For decades, the Altman Z-Score has been the gold standard for assessing corporate distress. It takes five weighted financial ratios—working capital to total assets, retained earnings to total assets, EBIT to total assets, market value of equity to book value of total liabilities, and sales to total assets—and compresses them into a single number. That number tells you whether a company is in the “safe zone” (above 3.0), the “grey zone” (1.8 to 3.0), or the “distress zone” (below 1.8). The problem? Most finance teams still calculate this manually in spreadsheets, pulling data from multiple systems, re-keying figures, and updating formulas by hand each quarter. This is slow, error-prone, and reactive. By the time the Z-Score drops, the damage is often already done.
The deeper friction is interpretation. A Z-Score of 2.1 tells you the company is in the grey zone, but it doesn’t tell you *why* or *what to do about it*. Is the deterioration driven by shrinking margins, a rising debt burden, or a one-off asset write-down? Traditional models stop at the number. They don’t flag which ratio is dragging the score down, nor do they suggest corrective actions. This is where ChatGPT changes the game. By feeding the model the same financial inputs, you can get not only the Z-Score but also a narrative breakdown of the drivers, a sensitivity analysis, and a prioritized list of actions—all in seconds. The AI doesn’t replace your judgment; it amplifies your ability to act before creditors or investors force your hand.
In this post, you’ll learn two practical ways to use ChatGPT for bankruptcy prediction. First, we’ll build a structured prompt that turns raw financial statements into a full Altman Z-Score analysis with driver commentary. Second, we’ll create a prompt for scenario stress-testing, so you can see how a potential downturn or a new capital raise would shift the score. Both prompts are designed for CFOs, controllers, and financial analysts who need speed and clarity, not academic theory.
Why AI Beats Spreadsheet-Only Approaches
Spreadsheets are fine for calculation, but they’re terrible at context. A Z-Score is a point-in-time snapshot; it doesn’t tell you the trajectory. ChatGPT can compare your current ratios to prior periods and flag whether the decline is accelerating. It can also reason about industry nuances—a Z-Score of 2.5 might be healthy for a utility but alarming for a tech startup with high burn. The model can incorporate qualitative factors (management changes, litigation, supply chain disruptions) that no formula captures. The result is a more complete picture, delivered in plain language that boards and lenders can understand.
Another advantage is speed of iteration. In a spreadsheet, testing five different scenarios means copying tabs, adjusting formulas, and re-checking references. With ChatGPT, you just change a few numbers in a prompt and get a revised analysis with new commentary. This turns bankruptcy prediction from a monthly compliance exercise into a dynamic, forward-looking tool. You can run it before every board meeting, after every earnings release, or whenever a major contract is signed or lost.
Prompt 1: Full Altman Z-Score Analysis with Driver Breakdown
The first prompt is your workhorse. It takes the five raw inputs, calculates the Z-Score, and then goes deeper—explaining which ratio is the biggest drag, whether the trend is improving or worsening, and what specific operational levers could move the needle. This is the prompt you’ll use for quarterly reviews, credit covenant assessments, or when a new client asks for a financial health check.
First, read these files completely before responding:
[financial_statements_Q1_Q3.xlsx] — contains balance sheet and income statement data for the last three quarters
[industry_benchmarks.md] — contains median Z-Score components for our industry (manufacturing, revenue $50M-$200M)
Here is a reference for what I want to achieve:
A CFO-level executive summary that goes beyond just the Z-Score number. It should explain the “why” behind the score, highlight the single most damaging ratio, and provide 3-5 actionable recommendations.
Here’s what makes this reference work:
– Uses plain English, not accounting jargon
– Separates the calculation from the interpretation
– Prioritizes recommendations by impact and ease of implementation
– Flags any data quality issues or missing inputs
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A written analysis, 400-600 words, with a table showing the five ratio components and their contribution to the final score.
Recipient’s reaction: They should immediately understand which two ratios to focus on and feel confident presenting this to the audit committee.
Does NOT sound like: A generic textbook explanation of the Z-Score formula. No filler, no theory.
Success means: The board can make a clear decision on whether to renegotiate debt covenants or accelerate cost-cutting.
