AI costing prompts · By role
AI costing prompts for private equity and investors
In diligence you are trying to learn, fast, whether the revenue in the model is the revenue that actually earns a return. Headline margin rarely survives contact with cost-to-serve. AI can help you structure that analysis and frame the questions to put to management, but a number it invents in a data room is a number that detonates after close. These prompts are built for analysis that survives post-acquisition scrutiny.
In short
Use AI to interrogate margin quality, not to manufacture it. The prompts below help you test cost-to-serve across the customer base, find which revenue streams are genuinely profitable, and probe the durability of pricing. Each one forces the model to work only from the figures in the data room, to separate fact from inference, and to flag every gap as a diligence question rather than fill it with a plausible guess. The output is a list of things to verify, which is what diligence should produce.
What an investor should and should not ask AI to do
AI is useful for the analytical scaffolding of diligence under time pressure. It can structure a cost-to-serve analysis from the data provided, rank customers and revenue streams by true profitability, surface concentration and the long thin tail of marginal accounts, and turn a target's pricing story into a set of pointed questions for management. It is also a quick way to draft the cost section of an investment memo from your own validated figures, and to play sceptic against the seller's narrative.
It is dangerous when you let it stand in for verification. Asking it to "estimate the margin," "assume a normal cost-to-serve," or "benchmark this against the sector" produces confident numbers with no basis in the target's books, and in diligence the entire point is to know what the books say, not what is typical. Treat any figure the model offers that is not in the data room as an open question, not an answer. Keep the verification, the judgement and the conclusion firmly yours.
Three prompts to start with
1. Test cost-to-serve across the customer base
Where headline margin meets reality. Builds on the cost-to-serve page.
You are a diligence analyst testing the cost-to-serve of a target's customer base. Work only from the data in front of you, which I will paste. Do not invent any numbers, benchmarks or "typical" figures. Treat every gap as a diligence question. Label missing data as DATA MISSING with the exact ask for management. My data: - Customers or segments with revenue and gross margin: [paste] - The cost-to-serve components available (logistics, support, returns, rebates, terms): [paste] Steps: 1. For each customer, compute net profit after the cost-to-serve components provided; show the formula before the value, cite the source row. 2. Identify customers that are gross-margin healthy but net-margin poor once cost-to-serve is applied. 3. List every cost-to-serve component that is NOT in the data and frame it as a question for management. 4. Separate clearly: calculated from the data, inferred, and to be verified. 5. Reconcile customer net profit to the total provided and flag any difference.
2. Find which revenue streams are really profitable
Quality of revenue, not just quantity. See customer profitability.
You are helping an investor assess the quality of a target's revenue. Work only from the data I paste. Do not invent figures or sector benchmarks. Flag every gap as a question. Label missing data as DATA MISSING. My data: - Revenue streams, products or contracts with revenue and any available cost and margin: [paste] - Contract terms or recurring vs one-off split, where provided: [paste] Steps: 1. Rank revenue streams by profitability using only the data provided; show each calculation as a formula. 2. Identify concentration: how much profit comes from how few streams, cited to the rows. 3. Distinguish recurring from one-off profit only where the data supports it; otherwise mark as unverified. 4. Separate clearly: what the data shows, what you inferred, and what management must confirm. 5. Produce a short list of the highest-impact diligence questions this analysis raises.
3. Probe the durability of pricing
How much of the margin is structural and how much is fragile. Links to pricing decisions.
You are a critical diligence reviewer probing the durability of a target's pricing. Work only from the data I paste. Do not invent elasticities, competitor prices or volumes. Treat anything not in the data as a question, not an answer. My data: - Price, volume and unit cost by product or contract: [paste] - Any pricing history, discounting pattern or contractual price protection: [paste what exists] Steps: 1. Show contribution per line as a formula and flag any line where price barely covers cost. 2. Identify where margin depends on discounting that may not persist, citing the data. 3. List what is unknown about pricing power and frame each as a management question. 4. Separate clearly: evidenced in the data, inferred, and to be verified. 5. State the single pricing assumption that, if wrong, most threatens the investment case.
The one rule
Work only from the data in the room. Do not invent any numbers, benchmarks or volumes. Treat every gap as a diligence question and label it DATA MISSING.
A fabricated number in a diligence memo is a liability that surfaces after close. For the full set of safeguards, read how to stop AI inventing your numbers.
Margin analysis that survives the first 100 days
A prompt can structure your diligence and sharpen your questions, but it cannot rebuild a target's costing the way the value-creation plan will need. That is the work we do alongside investors and portfolio companies: a defensible cost-to-serve and customer profitability model that holds up in diligence and becomes the operating tool after close. If you want margin analysis that survives scrutiny on both sides of the deal, start with a health check.