A customer service team of 50 people runs on Zendesk, paying per seat. The monthly bill is predictable. Now swap that setup for one operations person plus 50 AI agents. Ticket volume triples, but Zendesk collects revenue for a single seat.
This isn’t a thought experiment. It’s happening in 2026.
The subscription model didn’t lose to a competitor. It lost to its own logic. When AI lets one person do the work of five, charging by headcount stops making sense. Deloitte’s 2026 TMT forecast puts a number on it: seat-based revenue as a share of total SaaS income is sliding from 21% to 15%.
The question isn’t whether the old model survives. It’s what replaces it, who moves first, and whether buyers end up better off or worse.
The Unit Economics Problem
Many people frame this as “AI agents replaced users, so per-seat pricing doesn’t work.” That’s half right. The deeper issue is that AI broke the relationship between seats and value.
Traditional SaaS pricing rested on an assumption: one seat equals one unit of output. Ten sales reps using a CRM produce roughly ten units of sales work. Pricing anchored on user count because user count tracked value in a predictable way.
AI shattered that linearity. A marketing operator using AI tools can produce 30 pieces of content in a day. Five people used to be needed for the same output. One seat now delivers five times the value. If the vendor still charges per seat, the customer’s value grows exponentially while revenue stays flat.
Think of it this way: charging per seat in the AI era is like charging a self-driving car fleet by parking spots. The spot stays the same size, but the car runs all day.
A 2026 analysis from Korix shows the gap: if one seat paired with an AI assistant handles three times the workload, per-seat pricing leaves 67% of value on the table. Companies that saw this early started looking for alternatives.
Usage-Based Isn’t the Answer
When seat-based pricing failed, the first response was usage-based. Charge by API calls, by tokens consumed, by data processed. OpenAI’s API pricing is the archetype.
That’s progress. At least it ties revenue to usage. But it has a fatal flaw: usage doesn’t equal value.
An AI agent that burns through 100,000 tokens to resolve a support ticket creates value. An agent that burns the same tokens and solves nothing costs the same. The customer pays either way.
Salesforce’s Agentforce pricing history is instructive. When it launched in late 2024, pricing was $2 per conversation. By May 2025, it shifted to $0.10 per action, a finer-grained usage model. By early 2026, the company introduced Flex Credits, an attempt to align costs with outcomes. Three pricing models in 18 months. Salesforce is still searching for the right answer.
The problem with usage-based pricing is risk transfer. An inefficient AI agent that makes more calls costs the customer more. That’s the opposite of paying for results.
Usage-based is a transition, not an endpoint. The real destination is more radical.
Outcome-Based Pricing: The Endgame
Outcome-based pricing is simple: you don’t pay for the software’s existence or the number of API calls it makes. You pay for results.
Resolved a ticket? Pay. Generated a qualified lead? Pay. The AI ran 100 API calls but produced nothing? Free.
Between 2024 and 2026, this model moved from concept to reality.
Intercom Fin charges $0.99 per successful resolution. The word “successful” matters. If the AI can’t solve the issue and escalates to a human, there’s no charge. As of December 2025, Fin had handled over 40 million conversations with a 67% resolution rate. Intercom only charged for that 67%. The other 33% was cost the company absorbed.
Zendesk announced outcome-based pricing in August 2024. By its 2025 Relate conference, the model was live, starting at $1.50 per automated resolution. Plans include a set number of AI resolutions, with overage billed separately.
HighRadius went further. At its February 2026 Radiance conference, the company ditched per-seat pricing entirely. Zero implementation fees, zero pre-launch subscription costs. Customers pay only after they achieve measurable savings, as a percentage of those savings. This is the most aggressive outcome-based model in production.
Salesforce Agentforce evolved from $2 per conversation to $0.10 per action to Flex Credits. Three models running in parallel, balancing usage and outcomes.
Gartner predicts that by the end of 2025, over 30% of enterprise SaaS deals will include outcome-based components. In 2022, that figure was 15%. Chargebee’s 2025 subscription report shows 43% of companies have already mixed usage-based elements into their subscriptions, with expectations to hit 61% by year-end.
The trend is clear. But outcome-based pricing has its own problems.
Three Weaknesses in Outcome-Based Models
Attribution Is Hard
When a sales lead converts, was it the AI SDR tool, the brand campaign, or the final phone call from the rep? When multiple tools work together, pinning a result to one of them becomes a philosophical exercise.
Customer service is simpler. A ticket is closed or it isn’t. Attribution is clean. But as you move toward sales, marketing, or product work, attribution gets murky fast. That’s why customer service adopted outcome-based pricing first.
Unpredictable Budgets
CFOs hate surprises on the monthly bill.
Seat-based pricing offered certainty. Fifty seats at $100 each equals $5,000 per month. It goes in the budget and stays there. With outcome-based pricing, a busy month might cost $8,000, a slow month might cost $2,000.
Solvimon’s research found that 78% of IT leaders reported unexpected cost fluctuations when deploying AI. For a CFO presenting to the board, “I don’t know what next month’s software bill will be” is unacceptable.
Gartner’s 2023 survey found that 67% of CFOs listed “difficulty measuring ROI” as their top concern when approving AI investments. Outcome-based pricing theoretically solves ROI, you only pay when you get results, but it creates budget predictability problems.
Vendor Revenue Volatility
Buyers aren’t the only ones with headaches. For SaaS vendors, outcome-based models tie revenue directly to the customer’s business cycle. When the customer’s business slows, your revenue slows with it. Wall Street’s favorite narrative, predictable annual recurring revenue, becomes fragile under outcome-based pricing.
