This is The Agentic Commercial Model, Season 2: the build. Issue 03 built the instrument. This is what you do with it.
There is a contradiction sitting in the middle of this market that nobody wants to resolve.
A Cruxy survey of 300 SaaS chief executives in April found 97% plan to retire seat based pricing within two years. The same survey found 94% believe seat-based pricing currently aligns with the value their product delivers.
Both of those cannot be true. Either the seat reflects your value, in which case why abandon it, or it does not, in which case it has not aligned with anything for some time and you have been telling yourself a story.
What the contradiction actually reveals is that almost nobody is designing pricing. They are layering it. A model the industry has already declared obsolete stays exactly where it is, a credit meter gets bolted on top, and the company calls the result an AI pricing strategy. The base was never revisited. It was decorated.
The PricingSaaS 500 Index logged over 1,800 pricing and packaging changes across the top 500 B2B and AI companies in a single year, an average of 3.6 per company. That is not iteration. That is an industry metering and guessing it lands, lazy does not win the long game.
Design means starting somewhere deliberate, and there is only one deliberate place to start. The value exchange. What the customer gets, what they give you for it, and whether those two things still track each other in a world where the product does the work. Everything else in your pricing is downstream of that answer, and almost nobody has written it down.
So start there and work backwards.
Layer one: the outcome
Ask what your product does for the customer, stated as something they would recognise as measurable valuable and could describe to their own board. A ticket resolved. An invoice reconciled without error. A brief accepted. A meeting booked that turned into a qualified opportunity. Keep it simple, the value anchor is how you justify spend, not price.
That is your pricing metric. Not a token, not a credit, not a seat. The thing that happened.
The market is moving here, slowly and unevenly. Zendesk charges for successful AI driven resolutions rather than AI seats. Intercom applies the same logic to Fin. Decagon structures its pricing around resolved interactions, and interestingly lets customers choose between per conversation and per resolution, which is a rather elegant way of letting the buyer pick their own risk appetite.
However the honest constraint is measurement, not ambition. Kyle Poyar frames the requirements as consistency of outcomes, attribution, measurability and predictability. Fail any one of those and outcome pricing produces disputes rather than alignment. Which is precisely why Issue 03 came first. You cannot price an outcome you cannot verify, and most companies attempting this are pricing an outcome they have merely asserted.
Get the metric right and everything downstream is arithmetic. Get it wrong and no amount of packaging saves you.
Layer two: the meter
Now work backwards to what it costs you to produce that outcome.
This is the layer traditional SaaS never needed. At 80 to 90% gross margin the marginal cost of delivery was rounding. AI products frequently run at 50 to 60%, and every additional workflow generates inference, retrieval, orchestration and monitoring costs that vary by customer and by day.
So you meter. Not to bill the customer in tokens, which is meaningless to them, but to know your cost per outcome at account level. That number tells you whether the price in layer one is a business or a hobby.
Clay did something worth copying in March, splitting monetisation into two tracks and deliberately separating value, the platform, from cost, the tokens. That separation is the point. The customer buys value. You manage cost. Conflating the two in a single credit balance is how you end up explaining your infrastructure bill to a procurement team who correctly does not care.
Layer three: the floor
Finally, work back to what both finance functions need to plan.
A committed base. Not because it is elegant, but because pure consumption is unforecastable for the buyer and unbankable for you. Enterprise procurement needs a number to approve. Your board needs revenue it can predict. Neither gets that from a pure meter.
This is where the market has landed and the data is unambiguous. Hybrid is the most common primary structure at 37% in Poyar's survey of over 230 software companies, per-seat has collapsed from 21% to 15% in twelve months, and Futurum's 1H 2026 buyer research found 43% of buyers prefer consumption models, 27% prefer outcome-based, and fewer than one in five still prefer classic per-user. Their conclusion is worth quoting directly to anyone still defending a seat-only price list. Vendors restricted to seat-only pricing risk immediate disqualification.
