This is the finale of The Agentic Commercial Model, a newsletter about the operating model crisis hiding inside the AI gold rush. Six issues.
We've spent five issues taking things apart. Seat pricing that doesn't fit consumption. Comp plans built for a moment that no longer exists. A retention metric that lies. A moat most companies are digging by accident. A go-to-market motion optimising for a buyer who has already left the room. Five demolitions.
This issue is about what you build on the cleared ground. And it comes down to a single sentence that I think is the most important thing I can leave you with after six issues:
Whoever measures value owns the customer.
Not whoever ships the best model. Not whoever prices the lowest. Not whoever has the biggest sales team. In the agentic era, the company that can define, measure, and prove the value its product delivers is the company that controls the commercial relationship, sets the price, wins the shortlist, and keeps the customer. Everyone else is negotiating in the dark against someone who can see. Let me show you why, using the fight that's happening in the market right now over exactly this question.
The pricing everyone agrees is coming, and nobody agrees how to do
Start with where the market has landed, because there's rare consensus on the destination and open warfare on the route.
The destination is outcome-based pricing: you charge for a result delivered, not access granted or activity consumed. The whole series has been walking toward this. Issue 02 argued the progression is seat to consumption to outcome, and that most companies stop at consumption and call it done. The market has now caught up to the endpoint. Gartner projects 40% of enterprise SaaS will include outcome-based elements by 2026. The examples are real and multiplying: Intercom's Fin charges $0.99 per resolved conversation. Zendesk charges around $1.50 per automated resolution, defined by a 72-hour quiet period where the ticket stays closed. Sierra gets paid only when it completes the task. HubSpot has moved its Breeze agents to per-outcome pricing. This is no longer theoretical. It's the leading edge of how AI gets sold.
And the logic is beautiful on the surface. As Bessemer put it, Intercom's per-resolution price aligns every team, sales, product, engineering, success, around one outcome: resolved tickets. Pricing becomes a north star. You pay for value, the vendor is incentivised to deliver value, incentives align, everyone wins.
Except a serious, credible voice in the market is calling this a trap. And they have a point sharp enough that you cannot build an outcome model without answering it.
The elegant trap: the argument you have to beat
Parloa, in a public and deliberately provocative piece, called outcome-based pricing "the most expensive myth in enterprise AI." Their argument is not lazy contrarianism. It's the strongest critique of outcome pricing in the market, and it lands two genuine punches.
The first punch is attribution. AI agent performance never lives in isolation. It depends on the CRM, the routing logic, the knowledge base, the process design, the customer's own team. Outcome-based pricing tries to treat a multi-variable system as a single-variable bill. So who gets credit when a ticket resolves? The AI, or the process redesign that fed it clean data? When a deal closes, was it the agent or the salesperson who followed up? This isn't a footnote. It's why, by one estimate, only around 17% of enterprise SaaS vendors had implemented true outcome-based pricing as of a few years ago: it's genuinely hard to measure outcomes consistently, and unclear attribution produces disputes, reconciliation meetings, and measurement debates that poison the relationship.
The second punch is the one that should stop you cold, because this entire newsletter has been circling it. Parloa argues that outcome-based pricing shifts the value of efficiency away from the customer and toward the vendor. Here's the mechanism: as your agent gets better and resolves an issue in less effort, the customer created that efficiency, but under a per-outcome model the vendor captures it, because the fee per outcome stays the same while the cost to deliver it falls. Improvements the customer generates become revenue the vendor pockets. Parloa calls this the principal-agent problem in pricing, and their answer is to charge only for raw consumption, minutes and calls, so that when agents get more efficient, the customer's bill goes down and the gains stay in the customer's P&L.
That is a serious argument, made well. And if you can't answer it, you should not be running an outcome model. So let me answer it, because the answer is the whole thesis of this series.
Why Parloa is right about the problem and wrong about the solution
Parloa has correctly diagnosed the disease. Their cure is to amputate.
Look closely at what their solution actually does. By retreating to pure consumption pricing, minutes and calls, they've walked all the way back to the model Issue 02 dismantled. They've solved the efficiency-capture problem by abandoning value-based pricing entirely and returning to activity-based pricing. And activity-based pricing has its own fatal flaw, the one this series opened with: it prices the meter, not the outcome. It cannot tell the customer whether they're getting value, only how much they consumed. It reintroduces exactly the unpredictability and value-blindness that pushed the market toward outcomes in the first place. Parloa escapes the trap by climbing back into the older, colder cage and calling it freedom.
The efficiency-capture problem is real. But the answer isn't to stop pricing on value. The answer is to price on value and build the measurement and the fairness that make it legitimate. Both of Parloa's punches, attribution and efficiency capture, are not arguments against outcome pricing. They are arguments against outcome pricing done without the measurement infrastructure and the fairness architecture that make it work. Which is precisely the thing almost nobody is building, and precisely the thing that becomes the durable advantage for the companies that do.
This is where the series converges. The reason outcome pricing is hard is the reason it's valuable. The measurement problem isn't a bug to be avoided by retreating to consumption. It's the moat.
