A practical framework for pricing AI services and info products based on value delivered, not hours spent.
I spent my first six months pricing AI work wrong.
Not slightly wrong. Fundamentally wrong. I was pricing based on effort: how many hours the work took, how much compute I burned, how complex the code was. The problem with effort-based pricing for AI is that a system that saves a client 20 hours per week does not become less valuable because you built it in three days.
After two years of running AI businesses across two countries, I have a pricing framework that works. I am going to share the mental model and the actual numbers, because I wish someone had done the same for me when I started.
Traditional consulting prices on time. You charge per hour, per day, or per project with the hours baked into the estimate. The client is really buying your time.
AI work breaks this model for a specific reason: the relationship between effort and value is nonlinear. A well-designed AI agent might take 20 hours to build but save 200 hours per month. A poorly scoped AI project might take 200 hours and save nothing.
If you price on time, you get punished for being good. The faster and more efficiently you build, the less you earn. This is backwards.
The alternative is value-based pricing. You price based on what the outcome is worth to the client, not what it costs you to deliver. This sounds straightforward in theory. In practice, most builders struggle with two questions: how do I calculate the value, and how do I justify the price to someone who thinks in hourly rates?
The exact architecture, memory layers, and delegation patterns I use to run 50 agents across two businesses.
Get the AI Agent Blueprint →My pricing framework uses four variables. Every AI engagement can be evaluated through these.
Hours saved per month. This is the most concrete value metric. If the current process takes 40 hours per month of human labor and the AI system reduces that to 5 hours, the savings are 35 hours per month. Multiply by the loaded cost of those hours (salary plus benefits plus overhead divided by monthly hours) and you have a monthly dollar value.
For a $60,000/year employee spending 35 hours per month on a task, the loaded cost is roughly $35/hour. That is $1,225/month in direct labor savings. Over a year: $14,700.
Error reduction. Harder to quantify but often more valuable than time savings. If a manual process has a 5% error rate and each error costs $500 in rework or lost revenue, that is $2,500 per month for a process that runs 100 times. If AI reduces the error rate to 0.5%, the monthly value of error reduction is $2,250.
Clients rarely track their error rates. Part of the discovery process is helping them calculate this number. Once they see it, the price conversation changes.
Speed to outcome. Some tasks are not about saving hours. They are about getting results faster. A research report that takes a week to compile manually but two hours with AI does not save 38 hours (the researcher was doing other things during that week). It delivers the insight five days earlier. If that earlier insight affects a decision worth $50,000, the speed premium is significant.
Capacity creation. The most overlooked variable. If your AI system lets a three-person team handle the workload of a ten-person team, the value is not just the salary of seven people. It is the ability to grow without proportional headcount increases. For a growing business, capacity creation is worth more than cost reduction.
Here is how I calculate a price for any AI engagement.
Step 1: Quantify the monthly value. Add up the four variables. Hours saved multiplied by loaded hourly cost, plus error reduction value, plus speed premium (if applicable), plus capacity creation value. Call this the Monthly Value Delivered.
For most small to mid-size engagements, the Monthly Value Delivered falls between $2,000 and $15,000. For enterprise engagements, it can be much higher.
Step 2: Apply the value capture ratio. You do not charge 100% of the value you deliver. That would leave the client with zero ROI and no reason to buy. The standard range I use is 10% to 30% of annual value for the build, plus 5% to 15% of annual value for ongoing maintenance.
If the Monthly Value Delivered is $5,000, the Annual Value is $60,000. At a 20% value capture ratio, the build price is $12,000. At a 10% maintenance ratio, ongoing support is $6,000/year or $500/month.
Step 3: Set a floor. No matter what the value calculation produces, I have minimum prices. For a custom AI agent build: $3,000. For a multi-agent system: $8,000. For ongoing support: $500/month. These minimums exist because any engagement below them does not justify the overhead of client communication, onboarding, and support.
Step 4: Reality check against market rates. The value-based price should fall within a reasonable range of what the market expects. If the calculator produces a $50,000 price for a simple chatbot, something is off. Either the value was overestimated or the solution is more complex than a simple chatbot.
The framework above applies to service engagements. For info products (courses, templates, guides), the math is different because you sell once and deliver infinitely.
I use a three-tier structure.
Entry tier ($17 to $47). A specific tool or template that solves one problem. An AI pricing spreadsheet. A prompt library for a specific use case. An agent configuration template. The buyer gets immediate value and starts to trust your expertise.
Core tier ($97 to $297). A complete system or course that teaches a methodology. How to build a multi-agent system from scratch. How to price and sell AI services. The buyer invests real money and gets a real framework they can apply repeatedly.
Premium tier ($497 to $1,997). A comprehensive program with direct access, community, or done-with-you components. A cohort-based course with live sessions. A template pack plus monthly group calls. The buyer is paying for proximity and accountability, not just information.
The key insight for info product pricing: price based on the value of the outcome, not the volume of the content. A 10-page guide that saves someone $5,000 in pricing mistakes is worth more than a 200-page course that teaches theory without application.
“Why does this cost more than [competing service]?” Because I am not selling the same thing. A cheaper provider prices on time and delivers a technical output. I price on value and deliver a business outcome. If the competing service delivers the same outcome, the client should hire them.
“Can you give us an hourly rate?” I can, but it will cost you more. An hourly rate incentivizes me to work slowly. A project rate incentivizes me to work efficiently. Pick which incentive you want your vendor to have.
“We need to see a detailed time breakdown.” I provide a detailed scope breakdown, not a time breakdown. Every deliverable is specified: what it does, what it replaces, what value it creates. How long it takes me to build is my business decision, not the client’s cost driver.
“That is more than we budgeted.” What did you budget based on? If the budget was based on hourly rates from a different type of work, it will not align with value-based pricing. The relevant question is whether the ROI justifies the spend, not whether it fits a pre-existing budget that was calculated differently.
Pricing too low to win the deal. Underpricing does not build trust. It signals that you do not believe in the value of your own work. Two of my earliest clients later told me that my low price almost made them choose a competitor because they assumed the quality would be proportional.
Not quantifying value in the proposal. Early proposals described features: “The system will do X, Y, and Z.” Later proposals describe value: “X will save 15 hours/month ($525/mo), Y will reduce errors by 80% ($1,200/mo), Z will enable same-day turnaround (currently 3-day).” The second format closes at a higher rate because the client can do the ROI math themselves.
Bundling maintenance into the build price. For the first year, I included three months of free support in every project. This trained clients to expect free support. Now, support is a separate line item from day one. Clients who value ongoing reliability pay for it. Clients who do not need it save money. Both are happy.
The calculator spreadsheet, complete with the formulas and example scenarios described here, is available as part of the blueprint.
Pricing is a skill, not a talent. You get better at it by studying the math, testing different approaches, and tracking what works. Two years in, I am still adjusting. But the framework holds, and it consistently produces prices that are fair for the client and sustainable for me.
Subscribe to The Signal
One issue per week. The biggest AI build of the week, the top five videos worth your time, and an operator's read on what actually matters. For builders who ship, not dreamers who scroll.
No spam. Unsubscribe anytime. Just The Signal.