In the old world, you bought customer service software the way you bought office chairs: count the people, count the seats, sign the contract, and move on with your day.
Now we’re being asked to count “seats” for things that do not sit, do not sleep, and do not complain about lunch breaks. Which is convenient for procurement, right up until it isn’t.
Because once AI agents take a meaningful slice of work away from humans, the per-user-per-month model stops being a pricing mechanism and starts looking like a small accounting fiction everyone has agreed not to mention. As Zeus Kerravala from ZK Research puts it:
“It’s not really a virtual agent, it’s just a computer program. So charging by utilization, I think, is the fairest way to do it.”
Per-seat pricing has underpinned UCaaS and CCaaS for years, but it is becoming misaligned with how contact centers now operate. As AI agents handle a growing share of customer interactions, enterprises can increase service capacity without adding human headcount, which undermines the logic of paying “per agent.” This shift affects enterprise buying committees, vendor economics, and even investor confidence, because the unit of value is moving away from seats and toward usage and outcomes.
Why “AI Agent Seats” Break the Per-Seat Pricing Model in Contact Centers
There is a nightmare scenario playing out in slow motion inside enterprise buying committees.
A 500-agent contact center deploys agentic AI that autonomously handles 40 percent of interactions.
Service levels improve. Customer satisfaction rises. The business grows.
Then the CFO asks a question that is hard to answer without sounding like you’re charging rent on an idea:
If AI is doing 40 percent of the work, why are we paying the same price?
This is the practical consequence of “digital labor” entering a licensing model designed for human labor. Seat pricing assumes that value scales with headcount. Agentic AI breaks that assumption by making capacity scale with compute, orchestration, and automation coverage—none of which maps neatly to a named user.
The deeper issue is incentive alignment. With seat-based pricing, the vendor’s commercial upside is tied to the customer hiring more people. But the customer’s operational goal is often the opposite: improve containment, reduce handling time, and shift work left into automation.
That tension existed before AI, but it becomes sharper once autonomous agents are doing meaningful work.
“Investors love predictability… Consumption models are spiky. They go up and down.”
The arbitrage nobody wants to own
Here’s the part vendors often try not to say out loud: many providers increasingly buy AI compute in variable, consumption-based ways (tokens, GPU hours, LLM usage) but still sell on fixed per-seat contracts.
When AI usage spikes, costs can spike.
But revenue does not.
That creates a commercial arbitrage that can quietly eat margins, especially as autonomous systems become more capable and more widely used. It also explains why this isn’t just a procurement debate; it’s a business model debate.
Related Stories
- The Demise of Per-Seat Contact Center Pricing, and What Comes Next?
- Why Wall Street is Suddenly Nervous About Your Contact Center Vendor
- How Data Layers and AI Are Rewriting the CCaaS Market
Why Wall Street Cares: Predictable Seat Revenue vs Volatile AI Consumption
Seat pricing isn’t popular because it’s philosophically pure.
It’s popular because it’s predictable.
Wall Street understands “5,000 seats.” It understands recurring revenue tied to people. It understands renewals that look like last quarter plus a bit extra.
Consumption is different. It moves with seasonality, marketing campaigns, outages, product launches, and the very efficiency gains AI is supposed to create. It can be fairer. It can also be spikier.
That is why Zeus Kerravala’s framing lands: legacy vendors are often carrying “Frankenstein” portfolios stitched together through acquisitions, and those portfolios were priced to protect a seat-based revenue stream. AWS, by contrast, built Amazon Connect with a cleaner slate and can experiment with consumption models without destabilizing a legacy licensing base. That asymmetry matters.
“They… Frankenstein their contact center… [AWS has] an inherent advantage… because they can play around with pricing… If you were a standalone company, that would be very detrimental to the business.”
The uncomfortable implication is that pricing innovation is not just about customer fairness. It’s about which companies can survive the transition without triggering investor panic.
The Hybrid Model Emerging in CCaaS Pricing: Base Platform + Metered AI
Pure per-seat is becoming deflationary.
Pure consumption risks bill shock.
So most roads lead to a hybrid.
You can see it in how the market now talks about packaging. Even when vendors keep a “seat” concept for human agents, AI features are increasingly layered with metered credits, tokens, minutes, or usage bands—sometimes in ways that make forecasting harder, not easier.
CX Today has already pointed to the limitations of seat models: they can be inflexible, penalize efficiency, and discourage innovation. The counter-proposal—consumption-based pricing—aligns cost with usage, but it also introduces volatility that procurement teams and finance teams are trained to resist.
This is where the next generation of models is going to converge:
- A base platform fee that pays for predictability and core capability.
- A metered AI component that reflects variable compute and variable value.
- A governance layer that prevents bill shock (caps, tiers, alerts, “burst” rules).
If this sounds like cloud economics arriving in customer service, that’s because it is.
“Seat-based pricing has become increasingly misaligned with the realities of modern customer service operations.”
Why “outcomes” keep showing up in the pricing conversation