How to use AI to create a pricing strategy for your…

Most service business owners set prices by guessing, copying competitors, or charging what they charged three years ago. If you want to use AI to price your services for your small business properly, this post walks you through a repeatable process: calculating your actual floor price, benchmarking against your market, and testing pricing scenarios before you commit. The process works because it forces AI to reason through your numbers sequentially — costs first, margin second, market comparison third.

What you need before you start

ChatGPT (GPT-4o) or Claude (3.5 Sonnet): Either model handles the calculations in this guide. Both require an account. ChatGPT Plus costs $20/month; Claude Pro costs $20/month. Both have free tiers with usage limits — the free tier may work for a single session but can cut out mid-analysis. Check current pricing, as plans change.

A list of your business expenses: Monthly fixed costs, variable costs per project, and your target hours worked per month. A spreadsheet is ideal. A rough list on a notes app also works.

Local competitor pricing data: AI does not have access to your competitor's rates. You need to gather this yourself. More on that in Step 4.

Time required: 2–4 hours for a full first pass. Faster once you have your expense data organized.

Skill level: No coding required. You need to be comfortable copying and pasting text into a chat window.

Gather and sanitize your cost data before you touch the AI

This step happens before you open any AI tool. Your financial data contains information you do not want sitting in a chat interface.

Open your accounting software — QuickBooks, Wave, a spreadsheet, whatever you use — and export or manually list your monthly expenses.
Categorize every expense into three columns: Fixed Costs (rent, software subscriptions, insurance, phone), Variable Costs (materials, subcontractors, delivery fees per project), and Time Costs (hours you spend on admin, sales, and invoicing that you do not bill to clients).
Remove all sensitive identifiers before you copy anything. Delete bank account numbers, client names, tax IDs, and your business's full legal name. Use placeholders: "Client A," "Business Name," "Account ending 4422" becomes nothing at all.

You should end up with a clean expense list that tells the story of your costs without identifying you or anyone you work with.
Add your target: how many billable hours or completed projects you want to deliver per month at full capacity. Write this as a single number. It anchors every calculation that follows.

Data privacy is not an abstract concern here. OpenAI's enterprise privacy standards give you the technical picture. The practical rule: if a number identifies a person or an account, remove it.

Use AI to calculate your floor price (and see the math)

Your floor price is the minimum you can charge without losing money. AI calculates this accurately when you give it the right structure. The technique is called Chain of Thought prompting — you ask the model to show each calculation step before moving to the next.

Open ChatGPT or Claude and start a new conversation.
Paste your sanitized expense data directly into the chat, followed by this prompt:
I am calculating the floor price for my services. Work through this sequentially and show each calculation step.
- Step 1: Add up my total monthly fixed costs from the data below.
- Step 2: Estimate my total monthly variable costs assuming I complete [X] projects per month.
- Step 3: Calculate the total cost of my non-billable time using an assumed hourly rate of $[your target hourly rate].
- Step 4: Add all three together to get my total monthly cost.
- Step 5: Divide by [X] projects to get my cost per project.
- Step 6: Add a [20%] profit margin to get my floor price per project.
[Paste your sanitized expense data here]
You should see the model work through each step with visible arithmetic, ending with a floor price per project or per hour.
Review each step for errors. AI makes arithmetic mistakes on large number sets. Check the addition in Step 1 yourself.
Adjust the profit margin in the prompt if 20% does not match your business goals. Re-run the prompt with the new number.

The floor price is not your final price. It is the number below which you stop making money. Research from Inc. on cost-plus pricing confirms what the math shows: most service businesses underprice because they forget to include the cost of non-billable time. Step 3 in the prompt above fixes that.

Benchmark your floor price against your local market

AI cannot tell you what your competitors charge. You have to bring that data to the conversation.