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
This prompt works because it forces the AI to act like a senior financial advisor, not a calculator. The “success brief” section is critical—it tells the model exactly how the output will be used, which shapes the tone and depth. The instruction to ask clarifying questions first prevents the model from running off with incomplete data. In practice, you’ll upload your actual financials as the reference file, and the AI will ask for missing items like share price (for market value of equity) or preferred stock value before proceeding.
Prompt 2: Scenario Stress-Testing for Forward-Looking Risk
The second prompt is for what-if analysis. Bankruptcy prediction isn’t just about where you are today; it’s about where you’ll be in six months if revenue drops 20%, or if interest rates rise 200 basis points, or if a major customer defaults. This prompt lets you feed in multiple scenarios and get a revised Z-Score for each, along with commentary on which scenario is most dangerous and what early warning signs to monitor.
First, read these files completely before responding:
[base_case_projections.xlsx] — our current Q4 forecast with revenue, COGS, SG&A, and debt schedules
[scenario_definitions.md] — describes three scenarios: mild recession (revenue -10%), severe recession (revenue -25% with 20% customer defaults), and rapid recovery (revenue +15% with new credit line)
Here is a reference for what I want to achieve:
A side-by-side comparison table showing the Z-Score under each scenario, plus a narrative that identifies the single biggest risk factor in each case and the earliest observable leading indicator.
Here’s what makes this reference work:
– Each scenario is clearly parameterized (revenue growth, margin impact, debt level)
– The output separates “mechanical” changes (formula-driven) from “behavioral” changes (e.g., management pulling levers)
– It flags which scenario would trigger debt covenant violations
– It suggests specific monitoring metrics (e.g., days sales outstanding, gross margin trend) to watch weekly
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A table with three columns (scenario, Z-Score, key driver) followed by a 300-word risk narrative.
Recipient’s reaction: They should feel prepared for the worst case and know exactly which metric to watch in the next 30 days.
Does NOT sound like: A doom-and-gloom report. It should be pragmatic and action-oriented.
Success means: Our CFO can take this to the bank and negotiate a covenant waiver proactively, before any breach occurs.
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
This prompt is particularly powerful when you combine it with real-time data. For example, if you’re a controller at a mid-sized manufacturer, you can take your latest monthly actuals, project them forward, and run this stress test before every lender meeting. The AI will not only compute the Z-Score under each scenario but also reason about which assumptions are most sensitive. If the severe recession scenario drops your score below 1.8, the AI will suggest specific actions—like extending supplier payment terms, selling non-core assets, or converting short-term debt to equity—and estimate the impact of each on the score.
Practical Tips for Getting the Most Out of These Prompts
First, be precise with your data inputs. The Z-Score is unforgiving: a wrong market capitalization figure or an outdated total liabilities number will skew everything. Before running the prompt, reconcile your numbers against your latest trial balance. Second, use the “ask clarifying questions” step to your advantage. When the AI asks for clarification, it’s not being difficult—it’s ensuring the output is relevant. Answer those questions thoroughly. Third, run the first prompt monthly, not quarterly. The cost is near zero, and the early warning signal is worth ten times the effort. Finally, don’t stop at the Z-Score. Use the AI’s narrative to dig into operational drivers. If EBIT is the weak ratio, ask a follow-up prompt: “Based on our last three years of income statements, which product line has the lowest contribution margin, and what would happen if we discontinued it?”
One caution: the Altman Z-Score was originally developed for publicly traded manufacturing firms. If you’re analyzing a private company, a service business, or a startup, the model will still work but the thresholds (1.8 and 3.0) may need adjustment. You can ask ChatGPT to recalibrate the zones based on your industry’s historical default rates. Similarly, for non-US companies, you may need to adjust for different accounting standards. The AI can handle this if you mention it in the context file.
Your next step is simple. Take one company you’re analyzing—yours, a supplier, or a potential acquisition target—and run the first prompt today. Use the last four quarters of data. See what the AI produces. Then, in a follow-up conversation, ask it to explain the trend in the retained earnings ratio or to suggest three ways to improve working capital management. The more you iterate, the more the model learns your context and the sharper its insights become. Bankruptcy prediction is no longer a static, backward-looking exercise. With ChatGPT, it’s a real-time, forward-looking tool that puts you in control.
Published on 11 August 2026 on growwithgpt.com