The three tensions, and the one answer that resolves them
Everything difficult about outcome pricing reduces to three tensions. Watch how the same capability resolves all three.
Tension one: attribution. Can you prove your agent caused the outcome? The companies that win outcome pricing are the ones that invest in verifiable outcome definition. Look at what the credible players actually do. Zendesk doesn't just claim a resolution; it defines one, a 72-hour quiet period with no reopen, a technically verifiable signal. Intercom defines resolution through explicit confirmation or the absence of follow-up. The pattern among everyone succeeding here is the same: a technically verifiable outcome, agreed by both sides before the contract starts, detected by system events rather than argued after the fact. Attribution stops being a dispute when you've built the infrastructure to measure the outcome unambiguously. The vendors who can do that win. The ones who can't, retreat to consumption, as Parloa did.
Tension two: efficiency capture. Who keeps the gains when the agent improves? This is a fairness problem, and it's solvable by design, but only if you can measure the value delivered to the customer, not just the outcome billed. If you can measure the actual value the customer realises, you can build a model that shares the efficiency gain instead of quietly confiscating it: pricing that steps down as the cost-to-deliver falls, or that ties the fee to the customer's realised value so both sides win as efficiency rises. Parloa is right that a naive per-outcome fee captures the efficiency for the vendor. But that's a choice, not a law, and it's only avoidable if you're measuring realised value richly enough to share it fairly. Which, again, requires the measurement capability.
Tension three: is it real value or just billable activity? This is the oldest question in the series, and it's the one that separates the companies that will thrive from the ones running cosmetic outcome pricing, a consumption model with the word "outcome" painted on it. The only way to answer it honestly is to measure what the customer actually realised.
And that is Value-Realised Retention, the construct I introduced in Issue 04, now revealed as the foundation the whole model stands on. VRR measures the value the customer realised in outcomes, weighted by real margin, with usage depth as a live signal, rather than the revenue you happened to bill. It's the instrument that resolves all three tensions at once. It proves attribution, because you're measuring verified outcomes. It enables fairness, because you can see the efficiency gain and choose to share it. And it distinguishes real value from billable activity, because that distinction is the only thing it measures. Outcome pricing without VRR is Parloa's trap. Outcome pricing built on VRR is the model that wins.
Why some companies win exponentially and the rest win single digits
Here is the divide that will separate the software winners of this decade from the survivors, and it is not the one most leadership teams are watching. It shows up as a chasm in revenue growth, in customer lifetime, and ultimately in valuation multiple, and it traces back to a single architectural choice most boards never inspect.
Look at what a company is actually selling underneath the AI branding, because there are two architectures with completely different economics.
The first sells augmentation. A chatbot answers a question. RPA runs a pre-written script. RAG retrieves better context before it answers. These feel like AI, and they are, but they share one economic trait that caps their entire value: the human still does the work of deciding. The system assists, the person decides. When you sell assistance you can only capture productivity gains, which is why this gets priced per seat or per query, why it commoditises fast, and why its pricing power erodes. Anyone can bolt an LLM onto a knowledge base. The switching cost is near zero. This is the top half of the market, and it wins single-digit growth, because single digits is what augmentation is worth.
The second sells completed outcomes, and it is a different business entirely. Three components carry the commercial weight, and each maps to a distinct, durable advantage.
The orchestrator, with planning and feedback, is the shift from answering to deciding. The system decomposes a goal, sequences the work, checks its own output, and iterates until the job is done. Commercially, this is where per-seat pricing dies, because the agent is now doing the work the seat used to do, and you can finally price against the completed outcome rather than the tool. This is the architecture that earns the right to charge on outcomes. Everything this series has argued about outcome pricing only works if this is what you've actually built.
Memory is the moat, the same moat the special edition described, now located precisely. Every cycle through the loop, the system accumulates context about the customer's business, their preferences, their edge cases. Switching vendors stops meaning "migrate the data" and starts meaning "abandon the accumulated judgement." That's an epistemological switching cost, not a logistical one, and it compounds daily. It is the strongest retention asset in the whole architecture, and most buyers won't recognise it until they try to leave and find they can't.
The multi-agent protocol is the ecosystem play. Once agents discover each other's capabilities and route tasks between them, the vendor controlling the orchestration layer controls the demand routing to every specialist agent. That is a platform position with take-rate economics, the way app stores and ad exchanges monetise. The specialist agents are interchangeable suppliers. The protocol owner is the toll booth, and toll booths do not win single digits.
So here is the one-line version. The top half sells labour augmentation priced per user and grows single digits. The bottom half sells completed outcomes priced per result, with memory as the retention moat and the orchestration protocol as the platform toll booth, and it compounds, on revenue, on customer lifetime, on valuation. The value has migrated from the model to the loop.
And now the diagnostic that ties this straight back to everything this series has said about pricing. Most vendors are selling bottom-half stories on top-half architectures. They market outcomes and autonomy while they've built assistance and retrieval. And the tell, the thing that gives them away instantly, is their pricing model. A company selling per seat is telling you, whatever the marketing claims, that it has built augmentation, because if it had built an outcome engine it would be pricing on outcomes. The pricing model is the confession. It reveals the architecture beneath the branding, and therefore which half of the economics the company actually lives in.
