Commodity vs. Value-Based Pricing: AI Just Erased One Side of the Line
An agency quote lands in your inbox. Sixty thousand dollars for a first version, four months to ship, the same shape of estimate you would have read in 2022. Except last month you watched a friend build something close to it over a weekend with Cursor, for the price of a nice dinner. The number on the quote did not move. The thing it is charging you for did.
Or you are standing on the other side of the table. You are about to put a price on the product you have spent six months building, and the hourly rate you had in mind is now lined up against a twenty-dollar-a-month subscription that does the same task in seconds. Your number suddenly looks insane, and you cannot tell whether the problem is your price or your product.
Both founders are feeling the same thing before they have words for it. A line moved. Here are the words.
Most founders treat pricing as a number to set. It is a category to choose, and there are only two. AI just erased the floor under one of them.
What is the difference between commodity and value-based pricing?
Commodity pricing charges for the work, indexed to inputs: hours, deliverables, units of output. Value-based pricing charges for the result, indexed to what the outcome is worth to the buyer, with no fixed tie to what it costs to produce. The work can be identical. What the price is anchored to is the whole game.
Picture both sides with real examples.
On the commodity side: a dev agency at $150 an hour. A freelance copywriter at fifty cents a word. A developer who quotes per feature. A designer who bills per page. In every case the price tracks the input. Double the hours, double the bill. The customer is buying units of effort.
On the value side: Stripe takes a percentage of the payment volume it processes, not a fee per API call indexed to its compute cost. A fractional CTO is retained monthly for the result that the technical side is handled, not for a timesheet. A studio ships a fixed feasibility decision for a fixed fee, regardless of how many hours the decision took. The price tracks what the result is worth to the buyer.
Hermann Simon, who spent a career on this question and wrote Confessions of the Pricing Man, puts the distinction at the center of pricing strategy: a price either reflects your cost or it reflects the customer's value, and the two lead to completely different businesses. Madhavan Ramanujam and Georg Tacke push it further in Monetizing Innovation. Willingness to pay is a design input, something you discover before you build, not a number you reverse-engineer from your costs after the fact.
Most founders drift to the commodity side without ever choosing it, because it is legible. An hour is a number you can defend. A deliverable is a thing you can point at. Value is harder to name and harder to charge for, so the safe move is to price what you can count. That instinct was survivable when the work itself was scarce. It is not anymore.
Hold that frame, because the rest of this only makes sense once you can locate yourself on it. Until 2024, choosing between the two models was a margin decision. Now it decides whether you have a business at all.
Why is AI collapsing the price of commodity work?
Because when a model can do the work, the price of the work falls toward the cost of running the model, and that cost is approaching zero. A thin margin sits on top. That is the whole mechanism, and it does not care which industry it is eating.
Start where the data is hardest. Cursor crossed $2 billion in annual revenue by February 2026, the fastest climb to that mark in B2B software history. Lovable reported $200 million in annual recurring revenue by November 2025. By early 2026, 51 percent of all code committed to GitHub was AI-generated or AI-assisted, and 92 percent of US developers reported using AI tools daily. Baytech Consulting estimates an 80 percent reduction in the initial capital needed for a typical software build. We laid out what this did to the cost of one MVP in an earlier post on the pre-seed bar. The build is no longer where the money is.
The same wave is breaking over every service whose unit was an hour. Forrester reports that advertising agency headcount fell 8 percent in 2025 and forecasts a 15 percent cut in 2026, driven by AI and automation (Forrester, Predictions 2026). Gartner's 2025 CMO survey found 39 percent of marketing chiefs planning to cut agency budgets, and 22 percent saying generative AI had let them pull creative and strategy work back in-house, away from agencies. Goldman Sachs estimated in 2023 that 300 million jobs globally were exposed to some AI automation, with 44 percent of legal tasks and 46 percent of administrative tasks among the most exposed. Copy, design, first-pass legal, tier-one support, basic analysis: the model does it, and the price of doing it falls toward the model's running cost.
The freelance market shows the whole split in two numbers. Writing projects on Upwork fell 32 percent year over year in 2025, the steepest drop of any category, in an analysis of 2.2 million postings (Vollna, 2025). In the same market, freelancers doing AI-related work earned 44 percent more per hour than everyone else (Upwork, 2025). The floor and the ceiling moved in opposite directions at the same time. Generic writing is commodity, and it collapsed. Specialist work that uses the models instead of racing them is value, and it got more expensive.
People are losing work as this plays out. That is real, and pretending otherwise would be dishonest. The point here is narrower than the disruption itself. It is about where the money goes when the work gets cheap, so you can decide which side of that move your own business sits on.
Where does the profit go when the work gets cheap?
It moves to the adjacent layer that did not commoditize. This is the part most pricing advice misses, and it is the most useful thing in this post. The money does not vanish when a layer gets cheap. It migrates.
Clayton Christensen named the pattern the law of conservation of attractive profits, in The Innovator's Solution (Christensen and Raynor, 2003). When commoditization strips the profit out of one stage of a value chain, the opportunity to earn attractive profit reappears at an adjacent stage, usually the one that is now the bottleneck on performance. Profit is conserved. It just changes address.
The history is clean. Personal computers commoditized into interchangeable boxes, and the profit moved up to the operating system, where Microsoft sat, and down to the chip, where Intel sat. Smartphone hardware commoditized, and the profit moved to the app stores and the platforms running on top. Servers commoditized into rented compute, and the profit climbed to the platforms and software sitting on the cloud above them. Each time, the layer that got cheap stopped being where you wanted to stand, and the layer next to it got rich.
