The Per-Seat Model Had a Good Run

For two decades, every support tool charged per agent per month. Zendesk at $55-115/seat. Freshdesk at $15-79/seat. Intercom at $29-132/seat (before their pivot). Salesforce Service Cloud at $25-300/seat. The logic was simple: more agents means more usage, more usage means more value, more value means more revenue.

This worked when agents were the bottleneck. If each agent handled 50-80 tickets per day, adding a seat meant adding capacity. The vendor's revenue scaled with the customer's team size, which roughly correlated with support volume.

Then AI started resolving tickets without agents.

The Math That Broke Per-Seat

Imagine a 10-agent team paying $100/seat/month for their helpdesk. That's $12,000/year. They handle 3,000 tickets per month across those 10 agents.

Now they deploy an AI resolution tool that handles 45% of inbound volume. 1,350 tickets per month never reach an agent. The team could drop to 6 agents and maintain the same response times for the remaining 1,650 tickets.

Under per-seat pricing, they'd save $4,800/year by cutting 4 seats. But here's the catch: the helpdesk vendor just lost 40% of their revenue from this customer. So the vendor has zero incentive to help you reduce headcount. Their business model punishes you for getting more efficient.

Intercom saw this collision coming. In 2024 they launched Fin at $0.99 per resolution, decoupling their revenue from agent count entirely. When Fin resolves a ticket, Intercom makes money. When an agent resolves a ticket, Intercom makes money through the seat fee. Either way, they get paid for work done rather than chairs occupied.

The Current Pricing Picture

The market has fragmented into three tiers.

Supp charges $0.20 per classification, $0.30 per resolution. No seat fees at all. Your entire team uses the dashboard for free. At 1,000 resolutions per month, that's $300. Purpose-built classifier, not an LLM wrapper, so the cost stays low at any volume.

Intercom's Fin charges $0.99 per resolution. At 1,000 resolutions, that's $990/month. They've committed to this model publicly and their growth numbers suggest it's working. Customers who previously paid $2,000/month in seat fees might pay $1,500/month in resolution fees, but they also eliminated 3 agents.

Zendesk launched their AI resolution pricing at approximately $1.50 per resolution (varies by plan). Higher than Intercom, but Zendesk's bet is that their enterprise install base won't switch easily. They're right about that, but the premium is hard to justify when competitors charge 80% less.

Sierra uses outcome-based pricing for their AI agents but doesn't publish rates. They target enterprise brands with contracts reportedly starting around $150,000/year.

Why This Shift Is Happening Now

Two things changed simultaneously. First, resolution quality hit a threshold. Early chatbots (2018-2022 vintage) had maybe 15-25% successful resolution rates. Customers hated them. Support leaders deployed them reluctantly and pulled them back when CSAT dropped. You couldn't charge per resolution when most resolutions were bad.

Current AI systems resolve 40-60% of qualifying tickets successfully, with CSAT scores within 5-10 points of human agents for those ticket types. That's good enough to charge for. If the resolution sticks (customer doesn't reopen, doesn't call back, doesn't leave a negative review), it delivered value equivalent to a human handling it.

Second, CFOs started doing the math. A VP of Support at a 200-person company paying $115/seat for 30 agents is spending $41,400/year on helpdesk software. If AI can handle half their volume, they could theoretically cut to 15 agents and save $20,700/year in seat fees alone. But under per-seat pricing, the vendor captures none of the AI value they're delivering. Per-resolution pricing lets the vendor say "we saved you $20,700 in headcount, pay us $8,000 in resolution fees instead." Both sides win.

Winners of This Shift

Small teams win big. A 3-person support team paying $300/month in seat fees might pay $90-150/month in resolution fees with Supp, depending on volume. They get AI-powered classification and resolution without the overhead of per-seat licensing that scales with headcount they're trying to avoid growing.

High-volume, low-complexity businesses win. E-commerce companies, SaaS with self-serve tiers, subscription boxes, anything where 50%+ of tickets are repetitive questions with known answers. These companies see the highest automation rates and therefore the biggest savings from per-resolution pricing.

Companies with seasonal spikes win. A retailer doing 3x normal ticket volume in November and December doesn't need to buy extra seats for temporary agents. Per-resolution pricing scales automatically. You pay more during the spike, less during slow months.

Losers of This Shift

Legacy helpdesk vendors with large per-seat install bases lose revenue. Zendesk's 170,000+ customers are almost all on per-seat plans. Migrating them to per-resolution pricing means short-term revenue loss, even if it's the right long-term move. This is the classic innovator's dilemma.

Companies with mostly complex tickets lose. If 80% of your tickets require nuanced, multi-step human resolution, per-resolution AI pricing gives you a bad deal. You're paying the AI fee for the 20% it can handle while still paying agents for the 80% it can't. Per-seat pricing might actually be cheaper in that scenario.

Outsourced support providers lose the most. BPOs charge per agent hour. Their entire business model is selling human time. Per-resolution AI pricing makes their markup transparent. Why pay a BPO $22/hour per agent when AI resolves the same ticket for $0.30?

The Transition Isn't Clean

Most companies won't switch overnight. Zendesk and Salesforce will keep per-seat pricing as their default for years, adding per-resolution AI features as optional add-ons. The hybrid period will be messy. You'll pay seat fees for your agents and per-resolution fees for the AI on top of that.

Supp avoided this entirely by starting with per-resolution pricing from day one. No seats to grandfather. No legacy pricing tiers to protect. $0.20 to classify, $0.30 to resolve. The simplicity matters because pricing complexity is how vendors hide cost increases.

Five years from now, per-seat pricing for support tools will feel as outdated as per-minute cell phone plans. The work is what matters, not the number of people logged in.