As AI shrinks seat counts and exposes misaligned pricing, CX leaders need contracts that link vendor revenue to resolutions, retention, and cost-to-serve — not machine effort. Here’s a practical framework to get there.
The enterprise software market is about to hit the harsh wall of reality. For the longest time, enterprise software vendors relied comfortably on seat-based pricing. It was a beautifully simple model: more humans, more licenses, more revenue. SaaS scaled the model elegantly. Then AI emerged and began doing the work that humans had been doing, drastically changing the equation. Suddenly, the better the software gets, the fewer seats you need. And vendors find themselves in what one may call the seat cannibalization trap.
The numbers are unambiguous. According to Growth Unhinged’s 2025 State of B2B Monetization report, seat-based pricing dropped from 21% to 15% of companies in just twelve months, while hybrid pricing surged from 27% to 41%. This shift is far from subtle. And it is accelerating. A16z reports that the space is rapidly evolving and that vendors are experimenting with different pricing approaches. Granted, that was at the end of 2024, and focused on startups, but it is still a signal. The industry is clearly searching for a new atomic unit of value.
As I discussed at some length in my recent article on Salesforce’s Agentic Work Unit, usage-based pricing has emerged as the popular middle ground. Salesforce’s AWU, Zendesk’s conversation counters, various token-based billing schemes, and more all measure machine exertion rather than value. McKinsey’s September 2025 analysis of 150 global software vendors is instructive here: 68% of incumbent vendors still rely on flat-fee metrics, while only 2% have adopted successful outcome-based pricing. AI-native vendors fare somewhat better at 10%. The gap between vendor talk and what is actually delivered still remains big.
To put it bluntly: usage-based pricing is a very poor proxy for value delivered. A confused AI agent that loops fifty times before timing out generates fifty billable work units while delivering zero value. A poorly optimized workflow that burns through thousands of tokens to produce a mediocre email draft still rings up the vendor’s cash register.
The customer pays for effort, not results.
The vendors wash their hands in innocence, much like Pontius Pilate.
The CX vendors getting it right – and the ones pretending
To their credit, some CX vendors are working on genuine outcome-based pricing. Intercom was among the first movers when it shifted its Fin AI agent from traditional per-seat pricing to a flat $0.99 per successful resolution. If the AI fails, the customer pays nothing. Intercom allegedly reported 40% higher adoption rates and maintained healthy margins within six months. Zendesk followed suit in late 2024 with a $1.50 per automated resolution model. Sierra, the startup co-founded by former Salesforce co-CEO Bret Taylor, has pushed this further by tying pricing to a broader set of outcomes: resolved conversations, saved cancellations, upsells, and cross-sells. Sierra reportedly hit $150 million in ARR by January 2026, growing revenue more than fivefold in just over a year. Business customers vote with their wallets, and the wallet likes outcomes!
What makes these approaches work? Looking a little deeper, three success criteria emerge. First, the outcome metric must be highly correlated with the product’s actual work. A resolution is something an AI agent directly causes; a company’s annual revenue is not. As McKinsey’s researchers put it, the lag time between workflow execution and outcome must be short.
Second, the definition of success must be crystal clear and mutually agreed upon up front. According to an article by the Pragmatic Institute, Zendesk learned this the hard way when ambiguity around what constitutes a resolution led to customer disputes.
Third — and this is where things get interesting for the enterprise — the vendor must have skin in the game. This is the crucial difference between outcome-based pricing and usage-based pricing with a marketing hat on.
The strategic KPI problem: attribution in a world of interdependencies
Per-resolution pricing works well for narrowly scoped CX use cases. But the real challenge — and the real opportunity — lies in tying AI investments to strategic business KPIs. Such as: customer lifetime value, net revenue retention, first-contact resolution rates across the entire operation, customer effort scores, or cost-to-serve ratios. These are the numbers that boards care about and that budget holders can defend.
The attribution problem is non-trivial. When multiple systems, teams, and process changes all contribute to a KPI movement, isolating the contribution of a single vendor or initiative is difficult. McKinsey’s analysis of 150 global software vendors found that true outcome-based pricing remains rare precisely because the outcome must be highly correlated to the product’s actual work and the lag between execution and measurable result must be short. Bain & Company’s analysis of 30+ SaaS vendors reinforces this: none of the incumbents studied had fully shifted to outcome-based pricing.