One trap to avoid, because it is common and it is corrosive. Some vendors set overage rates two to three times the included tier rate. It looks clever in the model and it reads as a penalty to the customer, which is exactly what it is. You are charging your best customers most for succeeding with your product. That is not pricing, it is a fine. Build growth tiers, clear unit economics.
The problem nobody solves: your own product getting better
Here is the flaw sitting inside every outcome model, and almost nobody addresses it because addressing it costs money.
Your cost to deliver an outcome falls. Models get cheaper, prompts get tighter, caching improves. Say your cost per resolution falls from fourteen pence to four. Under a fixed per outcome fee, the customer's price is unchanged, their value is unchanged, and you have quietly pocketed the entire gain.
Parloa built a whole argument on this, and they are right about the mechanism. Efficiency created inside the system flows to the vendor by default. Their answer, retreating to pure consumption billing so the customer's bill falls automatically, throws out value based pricing to solve a fairness problem.
There is a better mechanism, and it is not a discount.
Reinvest the efficiency dividend into the value you deliver. Hold the price, and spend what you saved on more use cases, faster roadmap, deeper capability inside the same relationship. The customer does not get a smaller bill. They get materially more for the same money, every year.
Improving the value exchange is not the same as sharing the gain, and the difference matters enormously.
A price step down is a one way transfer. It shrinks your revenue base permanently, it funds nothing, and it is close to impossible to reverse. You have taught the customer to expect an annual reduction and given yourself less capital to build with. Two years of that and you are a cheaper version of the same product, competing with someone who spent the same money on capability.
Reinvestment compounds in both directions. Every efficiency gain becomes new capability, new capability creates new outcomes, and new outcomes are new things to price. The surface area of value grows rather than the invoice shrinking. Your customer's return on the same spend rises continuously, which is the only genuinely future-proof position in a market where the cost to deliver keeps falling.
It also answers the renewal ambush from Issue 04 far better than a discount does. When their audit agent runs the numbers, a falling bill looks like a vendor who was overcharging. A flat bill against visibly expanding value looks like a return that improves every year. One of those reads as an admission. The other reads as a compounding investment.
However this only works if you can prove it, which is where the instrument earns its keep again. Reinvestment that nobody can see is indistinguishable from pocketing the gain. So publish the value delivered per pound spent, track it, and show the curve moving. The commitment is not we will lower your price. It is your return on this spend will improve every year, and here is the evidence that it has.
That is a harder promise to make and a much harder one to fake. It is also the only one that leaves you with the capital to keep making it.
Where the blame sits
Three parties own this mess, in descending order of responsibility.
Vendors first. You have had two years to design something and most of you bolted a credit meter onto a seat price and called it an AI strategy. Multiple concurrent pricing models are not flexibility, they are the visible residue of decisions you avoided making. If your own commercial team cannot explain your pricing in one sentence, or your sellers cannot deliver an elevator pitch based on outcomes and the number, your customer's procurement function certainly cannot, and 3.6 pricing changes a year is not agility, it is an admission.
Buyers next, and this is the uncomfortable half. You have accepted pricing you cannot forecast, agreed outcome definitions you never read, and signed overage rates that punish you for succeeding. Get the resolution criteria in writing. If a vendor's definition of resolved includes cases where your customer followed up, you are paying for their false positives, at scale, monthly.
And boards last, because they approve all of it. Approving a pricing model nobody in the room can defend on unit economics is a governance failure, not a commercial one. The question is not what are we charging. It is what happens to this model when our cost to deliver halves, and if nobody can answer that, the model was never designed. It was assembled.
Start at the value exchange. Work backwards to the outcome, then the meter, then the floor. Reinvest what you save into what they get.
Or keep layering onto a model you have already admitted is obsolete, and changing your pricing page four times a year while calling it a strategy.
Next issue: Paying for the Motion. The comp rebuild, and where the pod model breaks.
First published in the Agentic Commercial Model newsletter on LinkedIn, September 4, 2026. Read the original on LinkedIn.