Pricing Your Work With AI: What ChatGPT and Claude Can (and Can't) Tell You
An honest look at what generic AI pricing advice gets right and where it stops, four prompts that sharpen an actual pricing decision today, and what changes when the AI applies a documented value-based fees framework instead of generic rate-setting advice.
Ask AI what to charge for a project and the default answer leans hourly or day-rate, because that's the dominant pattern in most freelance and consulting advice it learned from. It's a reasonable starting point and also, for a lot of skilled work, the wrong frame entirely: it prices your time instead of the client's outcome, and it caps your income at the number of hours you're willing to sell.
The more useful question isn't "what should I charge." It's "what is this actually worth to the client, and how do I price against that instead of against my calendar."
What's wrong with the price AI gives you when you just ask?
Describe your service to an AI and ask for a price, and you'll typically get a day-rate or hourly range pulled from general market data, "consultants with your experience level typically charge $X to $Y per hour." It's not wrong as a market anchor. It's also structurally the wrong question for a lot of skilled work, because it prices your input (time) rather than the client's output (the value the work actually creates for them), which is the entire distinction the value-pricing literature has been making for forty years.
The advice isn't bad. It's answering "what's the going rate for my time" when the more useful question is "what is this worth to them," and those two questions can produce wildly different numbers for identical work.
What is AI actually good for in a pricing decision, and where should you stop trusting it?
Unaided, it's genuinely good at generating a market-rate anchor fast, useful context even when you plan to price differently, and at helping you articulate the client's value case in their own terms once you describe the actual outcome your work produces for them.
It's weak at anything involving this specific client's actual budget reality, their alternative options, how urgently they need the problem solved, or what a competitor already quoted them. None of that is knowable from a generic prompt; it's only as good as what you feed it, and most people don't feed it enough specificity to move past a generic market-rate answer.
What four prompts actually sharpen a pricing decision?
All four work in Claude, ChatGPT, or Gemini exactly as written below, nothing extra to configure.
1. The value translation. "I'm pricing [project] for a client. Here's what the work actually produces for them: [describe the outcome, revenue impact, time saved, risk avoided]. Translate that into a dollar estimate of value to them, not a price for my time." Why it works: most sellers can describe what they do but struggle to translate it into the client's own financial terms; forcing the translation surfaces a number that's usually far higher than an hourly calculation would produce.
2. The range generator. "Given a value estimate of roughly [$X] to the client, and market rates for comparable work in [industry], suggest three pricing tiers, a conservative, a target, and an ambitious number, with the reasoning for each." Why it works: a single number invites a simple yes-or-no; three tiers turn the conversation into a choice, and the anchoring effect of the ambitious tier tends to make the target tier feel more reasonable by comparison.
3. The objection pre-mortem. "List the five most likely objections to this price, ranked by how likely each is, and a response to each that doesn't immediately discount." Why it works: most pricing conversations fail not because the price was wrong but because the seller wasn't ready for the specific pushback and reflexively discounted under pressure; rehearsing the objection in advance keeps the number intact.
4. The walk-away check. "If the client pushed back hard and asked for 30% off, what would I actually lose by holding the price, and what would I lose by discounting? Be specific." Why it works: this forces an honest look at whether your price is actually defensible or whether you'd cave under mild pressure, which tells you more about your own confidence in the number than any market-rate lookup does.
What's missing even after all four prompts?
These four prompts leave you genuinely better prepared for this one quote than a generic ask would. They don't leave you with a pricing philosophy that holds across every future client and project, one that resists quietly reverting to hourly thinking the next time someone asks "so what's your rate." Fixing this quote isn't the same as fixing how you price, on a standing basis, going forward.
What's actually different once a documented pricing framework replaces generic rate-setting advice?
Alan Weiss has published his value-based fees methodology across decades of books and case studies, most directly in Value-Based Fees and Million Dollar Consulting: establish the client's value first, in specific, client-stated terms, before any number is discussed, then price as a fraction of that value rather than a markup on hours, and refuse engagements where a genuine value case can't be established at all, because Weiss's documented position is that if you can't articulate the value, you shouldn't be pricing the engagement yet.
Before, generic prompting: asked to price a six-week marketing audit, a generic AI response estimates hours (roughly 80 to 100) times a market day-rate, landing on a number derived entirely from effort, with no reference to what the audit is actually worth to the client if it works.
After, Weiss's framework applied: the framework starts by refusing to discuss hours at all. It asks first what the audit will let the client do that they can't do now, redirect a specific amount of wasted ad spend, say, and asks for the client's own estimate of that waste. Only once that figure exists does a price get proposed, typically a defined fraction of the value created, explicitly decoupled from how many hours the work actually takes. If the client can't or won't name a value figure, the framework's documented advice is to treat that as a signal to walk away or restructure the engagement, not to fall back to an hourly rate as a default.
A generic pricing prompt has no reason to enforce that refusal, pricing against value instead of time, unless something is specifically making it do so. The framework does, as a fixed discipline rather than a one-off nudge toward a bigger number.
How does a framework like this actually get loaded into Claude, ChatGPT, or Gemini?