This is why measurement is the whole game. To price on outcomes you must be built for outcomes, and you must be able to measure and prove the outcome you deliver. Architecture earns you the right to charge on value; measurement lets you actually do it. The companies with both compound exponentially. The companies with neither grow single digits and call it a good year. And the companies that built the loop but can't yet prove what it delivers are leaving the exponential on the table, one unmeasured outcome at a time.
Whoever measures value owns the customer
Now stand back and see why this single capability, the ability to measure realised value, is the thing the entire series has been building toward.
The company that can measure value prices with confidence, because it can prove what it delivers instead of guessing (Issue 02). It compensates correctly, because it can pay its people on value realised rather than on the close (Issue 03). It sees the truth about retention, because VRR shows what NRR hides (Issue 04). It deepens the moat, because measuring usage depth is the same act as deepening the bespoke calibration that makes a customer unable to leave (the special). And it wins the shortlist, because it can make its value legible and specific in exactly the way the machine now rewards (Issue 05).
Every single thread of this series runs through one capability: the ability to define, measure, and prove value delivered. That's why the thesis isn't "price on outcomes." It's deeper. Whoever measures value owns the customer, because measurement is what makes every other part of the commercial model work. The vendor who can prove value sets the terms. The vendor who can't is left arguing about minutes and calls, retreating to activity pricing, hoping the customer renews, and quietly losing the accounts to whoever can see what they cannot.
This is also why it's a moat and not just a feature. Measurement infrastructure is hard, slow, and expensive to build. It requires instrumenting the product to capture outcomes, integrating the cost data to weight by margin, and earning the customer's agreement on what success means. That difficulty is exactly why it's defensible. Anyone can copy a price. Almost nobody can quickly replicate the capability to measure and prove realised value across a customer base. The company that builds it first, in a given category, doesn't just price better. It owns the definition of value in that category, and everyone else has to sell against its scoreboard.
The manifesto: what this whole series has been trying to tell you
So here is the argument, all six issues of it, in one breath.
A world-class AI product is no longer enough. It never was. The product was necessary and it was never sufficient, and the companies about to discover this the hard way are the ones who poured everything into the model and left the commercial operating model exactly as it was.
Because everything around the product has changed. The way you price has to change, because seats and even consumption don't capture agentic value. The way you pay your people has to change, because rewarding the close is rewarding the wrong moment in a forever relationship. The way you measure has to change, because your retention metric is lying to you and hiding structural decay behind a comfortable number. The way you defend your position has to change, because the real moat is the bespoke value accumulating in every customer, measured or squandered. The way you reach a buyer has to change, because the shortlist now forms in a machine before your sales team knows the buyer exists. And the way you charge has to change, because value that can't be measured won't be paid for, and whoever measures value owns the customer.
None of these are product problems. Every one of them is a commercial operating model problem. And they are not six separate problems. They are one problem, wearing six faces: the operating model was built for a world that no longer exists, and shipping AI on top of it changes nothing, because the thing that needed to change was never the product.
This is the lazy trap the whole industry is walking into, the same laziness that funds AI by cutting people instead of creating value, that adds a consumption SKU and calls it transformation, that ships a brilliant agent onto a commercial chassis designed in 2015 and wonders why the value leaks out at every seam. Bolting AI onto an unchanged operating model is not strategy. It's the avoidance of strategy, performed confidently, and the market is about to start pricing the difference.
To the operators: stop optimising the old machine at the margins. Rebuild it. Price on value, measure what you deliver, pay your people for realised outcomes, and make your value legible to the machine and the buyer alike. To the boards and PE operating partners: stop taking comfort from the green numbers on the old dashboard. Ask what they're hiding. The value-creation thesis for the next decade of software is not "which portfolio companies shipped AI." It's "which ones rebuilt the commercial operating model around it." That is the line between the companies that compound exponentially, on revenue, on customer lifetime, on valuation multiple, and the ones that grow single digits and congratulate themselves for it. That is where the returns are, and where the wipeouts will be.
I've spent twenty years building commercial engines inside software businesses, and I've never seen a moment where the gap between the companies that understand this and the companies that don't will open as fast, or as permanently. The agentic transition is not a product race. It's a commercial operating model race. The winners are already measuring what they deliver. Everyone else is about to find out that value which can't be measured won't be paid for, and that the customer belongs to whoever can prove it.
That's the work. That's what FR Advisory exists to do, and it's what this series has been arguing from the first issue. If it resonated, the conversation doesn't end here. It starts.
Build the model the product deserves.
Thank you for reading The Agentic Commercial Model. Six issues, one argument: a world-class product is not enough, and the commercial operating model is where the agentic era is won or lost. If it changed how you see the problem, share it with the operator, CCO, or board member who needs it.
This series is the short version. The full thesis, every model taken apart and rebuilt, is in my book Dead Model Walking. Link in the comments.
First published in the Agentic Commercial Model newsletter on LinkedIn, July 3, 2026. Read the original on LinkedIn.