re likely to prompt further questions - and this is the real value of generative AI. Ask for new visualizations or scenarios based on what you observe.

Claude created the above visualization after a bit of back-and-forth with me. Follow this link to explore it further; make sure to scroll down to see Claude's full analysis and recommendations, including how pulling forward the launch of Acme Corp's new "earthquake pills" would have a huge impact 😂.

Notice how the entire dashboard is interactive; hovering over a point gives you further details. Producing a dashboard like this used to take hours and loads of specialized knowledge; now it takes just minutes and almost anyone can do it.

This doesn't mean you should only spend a few minutes on forecasting; rather, shift how you spend your time and delve deeper into scenarios. Trade the time you used to spend building dashboards for increasing the rigor of your analysis. 

So many people miss the real opportunties with generative AI because they see it only as a tool to drive efficiency. Yes, efficiency and speed matter and they're a great place to start - but don't stop there. Now with faster and cheaper approaches, it's time to re-think every business process - including how you forecast cash flows.

Step 6:
Repeat Quarterly

Set aside a couple of hours each quarter to revisit this activity. I live and die by my calendar, so I've got a quarterly recurring event that lasts 90 minutes to repeat this activity in the middle of the last month of the quarter.

Over time, you’ll build more and more data inside of your Cash Flow Advisor. Pair this accumulation of data over time with the inevitable scaling of AI capabilities, and your Advisor will only get more insightful and helpful each time you use it. 

The Bottom Line (🙄)

Cash flow forecasting with AI isn't just about avoiding problems – it's about building the financial confidence to make bold moves at the right time. When you have clarity about your cash position weeks to months in advance, you can:

• Negotiate with suppliers from a position of strength
• Proactively reach out to customers before their payments become critical
• Make hiring decisions with full awareness of the financial implications

Whether you're a solopreneur managing everything yourself or a growing business with a dedicated finance team, AI-powered forecasting gives you the financial visibility that was once available only to large enterprises with sophisticated FP&A departments.

While the days of cash flow surprises will never completely disappear, this approach should help mitigate many of the common challenges SMBs face. With these techniques, you can transform cash flow management from stressful and haphazard to easy and rigorous.

✨ ✌🏻 ✨

The AI Models for SMBs Comparison Chart is now updated to include Claude 3.7 and ChatGPT 4.5 plus a handful of other items.

Stay subscribed for a Q2 update in mid-April.