Two conversations are happening in parallel right now about AI value. In SaaS boardrooms, the question is why companies that built genuinely useful AI features can’t turn them into pricing power. In professional services partnerships, the question is why clients are demanding lower fees for work that still requires skilled people.
These sound like different problems. They are the same problem.
Monevate’s 2025 B2B SaaS Monetization Benchmarking Survey provides the data on the vendor side. The firm surveyed 75 B2B SaaS companies with annual revenues between $20 million and $500 million. Nearly half of those introducing AI features raised prices by less than 10%. Only one in five achieved increases above 20%.
The customers are saving time, improving efficiency, getting better outputs. The companies that built the capability can’t capture a proportionate return. That gap is not a coincidence. It is a structural consequence of what AI did to the thing being priced.
The vendor side: value created, value escaping
Monevate identifies three failures in how SaaS companies are approaching AI monetisation. Each one sounds like a tactical error. Together they reveal something more fundamental.
The first is over-reliance on “AI-powered” messaging. Customers no longer pay a premium for the label. An Irrational Labs study found that products explicitly branded as AI-driven were trusted less, and customers were willing to pay less for them. Buyers want to know what the feature helps them achieve, not how it was built.
The second is defaulting to cost-recovery pricing metrics. Because AI workloads are expensive, many companies charge per query, per token, or per transaction. Two-thirds of the 35 AI pricing models Monevate evaluated skewed toward cost recovery rather than value alignment. These models cover costs neatly, but they train customers to second-guess every interaction and treat AI as a metered utility rather than a source of competitive advantage.
The third is packaging that gives AI away without monetisation leverage. Over a quarter of firms in the survey include AI in their lowest tier with no usage gating. Customers who want the capability consume unlimited value without incremental charges. Margins erode and AI is perceived as a commodity.
Monevate’s prescriptions are sound: lead with outcomes rather than technology, design for adoption before monetisation, and anchor pricing to value rather than compute costs. What the analysis does not ask is why this pattern keeps recurring. Why vendors who built genuinely valuable features still cannot capture a fair return.
The services side: the execution premium collapses
The answer becomes visible when you look at the other side of the table.
In February 2026, KPMG International negotiated a fee reduction with its own auditor, Grant Thornton UK. The argument was blunt. AI and automation should be making audits faster and cheaper.
Those efficiencies should show up in pricing. Grant Thornton’s audit fee fell from $416,000 to $357,000, a 14% reduction.
The numbers are small. The signal is not.
Grant Thornton’s value proposition was skilled execution. KPMG’s counter was that the execution component is now worth less than it was twelve months ago. Not worthless. Worth less.
The audit itself, the independent verification, the regulatory accountability, the professional judgement, still matters. What KPMG repriced was the production work wrapped around it.
This is what I have been writing about as the collapse of the execution premium. The concept traces back to Robert Kaplan and David Norton’s 2008 book The Execution Premium, which argued that the firms capable of reliably translating strategy into operations earned a competitive advantage precisely because that capability was scarce.
AI is collapsing that scarcity. The premium does not shift to a different team or geography. It evaporates from the transaction.
Same phenomenon, different seat
The parallel to Monevate’s findings is structural, not superficial. Professional services firms bundled trust and execution into a single fee for decades. SaaS vendors bundled AI capability and customer outcomes into a single product tier. In both cases, the bundling obscured where the value actually sat.
The SaaS vendor who includes AI free in the base tier is giving away the execution layer. The professional services firm watching its fees compress is having its execution layer repriced by clients who know AI makes the work cheaper. The mechanism differs. The economics are identical.
Both are trying to price execution in a market where execution is approaching free. Neither has figured out how to price what sits above it on its own.
What sits above execution
If execution is approaching free, what justifies a premium?
Not the technology. Monevate’s own research shows that “AI-powered” labelling reduces customer willingness to pay. Not the volume of output. Per-query and per-token pricing commoditises the very capability vendors are trying to monetise.
What justifies a premium is trust. Specifically, the architecture that makes AI outputs reliable enough to act on.
Consider the domains where the output carries consequences: finance, audit, investment decisions, regulatory compliance. The question in those contexts is no longer “can you do the work?” AI can do the work. The question is “can anyone verify the work was done correctly?” That requires something AI alone cannot provide: deterministic constraints that do not flex based on who is asking or which model is running.
AI is powerful precisely because it is flexible. It reasons across ambiguity, fills gaps, synthesises patterns from incomplete data. That same flexibility is why its outputs cannot be guaranteed to be consistent. The capability that makes it useful is the same capability that makes it unreliable in domains where consistency matters.
I call this architecture a Deterministic Sandwich. AI sits in the middle, handling extraction, pattern recognition, and synthesis. It is flexible and probabilistic, which means it is also non-deterministic by nature. The same prompt can produce different outputs on different days.
So we bound it. At the base: a fixed taxonomy. Standardised definitions that ensure “revenue” means the same thing across every entity and every query, regardless of the underlying accounting system or reporting convention.
At the top: verification logic. Outputs checked against known constraints before they are released for decision-making.
Does this margin figure fall within the expected range? Does this classification match the taxonomy? If not, it gets flagged before anyone acts on it.
The sandwich does not slow AI down. It is what makes AI safe enough to deploy in environments where being wrong has consequences.
From value-aligned to trust-aligned pricing
Monevate calls their resolution “value-aligned pricing.” I would go further. What they are describing, at its best, is trust-aligned pricing.
Their example is instructive. An AI assistant creates value in the answers it delivers, even though costs accrue with every query processed. Rather than charging per query, they suggest charging per answer, priced to reflect the value of the output rather than the cost of producing it. That is a move in the right direction.
The reason it works is that the “answer” carries implicit trust. The customer is paying for an output they can act on. Not a probabilistic guess.
Not a raw model response. An answer they are willing to stake a decision on.
The metric that justifies the premium is not “insights generated.” It is “insights verified.” Not “tasks automated.” “Tasks automated with an audit trail.”
The SaaS vendor who builds deterministic verification into their AI features has something to charge for beyond the model. The professional services firm that builds trust architecture, the taxonomies, the validation layers, the deterministic constraints, has a value proposition that survives the execution premium collapse.
Everyone else is competing on execution. And execution is approaching free.
Two problems, one resolution
The monetisation gap on the vendor side and the fee compression on the services side are not two separate problems requiring two separate solutions. They are the same structural shift, viewed from different seats at the same table.
The execution layer got cheap. The organisations that had been bundling execution with trust, whether as software features or as professional services, are watching the bundle come apart. The ones trying to re-price execution will keep losing ground. The ones who build and price around trust architecture will capture the premium that execution used to carry.
The trust premium is where the value went. The only question is who builds the architecture to claim it.