HighRadius’s model, zero upfront fees plus revenue share on savings, is attractive to customers but a cash flow management challenge for the vendor. Only companies with strong cash reserves and extreme confidence in their product can afford to play this way right now.
The Real Answer: Hybrid Three-Tier Pricing
Pure seat-based pricing is obsolete. Pure usage-based is unfair. Pure outcome-based is unpredictable. So what works?
The answer emerging in 2026 is a hybrid three-tier structure:
Tier 1: Fixed Platform Fee
A predictable monthly base that covers platform access, data storage, and core features. This gives the CFO a number they can budget.
Tier 2: Included Volume Cap
The plan includes a set amount of AI processing, maybe 1,000 resolutions or 5,000 actions. Within that range, no extra charges. Beyond it, usage-based billing kicks in.
Tier 3: Outcome Bonus
When AI output exceeds a threshold or hits specific business goals, the vendor charges a premium tied to results.
L.E.K. Consulting calls this a “burstable reserve” model: a predictable baseline plus outcome-based elasticity. Zendesk’s implementation fits this pattern. Plans include a set number of AI resolutions, with $1.50 per additional resolution. Salesforce’s three models running in parallel let customers pick the hybrid ratio that fits their needs.
My prediction: within three years, 70% of AI-native SaaS will use some form of hybrid pricing. Pure seat-based will retreat to sectors where AI penetration is low and headcount still tracks value, compliance audit tools, legal workflow management. Pure outcome-based will stay in verticals where attribution is clean and results are quantifiable, customer service, collections. Most products will land in the middle.
Who Moves First, Who Moves Last
Pricing shifts won’t hit every sector at once. There’s a clear sequence.
Wave 1 (2024-2025): Customer Service
- Intercom Fin: $0.99 per resolution
- Zendesk: $1.50 per automated resolution
- Why first: attribution is cleanest, tickets close or they don’t, results are quantifiable, AI replacement rates are highest.
Wave 2 (2025-2026): Sales and Lead Generation
- Salesforce Agentforce: conversation-based to action-based pricing
- Clay: credit-based model, a usage-based variant
- Why second: lead value is quantifiable, but attribution is more complex than customer service.
Wave 3 (2026-2027): Content Generation and Marketing
- Jasper: still $59 per seat as of now, but pressure is building
- Why later: defining a content “outcome” is hard. Is it publication count? Engagement? Conversions? Attribution chains are long and tangled.
Last to Move: Compliance, Security, Infrastructure
- Why they’ll stay seat-based longest: value in these sectors is preventing bad outcomes, not producing good ones. You can’t easily charge based on “days without a breach.” Seat-based or flat-fee models will persist here.
What Buyers Should Do
If you’re renewing a SaaS contract in 2026, a few things matter:
- Ask for hybrid options. If the vendor only offers seat-based, ask whether usage or outcome components exist. Many vendors have them but don’t lead with them.
- Negotiate a spending cap. If you accept outcome-based pricing, insist on a monthly ceiling. Outcome-based without a cap is a blank check.
- Define “outcome” in the contract. What counts as a resolution for Intercom? Does it mean the customer didn’t come back? Does it require explicit confirmation? The definition determines your bill.
- Retain audit rights. In outcome-based models, vendors have an incentive to mark “unresolved” as “resolved.” The contract should give you the right to audit outcome data.
FAQ
Q1: Will outcome-based pricing push vendors to cut quality to control costs?
No. In outcome-based models, vendors only get paid when they deliver results. The incentive is to raise resolution rates and output quality. The real risk runs the other way: vendors may loosen the definition of “success” to bill more often. That’s why buyers need tight contractual definitions of what counts as an outcome.
Q2: Is outcome-based pricing right for small companies?
It depends. If your usage is small but steady, a fixed subscription might be cheaper. Outcome-based pricing often has higher per-unit costs than a seat price amortized across the year. But if your usage swings wildly, with clear peaks and valleys, outcome-based pricing can save you money in the slow months.
Q3: What’s the core difference between usage-based and outcome-based?
Usage-based measures input. How many calls did the system make? How many resources did it consume? Outcome-based measures output. How many problems did it solve? How much value did it create? An AI that makes 100 API calls and solves nothing still generates a bill under usage-based. Under outcome-based, it’s free.
Q4: How do you budget for hybrid pricing?
Use “baseline plus estimated elasticity.” The fixed platform fee is certain. The elastic portion should be estimated from historical data as a range, P50 to P90. Budget at P75 and leave 15-20% flex room.
Q5: What signals that a vendor is about to change pricing?
Three signs: (1) They start tracking your AI usage and resolution rates in the background. (2) They offer a “new pricing pilot” at renewal. (3) Competitors have already switched to outcome-based. If you see any one of these, prepare your negotiation strategy.
Conclusion
SaaS pricing isn’t being disrupted by a competitor. It’s being disrupted by math. The assumption that headcount tracks value no longer holds when AI multiplies individual output by five. Seat-based pricing worked for 20 years because that assumption was stable. AI made it obsolete.
Usage-based was the first patch, but it charges for input, not output. It’s not the final answer. Outcome-based points in the right direction, paying for results, but its attribution and predictability problems mean it won’t dominate in pure form.
The 2026 answer is hybrid: a fixed base for certainty, a usage allowance for elasticity, and outcome bonuses for value alignment. Intercom, Zendesk, Salesforce, and HighRadius are already moving in that direction, each with different balances.
For vendors still watching from the sidelines, the question isn’t whether to change. It’s whether you want to lead the change or be forced into it. When your customers see a competitor charging only for results, your per-seat invoice starts looking expensive.
The seat-based era is over. The question now is what you build next.
Stay updated with our latest AI insights