Research three to five competitors in your service area. Use their websites, Google Business profiles, Thumbtack, Houzz, LinkedIn, or any platform where rates appear publicly. Write down the prices or ranges you find.
Return to the chat and add this prompt:

Here is the local market pricing data I collected. Compare my floor price of $[X] against this market data. Tell me where I fall in the range — below market, at market, or above market. Then suggest what a market-rate price and a premium price would be for my service, based on this data.

[Paste your competitor pricing notes here]

You should see a clear comparison showing your floor price relative to market low, mid, and high ranges.
Note the gap between your floor price and the market rate. A 15–25% gap between your current rate and a defensible market rate is common. That gap is not an accident — it typically comes from hidden overhead costs that owners never counted.

Run what-if scenarios to stress-test your pricing

This is where an AI pricing strategy for your service business earns its place in your process. Sensitivity analysis — changing one variable and seeing how the outcome shifts — takes minutes with AI and hours on a spreadsheet.

Ask the model to run a client loss scenario:

If I increase my price by 10%, how many clients can I afford to lose before my total monthly profit stays the same as it is now at my current price and current client volume? Show your calculation.

You should see a specific number. That number tells you how much pricing room you have before volume loss cancels out the revenue gain.
Run a second scenario for a price decrease or a package offer:

If I offer a three-project bundle at a 15% discount, what is the minimum number of bundles I need to sell per month to match my current monthly profit?
Test one more variable that is specific to your business — a cost increase, a new equipment expense, a planned hire. Frame it the same way: one change, one outcome, show the math.

AI handles this class of problem well. It is arithmetic with conditions, and models like GPT-4o and Claude 3.5 Sonnet do not get bored running the same structure ten times with different inputs.

When something goes wrong

The AI gives you a floor price that looks obviously wrong (too low or too high). The model likely misread your expense data format or skipped a category. Paste your expense list again in a cleaner format — one item per line, with a dollar amount after each — and re-run the Step 1 calculation in isolation before continuing.

The model stops mid-calculation in the free tier. Free accounts on ChatGPT and Claude have output limits. If the response cuts off, type "continue" to resume. For a full pricing analysis session, a paid plan reduces interruptions.

Your competitor pricing data produces a nonsensical comparison. This happens when the data mixes hourly rates and project rates without labeling them. Go back to your competitor notes and label each figure as "per hour" or "per project" before pasting. Ask the model to normalize everything to one unit before comparing.

What to do next

Take your new floor price and one market-rate comparison figure and update your next client proposal using those numbers. One proposal is enough to test whether the pricing holds in a real conversation.

If you want to build this analysis into a repeatable monthly process, read how to automate your financial reporting workflow with AI tools.

FAQ

Am I undercharging? How can AI tell me? AI can identify undercharging by comparing your calculated cost-per-project against your current rate. If your floor price — costs plus a reasonable margin — is higher than what you currently charge, you are losing money on each project. Feed your current rate into the prompt in Step 3 and ask the model to flag the gap explicitly.

Is it safe to paste my financial data into ChatGPT? The risk is manageable if you sanitize first. Remove identifying information: account numbers, client names, tax IDs, and your business's legal name. What remains is a list of numbers and categories, which carries minimal risk. For a higher privacy standard, use the ChatGPT Enterprise plan, which excludes your data from model training by default.

What if I charge hourly instead of per project? The process is the same. Replace "per project" with "per billable hour" throughout. In Step 4 of the prompt, divide your total monthly cost by your target billable hours per month instead of by project count. You get a floor rate per hour instead of a floor price per project.

Does AI account for taxes in the pricing calculation? Not unless you include tax obligations in your expense data. Add a line item for your estimated self-employment tax or corporate tax set-aside — typically 25–30% of net profit — before you run the calculation. Otherwise, your floor price will not cover your tax burden.

How often should I redo this analysis? Run it once when you set up the system, then re-run it whenever your costs change materially: a new software subscription, a rate increase from a supplier, a hire, or a significant shift in your project volume. Once a year is the minimum for a stable business.