Coding is the layer commoditizing now. So the profit is migrating up, to the layer AI does not touch: product judgment, system integration, accountability for whether the thing actually works, trust and regulatory weight, the customer relationship. Deciding what to build and owning the outcome was always worth more than typing the code. For twenty years that gap was hidden, because the typing was expensive enough to feel like the real work. AI pulled the cover off.
The supply of that top layer has not budged, and that is the other half of the mechanism. You can produce a thousand more units of code generation by renting more compute. You cannot produce a thousand more units of judgment, or accountability, or a real relationship with a customer, by renting anything. Cheap and expanding supply on one layer, fixed supply on the next: that is the whole reason the price splits the way it does.
You can watch the migration happen in the agency numbers. Forrester describes agencies being forced off the hourly model and toward selling outcomes, because clients will no longer pay input prices for work a model now does. That is not a trend report. That is profit relocating in real time, from the hours to the result, exactly where Christensen said it would go.
Should you still pay a dev agency now that AI can code?
Only for the part AI does not do. Before you sign anything, read the quote and ask one question: is this price anchored to hours and deliverables, or to a defined outcome you will own either way?
If it is anchored to hours, you are about to pay full retail for an arbitrage against AI. The agency's cost structure is built around timesheets, so it bills the commodity way, and the commodity is now cheaper than the cheapest agency on earth. There is no version of that math where the hourly quote is your cheapest path to working software.
The only reason left to pay an agency, a freelancer, or a studio anything is the part the model cannot deliver. The judgment about what to build. The accountability when it breaks. The decision, made by someone who has been in the room before, about which of your assumptions is the one that will sink you. That is the thing whose price has held, because its supply has not changed. So find it on the invoice. If the quote is hours and deliverables with no outcome attached, you are buying the part AI does for free, at 2022 prices. We broke down the specific buying decision between an agency, a CTO, and a studio in a separate post, and the whole frame there reduces to this: a studio sells judgment, an agency sells hours, and only one of those still commands a price.
Read the line items the way you would read a contract you expect to fight over. A line that says forty hours of frontend work is a commodity you can now buy from a model for almost nothing. A line that says we decide what ships and we carry it if it breaks is the thing you are actually short on. The first should make you wince at the number. The second is the only part worth paying full price for.
How should you price your own product in the AI era?
Match the price to the unit of value, not the unit of input. This is the same logic, flipped to face you as a seller. Alex Hormozi makes the point in $100M Offers: the customer pays for the offer, the promised result, not the raw features that produce it. Price the result.
The companies that get this never charge for their inputs. Stripe does not bill per API call indexed to compute. It takes a percentage of payment volume, because the value to the merchant is the money Stripe helps move. Linear charges per seat, because the unit of value is one engineer getting work done inside it. Webflow charges per published site, because the unit of value is a business going live. None of these prices has anything to do with the vendor's cost. Each is anchored to the customer's outcome.
The trap on this side is cost-plus: add up your time, mark it up, quote the total. It feels prudent, and it is the precise move that prices you against the model, because your cost is now the model's cost, and the model's cost is falling every quarter. Anchor to the result instead, and your price stops tracking a number sliding toward zero.
Here is the exercise, and it takes one evening. Write down what your customer is actually paying you to make happen, in their language, not yours. Not the features. The thing in their life that changes when your product works. Then build your price against that sentence. If you cannot finish the sentence, that is the real problem, and no pricing model will paper over it. Pricing against your inputs in 2026 is pricing yourself directly against AI, and that is a race to the model's running cost, which is to say a race to zero.
Which side of the line is your business on?
Answer it honestly, because the market will answer it for you if you do not. There is one question underneath everything above: which side of the commodity-value line am I on, and is my pricing model telling the truth about it?
If you price hours and dress it up as value, the market corrects you, because the buyer can now get the hours from a model. If you buy hours and tell yourself you are buying value, AI corrects you, because you are overpaying for the thing it does for free. Both corrections are running right now, on the same founders, often in the same week. Most are still pricing and buying as if the line sat where it did in 2022. It does not. It moved, and it kept moving.
We feel this inside our own engagements constantly. A founder will describe the agency quote on their desk and the price they are about to set on their own product in the same call, never noticing that both decisions are the same decision wearing two costumes. One is them overpaying for a commodity. The other is them underpricing a result. Same line, both sides, same person.
The reframe is not a stylistic preference anymore. It is the structural move that decides whether your business has margin twelve months from now. Choose the appreciating side on both sides of the table, or watch the depreciating side choose for you.
Waiting is not neutral here. The corrections compound monthly, because the models improve monthly, and every release pushes more work across the line from value to commodity. The price of a commodity quote you accept today will look worse next quarter, and the product you priced against your inputs will look more exposed with every model that ships. Sitting still is a vote for the side that is losing value, cast quietly, by default.
A founder who reads this and changes nothing has decided to compete on the half of the market that is on its way to zero, while the other half gets more expensive every month the models improve.
If you are about to spend money with an agency, or set a price on your own product, that single choice between commodity and value decides which side of the AI line your business sits on for the next year. We do feasibility and strategy work for early-stage founders, and the read is usually about exactly this. Talk to us before you sign the quote or set the number. Last verified: June 2026.