April 3, 2026
"If an AI agent can do the work of three support reps, why would you keep paying for three seats?"
That question is no longer hypothetical. Seat-based SaaS pricing, the dominant model for the last two decades, is under serious pressure. Not from a new startup or a clever pricing consultant, but from the basic economics of AI. When software replaces headcount, charging per head stops making sense.
This article walks through the pricing spectrum that's emerging in response, from traditional SaaS to full outcome-based pricing, and explains why the shift matters for every company buying or building software today.
For most of the SaaS era, software has been sold like an American buffet. You pay a flat fee per seat per month. Within that, you can use as much as you want. No usage caps, no metering, no surprises on your invoice.
This model drove massive adoption. It removed friction. It made budgeting simple. And it worked for a long time.
But it also created a hidden subsidy. Your power users, the ones logging in daily, running complex workflows, consuming real compute, pay the same flat rate as the person who logs in once a month. The heaviest users get the most value per dollar. The lightest users quietly subsidize them.
This dynamic was fine when the marginal cost of serving another user was close to zero. But AI changed the math. Every AI-powered feature carries real compute costs: inference tokens, API calls, GPU time. A flat per-seat fee can no longer absorb that. SaaS companies are being forced to rethink how they charge, and software buyers need to understand what's coming.
What's replacing the buffet isn't a single new model. It's a spectrum. Think of it as five rungs on a ladder, moving from the most abstracted pricing to the most aligned with actual value delivered.
This is the model everyone knows. You're paying for access. The number of users determines the price. Whether those users are active or inactive, productive or idle, the bill stays the same.
It's simple. It's predictable. And it's declining. According to Growth Unhinged's 2025 State of B2B Monetization report, seat-based pricing dropped from 21% of SaaS companies to 15% in just twelve months. Industry analysts broadly expect the majority of vendors to move away from pure per-seat models within the next few years.
The reason is straightforward: AI agents don't need seats. A company that once required 50 Salesforce logins might now need 15, with AI handling the rest. Salesforce's own Agentforce and Data 360 products hit nearly $1.4 billion in combined ARR, growing 114% year-over-year. But that growth is simultaneously cannibalizing their seat-based revenue. The irony is real: Salesforce built AI that helps customers need less Salesforce.
This is the first step away from the buffet. Instead of paying a flat fee, you pay for what you actually consume: CPU time, API calls, tokens processed, data stored.
AWS popularized this model for infrastructure. Now it's spreading into application software. The appeal is obvious: you only pay for what you use. But it introduces unpredictability. Enterprise buyers routinely report unexpected charges from consumption-based AI pricing, and CIOs consistently rank cost forecasting as one of their top challenges in AI deployment.
Micro usage-based pricing aligns cost with consumption, but not necessarily with value. You could burn through a million tokens on a task that produces nothing useful. The meter runs regardless of whether you got what you needed.
This is the middle ground. Instead of metering raw compute, you're charging for a meaningful unit of work, something that clearly matters to the customer, even if it's not the final outcome.
Credits are the mechanism most companies are using to get here. The PricingSaaS 500 Index found that 79 companies now offer credit-based models, up 126% year-over-year. HubSpot, Salesforce, Figma, and Adobe have all adopted credit structures. Clay, one of the most-watched companies in sales tooling, just restructured its entire pricing around a split between "Data Credits" and "Actions," separating the cost of data from the cost of platform orchestration.
Credits sit between access pricing and outcome pricing. They give customers more transparency than a flat seat license while being easier to implement and measure than pure outcomes.
A concrete example: a fintech platform that generates customized investment proposals for financial advisors. Each proposal is a unit of meaningful value. It's not a raw API call, and it's not the final outcome (winning the client), but it's clearly worth something to the advisor. Charging per proposal generated aligns price with a value event rather than raw consumption.
Now we're getting closer to the finish line. A micro outcome isn't the ultimate business result, but it's a measurable step on the path to it.
In sales and marketing, leads are the clearest example. Companies like Clay and Apollo sell access to enrichment and prospecting workflows. The value isn't in the API call or the data credit. It's in the qualified lead that comes out the other end. A booked meeting, a verified contact, a scored prospect: these are micro outcomes on the path to a closed deal.
The distinction from value-based pricing is subtle but important. A credit buys you a unit of work. A micro outcome pays for a unit of progress. The customer isn't buying the activity. They're buying a step closer to the result they actually want.
