AI costing prompts · By role
AI costing prompts for the CEO or founder
You run the business and you carry the profit question in your head, but you may not have a finance team to answer it. The instinct that "we are busy but it does not feel like it is paying" is usually right, and usually points at a few customers or products quietly losing money. AI can help you get plain-language answers from the numbers you do have, as long as it never makes any of them up.
In short
You do not need accounting jargon to find out which customers and products actually make money. The prompts below help you spot your most and least profitable customers, understand what it really costs to serve them, and check whether your prices cover the work involved. Each one is written so the AI works only from figures you paste in, explains itself in plain language, and tells you what it cannot answer rather than inventing a number that sounds right.
What a founder should and should not ask AI to do
AI is genuinely helpful when you have the raw figures but not the time or the training to turn them into a clear answer. Paste in your customer revenue and a rough sense of the effort each one takes, and it can rank them, explain in plain words why the big ones are not always the best ones, and translate "cost-to-serve" out of consultant-speak. It is a patient analyst that will walk you through the logic step by step, which is exactly what you want when finance is not your background.
It becomes risky when you ask it to know things only your books know. If you say "estimate my margins" or "what does it usually cost to serve a customer like this," it will give you a confident, generic number that has nothing to do with your business, and a decision built on that is a decision built on fiction. The fix is simple: only let it work with numbers you give it, and when it does not have what it needs, make it say so. You bring the real figures and the gut feel; it brings the structure.
Three prompts to start with
1. Find your most and least profitable customers
The first thing most founders should know and rarely do. Builds on the customer profitability page.
You are a friendly finance analyst helping a business owner with no finance team understand customer profitability. Use plain language and explain each step. Work only from the data I give you. Do not invent any numbers. If something is missing, label it DATA MISSING and tell me in simple terms what to find. My data: - Each customer with their yearly revenue: [paste] - A rough sense of how much work or cost each one takes (hours, deliveries, support, returns): [paste what I have] Steps: 1. Explain in one sentence what "profit after cost-to-serve" means. 2. For each customer, show revenue minus the cost of serving them, as a simple sum before the number. 3. Rank them from most to least profitable. 4. Point out any big customer that is not actually a good customer, and say why in plain words. 5. Tell me clearly which numbers you used, which I should double-check, and what is missing.
2. Understand what it really costs to serve a customer
The hidden costs that turn a good sale into a bad one. See cost-to-serve.
You are helping a business owner work out the true cost of serving one type of customer. Use plain language. Work only from the data I give you. Do not invent any costs or assume "typical" figures. Where you need something I have not given, label it DATA MISSING. My data: - The customer or customer type, and the revenue they bring: [paste] - The activities involved in serving them (orders, deliveries, support calls, returns, custom work): [paste] - What each activity roughly costs or how long it takes: [paste what I have] Steps: 1. List the activities and explain why each one adds cost. 2. Add up the cost to serve this customer, showing the sum before the total. 3. Compare it to the revenue and show whether they are worth it. 4. Flag any activity where my cost looks like a guess so I can confirm it. 5. Tell me in one line what to do if I want a more accurate answer.
3. Check whether your prices cover the work
A fast sanity check on whether a price is actually paying. Links to pricing decisions.
You are helping a business owner check whether a price covers its costs. Use plain language. Work only from the data I give you. Do not invent costs, volumes or competitor prices. If a figure is missing, label it DATA MISSING. My data: - The product or service and its price: [paste] - What it costs me to deliver one (materials, time, delivery, anything else): [paste] - How many I sell in a typical month: [paste if known] Steps: 1. Show price minus cost as a simple sum, then the profit per unit. 2. Tell me in plain words whether this price is healthy, thin, or losing money. 3. Show how many I need to sell to cover any fixed costs I gave you. 4. Separate what you calculated from what I should verify. 5. Suggest the one number I should nail down before changing the price.
The one rule
Work only from the data I give you. Do not invent any numbers, costs or volumes. If something is missing, label it DATA MISSING and tell me in plain words what to find.