Claude packages it as a Skill: upload the .zip from Settings → Skills → Add skill, and Claude auto-invokes it once your question matches the skill's description; leading a message with /alan-weiss-framework overrides that judgment call and applies it regardless. From Claude Code, the same package goes into ~/.claude/skills for use across every project, or into a single project's .claude/skills folder if that's all you need.
ChatGPT has no direct equivalent, so the plain .md file becomes a Custom GPT's Instructions: Create a GPT → Configure → Instructions, paste the content straight in. Don't reach for the default Custom Instructions fields here, they cap at 1,500 characters; a Custom GPT's Instructions field holds close to 8,000, which is what a full framework actually needs.
Gemini offers no upload mechanism at all for this, so the content goes into the system prompt directly if your interface supports one, or into the opening message of the chat: "Here is a thinking framework to apply throughout this conversation. Read it carefully, then answer using this framework's approach," with the file's text pasted underneath.
What comes after the number?
Weiss's documented method, alongside David Maister's and Blair Enns's, is available in the Consultant & Advisor category as .md files for Claude, ChatGPT, or Gemini, the better route if pricing well is an ongoing problem rather than a one-off. For the proposal already sitting in your drafts, Pricing Decision is a $99 tool built to structure that specific call before the quote goes out.
Frequently asked questions
Can AI actually tell me what to charge?
It can give you a plausible market range if you describe your service and market clearly, which is a genuinely useful starting anchor. It cannot tell you what a specific client will actually pay, because that depends on the value they place on the outcome, information the AI doesn't have access to unless you supply it. Treat any number it gives you as a starting range to test, not a verdict to accept.
Why does AI keep suggesting hourly or day-rate pricing when I ask about this?
Because hourly and day-rate pricing is the dominant pattern in the training data it learned from, most freelance and consulting advice online defaults to it, so it's the statistically likely answer to a generic pricing question. Getting past this requires explicitly ruling it out in your prompt, asking for value-based or outcome-based alternatives specifically, rather than hoping the AI volunteers them unprompted.
Is it safe to send my actual client budget or numbers to an AI for pricing help?
For working through the reasoning, yes, most people find this low-risk, treat it the way you'd treat notes to yourself. If you're using an enterprise or team account with data-retention or compliance requirements around client financial information, check your organization's policy first, the same way you would before pasting any client-confidential number into any external tool, AI or otherwise, since the risk here is organizational policy, not the AI itself.
What's the biggest mistake people make asking AI for pricing help?
Asking it to justify a number they've already quietly decided on, rather than asking it to help them find the number. The tell is a prompt like "is $5,000 a fair price for this project," which invites a validating answer instead of an independent one. A better prompt describes the value delivered and asks the AI to reason toward a range from there, unanchored by a number you've already committed to emotionally.
Does value-based pricing actually work, or is it consultant folklore?
It's documented, not folklore, though it isn't universal. Alan Weiss has published his value-based methodology and case results for decades, and the underlying logic, price against the client's value received rather than your time spent, holds up across consulting, agency, and freelance work generally. It works best when the outcome is genuinely measurable or high-stakes for the client; it works less cleanly for commodity work where the client has many equivalent substitutes.
How is a documented pricing framework different from just asking AI what to charge?
A framework fixes the sequence: define the client's value first, in their terms, before any number is discussed, then set a price as a fraction of that value rather than a markup on your time, and refuse engagements where the value can't be established at all. A one-off prompt gives you a number for this project. The framework gives you a repeatable method for every pricing conversation you'll have going forward, run the same way each time.
Won't clients push back harder on value-based pricing than on an hourly rate?
Often initially, yes, and the documented response to that pushback is part of the method, not an afterthought. Weiss's own published position is that a client questioning value-based pricing is usually questioning whether the value case was made clearly enough, not objecting to the pricing model itself, which is why the framework spends real effort on articulating value in the client's own terms before the number ever comes up.
What's the honest limit of using AI for a pricing decision?
It doesn't know this specific client's budget reality, their alternative options, or how badly they need this solved right now, all of which materially affect what they'll actually pay. Use it to reason through your value case and generate a defensible range; the final number still depends on read-the-room judgment about this particular client that no general-purpose prompt can supply for you.
Written by Gareth Hoyle. Last updated 26 September 2026. Part of the authority.md guides library.
More guides.
Bringing AI Into a Creative Project Without Ruining It
What generic AI brainstorming actually gives a creative project and what it quietly flattens, four prompts that help today without taking over, and what changes when the AI applies Rick Rubin's documented listening practice instead of generic idea generation.
AI for Hiring Decisions: Where Claude and ChatGPT Actually Help
A grounded look at what generic AI use gets right and wrong when you're deciding on a hire, four prompts that genuinely sharpen the decision, and what changes when the AI applies a documented evidence-based hiring framework instead of generic interview advice.
Using AI to Prep for a Negotiation: What Actually Works
Generic AI negotiation advice is genuinely useful and genuinely limited. Here's what Claude and ChatGPT get right before a negotiation, four prompts that work today, and what changes once the AI has a documented framework instead of a generic one.