This is the bottom of the spectrum, and it's where the industry is heading. You pay only when the software delivers the actual result you're looking for.
This is no longer theoretical. Intercom charges $0.99 per AI-resolved support ticket. Their Fin AI agent has processed over 40 million resolved conversations with a 67% resolution rate and is approaching $100 million in annual recurring revenue, growing at roughly 3.5x year-over-year. Zendesk charges $1.50 to $2.00 per automated resolution. Salesforce prices its Agentforce on completed actions, not human seats. Decagon offers per-resolution pricing for enterprise support.
Gartner projects that by 2030, at least 40% of enterprise SaaS spend will shift toward usage-, agent-, or outcome-based pricing. That timeline may be conservative: Gartner separately forecasts that 40% of enterprise applications will feature task-specific AI agents by the end of this year, up from less than 5% in 2025, which will only accelerate the pricing shift. Chargebee's 2025 State of Subscriptions Report found that 43% of companies already use hybrid models, with adoption projected to hit 61% by the end of this year.
The logic is clean. When software can track its own results in real time, charging for access stops making sense. Charging for results becomes the obvious model.
Why this matters for buyers: Outcome-based pricing shifts risk from the buyer to the vendor. If the AI doesn't resolve the ticket, Intercom doesn't get paid. That's a fundamentally different deal than paying $150 per seat per month regardless of whether the software actually does what you bought it to do.
Outcome-based pricing sounds ideal in theory. In practice, two challenges keep it from being a universal solution.
The first is attribution. Did the AI close the sale, or did the rep's follow-up email? Did the fraud detection platform catch the attack, or did the internal security team flag it? When outcomes depend on multiple systems and human actions, determining who or what deserves credit gets complicated fast.
The second is predictability. Enterprise buyers need to set budgets. If pricing is purely variable, finance teams can't forecast spend. This is why hybrid models are growing faster than pure outcome-based ones. A common structure: a predictable monthly platform fee for access and core features, with outcome-based charges layered on top when AI delivers measurable results above a baseline.
Most companies aren't choosing between per-seat and outcome-based pricing. They're combining them. And that's probably the right answer for most buyers, at least for now.
The pricing spectrum above applies to software products. But the same forces are reshaping how technology services get priced, from dev shops and agencies to legal and consulting firms.
Services have historically lived in the hourly world. The worker does the work, the client gets billed for the hours. It's a form of usage-based pricing: however many hours are consumed, mark them up, and that's the invoice.
This model has always had an uncomfortable misalignment. The client's goal is the output. The provider's revenue is tied to the input. Faster delivery means less money for the provider. The incentives work against each other.
AI is making that misalignment impossible to ignore. For the first time, AI is meaningfully divorcing hours from output in knowledge work. According to Jellyfish's 2025 State of Engineering Management Report, 62% of engineering teams report at least a 25% productivity increase from AI tools, with deeply engaged teams seeing 30-50% faster throughput. The gains are uneven: the most skilled engineers, the ones who can effectively direct AI and fix its output, see the largest multipliers. Less experienced engineers often see more modest improvements, because using AI well still depends on being fundamentally good at the underlying work.
The implication for pricing is direct. If a strong engineer augmented by AI can produce in two hours what used to take eight, hourly billing makes no sense for either side. The client overpays relative to the effort. Or the provider undercharges relative to the value. Either way, the hourly model breaks.
The traditional alternative in tech services is the fixed bid: a flat price to deliver an entire project. Fixed bids have existed for decades, and for good reason. They give the client cost certainty. But they create a different problem. Software projects have a lot of unknowns. To protect their margin, providers pad their estimates to account for overruns, bug fixes, scope changes, and QA. That padding can be significant, sometimes doubling the real cost. The client gets cost certainty, but at a steep premium for uncertainty they may never actually encounter.
The middle ground is pricing per unit of work rather than per hour or per project. In software development, that unit is the story point: a standardized measure of task complexity. Story point pricing works like slicing the fixed bid into thin, individually scoped increments. Each task gets estimated and priced before work begins. The provider doesn't need to pad a massive lump-sum estimate because the risk is distributed across many small, well-defined units. The client gets transparency into what each piece of work costs and can prioritize accordingly.
