There’s a weird disconnect in SaaS right now that keeps coming up in conversations. On one hand, everyone is poking holes in per-seat pricing. On the other hand, valuation multiples haven’t dropped the way you’d expect if the model was actually breaking.
Both things are true, but they’re not showing up in the numbers the same way.
Per-seat pricing is getting questioned because AI is compressing the relationship between headcount and output. Historically, software scaled with people. More employees meant more seats, more revenue, clean expansion math. That was the magic of SaaS. Now you can have a smaller team doing more work, or AI agents replacing entire workflows that used to require multiple humans. So the buyer looks at a per-seat model and starts asking why they’re paying more as efficiency increases. It feels backwards.
But valuations aren’t anchored to pricing models; they’re anchored to the durability of revenue and belief in future expansion. And right now, most SaaS companies haven’t actually seen a material break in their numbers. Net retention is holding up well enough. Contracts are still largely structured the same way. Budgets haven’t been fully reallocated. So, from the outside, the metrics still look like those of a SaaS company. The disruption is more visible in product usage than in reported revenue… at least for now.
There’s also a lag effect. Pricing model changes are among the last to move in a market. Founders don’t rush to reprice their entire business unless they’re forced to. Buyers don’t push hard enough until there’s a clear alternative. So we’re in this in-between phase where everyone agrees the model is imperfect, but it’s still the default.
What’s actually happening underneath is a shift from paying for access to paying for outcomes, even if it’s not fully expressed in pricing yet.
If you zoom out, the future of SaaS pricing for AI-heavy products probably breaks into a few buckets.
Usage-Based
But not in the old API sense. More like units of work. Reports generated, claims processed, leads enriched, migrations completed. The more pricing is tied to a clear, valuable output, the easier it is to justify. Especially when AI is doing the work instead of a human.
Outcome-Based
Everyone talks about it, but it is harder to execute. Charging on revenue lift, cost savings, or some KPI movement. This works best when the product sits close to a financial metric and attribution is clean. Payments, revenue cycle, and certain fintech workflows. Much harder in horizontal tools where value is diffuse.
Hybrid Models
A base platform fee plus usage or outcomes layered on top. This is probably where most companies land because it preserves some predictability while still capturing upside from AI-driven efficiency. Think of it as a hedge between old SaaS and what’s coming.
Agent-Based
Not seats, but “workers.” You’re not buying licenses for employees, you’re buying capacity from AI agents. A customer might pay for 10 agents that can handle a certain volume of work. This starts to look more like labor economics than traditional software pricing.
The interesting part is what doesn’t change. Investors still care about the same core things. Is revenue predictable? Does it expand over time? Is pricing aligned with value in a way that compounds as the customer grows?
If anything, AI raises the bar. If your product is saving customers money by reducing headcount, you need a way to participate in that value, or you’ll get priced down. If your product is driving new revenue, you have more leverage to move toward outcome-based models.
So the reason valuations haven’t moved dramatically is that the scoreboard hasn’t changed yet. But the rules of the game are clearly shifting. Over the next few years, I’d expect to see a divergence. Companies that successfully tie pricing to value creation, especially in AI-driven workflows, will hold or expand multiples. Companies clinging to seat-based models in areas where AI is actively reducing seats will feel pressure.
We’re not watching SaaS break. We’re watching it unbundle from headcount. And that’s a much bigger shift than it looks like in the numbers today.