L.E.K. Consulting identifies five prerequisites for making it work: a clear link between services and measurable benefits, robust tracking systems, defined timelines, stakeholder alignment on metrics, and adequate technical and operational capabilities.
In other words, outcome-based pricing is not a pricing decision. It is an operating model decision.
A buyer’s framework for outcome-based AI pricing
For buyers who want to move beyond paying for machine effort, here is a four-step framework.
Step 1: Baseline before you buy.
Before engaging any vendor, measure your current strategic KPIs with precision. What is your cost-to-serve per interaction? Your first-contact resolution rate? Your churn rate? Whatever they are. McKinsey’s 2025 State of AI survey found that only 39% of respondents can attribute any EBIT impact to AI. The ones who can quantify impact are the ones who measured first.
Step 2: Agree on metrics and attribution before ink hits paper.
This is where most outcome-based contracts fail. The Emergence Capital framework suggests evaluating AI investments along two dimensions: autonomy (how independently the AI operates) and attribution (how clearly its actions link to measurable outcomes). For CX applications with high autonomy and clear attribution, such as automated ticket resolution, pure outcome-based pricing works. For AI that augments human workflows with diffuse impact, a hybrid model is more realistic. The critical point remains: Agree on the attribution methodology before the contract is signed.
Step 3: Build short feedback loops with contractual checkpoints.
McKinsey’s research finds that large-scale technology transformations fail to meet objectives more often than they succeed. The ones that do succeed iterate rather than attempt a “big bang.” Apply this learning to your pricing agreement. Structure contracts with 90-day review windows where both sides evaluate whether agreed-upon outcomes are materializing. If they are, scale up. If not, recalibrate.
Step 4: Demand a transparent measurement infrastructure.
If your vendor cannot provide real-time dashboards showing the outcomes they claim to deliver, that is a red flag. Bain & Company finds that most software companies still lack the product telemetry and billing infrastructure to support outcome-based models at scale. Insist on shared analytics, ideally using an independent measurement layer rather than the vendor’s own reporting. Trust but verify, as a Russian proverb famously says.
Five suggestions for buyers
Let me distill this into actionable guidance:
1. Establish your KPI baseline before talking to vendors.
Document your current cost-to-serve, resolution rates, satisfaction scores, and retention metrics with statistical rigor. This baseline is your negotiation power and, more importantly, your value-tracking baseline.
2. Negotiate and agree on the attribution methodology upfront, not after deployment.
Decide how you will isolate the vendor’s contribution from other factors, like market conditions, team changes, and complementary technology investments. Write it into the contract. As said above, L.E.K. Consulting’s research identifies stakeholder alignment on metrics as one of the five critical prerequisites for successful outcome-based pricing. Get this wrong, and disputes are guaranteed.
3. Use hybrid pricing models as a bridge.
Pure outcome-based pricing is ideal in theory but difficult in practice for complex, multi-touch CX initiatives. Bain & Company found that roughly 65% of SaaS incumbents introducing AI capabilities have adopted hybrid models, layering usage or outcome metrics on top of existing seat pricing. Start with a reduced base subscription plus outcome-based variable fees tied to agreed KPIs. As attribution matures and trust builds, shift the ratio toward outcomes.
4. Contractualize regular review cycles and exit ramps.
Quarterly reviews with data-driven outcome assessments are non-negotiable. Include provisions to recalibrate scope, pricing, or exit if agreed-upon outcomes are not met. This incentivizes the vendor to invest in your success, not just your adoption.
5. Insist on an independent or shared measurement infrastructure.
Do not let the vendor be the judge of outcomes. Shared dashboards, third-party analytics, or, at the very least, access to raw outcome data, are prerequisites for any outcome-based deal. As the Pragmatic Institute’s analysis of Zendesk’s resolution pricing illustrates, ambiguity around what counts as a “successful outcome” is the fastest path to commercial disputes.
The bottom line
The AI era is having seat-based pricing for lunch, and that is a good thing for buyers. However, the industry’s rush to usage-based models, dressed up in fancy language like agentic work units, tokens, or flex credits, risks replacing one misalignment with another. The real opportunity lies in pricing that ties vendor success directly to customer success: strategic KPIs, agreed upfront, measured transparently, and reviewed rigorously.
Vendors that embrace this will earn trust, retention, and pricing power. Those that resist will discover that in a world where AI makes switching costs lower and outcomes easier to measure, the seat they’re selling might just be their own.
Just my $.02. What do you think?