AI makes this model more practical than it used to be. The same AI that's good at writing code is also effective at reading task descriptions, assessing complexity, identifying substeps, and estimating effort. That planning and estimation process, which used to be manual and inconsistent, can now be automated and standardized. The result is a pricing model where the cost of each task is scoped upfront, tied to output rather than hours, and small enough that the risk of any single estimate being wrong doesn't blow up the project economics.
These pricing models, both for software products and services, are at different stages of maturity. Traditional per-seat SaaS still dominates by installed base. Hourly billing still dominates in services. But the trend lines are unmistakable. Usage-based pricing is mainstream. Credits are proliferating. Outcome-based pricing is live in production at some of the largest software companies in the world. And in services, the pressure to move from hourly to output-based is accelerating with every improvement in AI tooling.
Here's what hasn't changed: the fundamental question for software buyers is still whether a given tool or service delivers value that justifies its cost. What's changing is that the cost structure is becoming more transparent and more aligned with actual value delivered.
For companies evaluating software investments today, the practical takeaways are straightforward. First, look at how your vendors are pricing AI features. If they're bundling AI into existing seat licenses, that won't last. Expect pricing changes. Second, understand what pricing model aligns with how you'll actually use the product. If you're buying a tool for its AI automation capabilities, a per-seat model is probably overcharging you for idle seats and undercharging you for compute. Third, negotiate with the new models in mind. Credits, usage tiers, and outcome-based components are all levers that didn't exist in most SaaS contracts two years ago. And if you're buying development services, ask the same question: are you paying for hours, or are you paying for output?
At Fraction, this is the model we've built. We charge $149 per story point, scoped before development begins, with a structured breakdown so you know exactly what you're paying for before you commit. AI helps automate the estimation process, which means tighter scoping and less padding. The engineer gets compensated for completing tasks, not for logging hours. The client pays for output, not input. As AI reshapes the economics of both building and buying software, that alignment between cost and value becomes more important, not less.
The buffet era isn't over yet. But the menu is changing. The companies that understand the new pricing landscape, whether they're buying software products or development services, will make better decisions, negotiate better contracts, and avoid locking themselves into models that won't survive the next two years.
Related reading: Build vs. Buy AI: When to Build Custom and When to Use Off-the-Shelf, The Future of Build vs. Buy: Throwaway Software, Dark Factories, and Liquid Code, AI Strategy for Non-Technical Founders
Growth Unhinged, "2025 State of B2B Monetization Report." Documents seat-based pricing declining from 21% to 15% of SaaS companies in twelve months, and hybrid pricing surging from 27% to 41% in the same period.
Gartner, August 2025. Predicts 40% of enterprise applications will feature task-specific AI agents by end of 2026, up from less than 5% in 2025. Separately, Gartner projects that by 2030, at least 40% of enterprise SaaS spend will shift toward usage-, agent-, or outcome-based pricing (cited via Deloitte, "SaaS Meets AI Agents," November 2025).
Chargebee, "2025 State of Subscriptions Report." Found 43% of companies already use hybrid pricing models, with adoption projected to reach 61% by end of 2026.
PricingSaaS 500 Index, February 2026. Tracked 79 companies offering credit-based models, up 126% year-over-year from 35 at end of 2024.
Salesforce Q3 FY26 Earnings Release, December 2025. Reported Agentforce and Data 360 combined ARR of nearly $1.4 billion, up 114% year-over-year, with over 9,500 paid Agentforce deals.
VentureBeat, "Intercom's new post-trained Fin Apex 1.0," March 2026. Reports Fin approaching $100M ARR, growing at 3.5x, with 67% average resolution rate across 40M+ conversations. Intercom overall at $400M ARR.
Mostly Metrics, "How Intercom Reaccelerated Growth with Outcome-Based Pricing," March 2026. Details Intercom's pricing transformation, including NRR jumping from 112% to 146% and Fin on pace to represent half of Intercom's revenue.
EY-Parthenon, "SaaS Transformation with GenAI: Outcome-Based Pricing," February 2026. Analyzes revenue recognition implications as SaaS companies transition from seat-based to outcome-based models.
Clay pricing restructure announcement, March 2026. Introduced split between Data Credits and Actions, separating data costs from platform orchestration.
Jellyfish, "2025 State of Engineering Management Report," July 2025. Found 90% of engineering teams using AI coding tools (up from 61% the prior year), with 62% reporting at least a 25% productivity increase and deeply engaged teams seeing 30-50% faster throughput.