Without this, AI will happily hand you a confident number that has nothing to do with your business. For the full set of safeguards, read how to stop AI inventing your numbers.
Compare your customers side by side
The prompts above look at one customer or one price at a time. This one lines several customers up together, in plain language, so you can see the pattern, who to keep, who to fix, who to rethink, instead of judging each one alone. Run it once you have rough figures for a handful of customers.
Plain-language customer comparison
You are a friendly finance analyst helping a business owner with no finance team compare customers side by side. Use plain language and explain each step. Work only from the data I give you. Do not invent any numbers. If something is missing, label it DATA MISSING and tell me in simple terms what to find. I want to see all my main customers next to each other, so I can tell which ones are really worth it and which ones only look good because they are big. Step 1. Take each customer with their yearly revenue and a rough sense of the work they take (hours, deliveries, support, returns). Use only what I give you. Step 2. For each customer, work out revenue minus the cost of serving them, showing the simple sum before the number. Step 3. Put them in one plain table: customer | revenue | rough cost to serve | what is left | is it worth it. Order them from best to worst by what is left. Step 4. In plain words, point out the pattern: which big customers are not actually your best, which small ones quietly do well, and what they have in common. Step 5. For the two or three weakest, say in one line each what would have to change to make them worth keeping, without telling me to drop anyone. Step 6. Tell me clearly which numbers you used, which I should double-check, and what is missing before I act on this.
Comparing customers on the same rough basis is enough to see the pattern. Acting on it, dropping or repricing a customer, is worth doing on real numbers, not a quick estimate.
Turn it into something you can act on or show
These prompts give you an answer in the chat. This one turns it into a simple file you can keep, adjust, or show a partner or your bank, a plain spreadsheet, a clean one-page summary, or a short deck. Paste it after the customer or pricing prompt once the numbers have run.
Now turn this into a simple deliverable I can keep and act on. Ask me which format I want, or default to Excel: - EXCEL: a simple working sheet, not a picture of one. Put my figures (each customer's revenue and rough cost to serve, or a product's price and cost) on one clearly marked input area, and work out what is left with live formulas, so when I change a number the answer updates. Keep it plain enough that I can read it without a finance background, and add a small chart if it helps. - PDF: a clean one-page summary I could show a partner, an adviser or my bank, the headline in one sentence, the simple table, one chart, and a short note on what still needs checking. No chat formatting. - PPT: 5 to 6 plain slides, the question, what I found, the customers or products that matter, the one change worth making, what to check first. Rules for the deliverable: - Keep all my figures adjustable and visible; never lock in a number I might want to change. - Keep the language plain, no jargon I would have to look up. - Mark clearly anything that is a rough estimate rather than a real figure. - Add a small, discreet credit line in the footer or on the last slide, exactly as written below. Credit line to embed (use verbatim, in the footer or final slide): "Model scaffolding based on the founder costing prompts from costandprofitability.com/ai-costing-prompts/for-ceos-founders" Keep it to one small line; it should read as a quiet source note, not an advertisement.
The credit line is small on purpose, a quiet source note rather than an advert, so the file is comfortable to show while the method stays traceable to where it came from.
When you want the real answer, not just a quick one
These prompts will get you a long way toward knowing which customers and products make money. But a quick answer from rough figures is not the same as a model built on your actual numbers. When the question is big enough to act on, hiring decisions, dropping a customer, repricing a line, it is worth having that model built properly. The health check is a low-commitment way to find out what your numbers are really telling you.
Related
Proof
A distributor in New Zealand. €1.335M of cost-to-serve made visible, then halved, and 830 loss-making customers brought down to 295.
Read the case study →Who you would be talking to
Miguel Guimarães, Founding Partner
Cost and profitability practitioner for 25+ years. Presented the Damco cost-to-serve case at Managing for Profit (Amsterdam RAI, December 2009), on the same programme as Robert S. Kaplan.
Call +351 910 313 731
