The 2026 shift to outcome-based AI pricing

For years, support software pricing was simple: you paid per agent per month. Zendesk Suite at $55–$115/seat, Freshdesk at $19–$89/agent, Help Scout at $25–$75/user. The vendor's incentive was to keep you buying more seats. The model was predictable, even if wasteful.

That's changing. In April 2026, HubSpot cut Breeze from $1.00 per resolution to $0.50 per resolution—and, just as importantly, retied the charge to a resolved conversation rather than any conversation. It's a clear signal: the industry is moving toward outcome-based pricing. Zendesk, the category leader, acquired Forethought (a pure AI deflection play) in March 2026 and is now pushing customers toward resolution-based billing. Intercom Fin, launched in 2024, has anchored the category at ~$1.00 per resolved ticket. Drift, the original conversational AI platform, is being sunsetted (March 2026) in favor of its successor 1mind, which also uses per-resolution models.

This shift mirrors what happened in marketing automation (HubSpot cut email per-contact pricing years ago) and analytics (vendors moved from per-user to per-event). But for support ops, the implication is sharper: your bill now rises when your AI gets better at deflection. That's good for long-term incentives but bad for budget predictability—unless you model it first.

Why are vendors moving to per-resolution?

Three forces are converging. First, AI deflection is finally mature enough to measure reliably. A resolution—ticket closed by bot, no escalation—is a concrete event, not a fuzzy metric. Vendors can price it confidently.

Second, per-seat pricing was misaligned with actual value. An agent on payroll works full-time. An AI bot that deflects 50% of tickets in a team of 10 might replace 3–5 agent-equivalents but was still billed as one "seat" or one user account. Vendors left money on the table; customers with high deflection rates got a deal.

Third, the proliferation of AI options forced a reckoning. When every vendor added an AI chatbot, per-seat pricing became a commodity race. Margins compressed. Zendesk's acquisition of Forethought signals a move upmarket: bundling AI with resolution-based pricing lets Zendesk charge more to customers who actually use deflection, instead of subsiding non-users.

From the vendor's perspective: per-resolution aligns incentives. The better the AI, the higher your bill—which sounds wrong until you realize it means vendors will invest in actual deflection quality, not just gate-keeping it behind opaque sales workflows.

What the current benchmark prices tell you

HubSpot Breeze at $0.50/resolution is now the low anchor. It's cheap per event, but Breeze is a chatbot for simple queries (order status, password resets, returns). Most teams won't hit >500 resolutions/month without heavy optimization.

Intercom Fin at ~$1.00 per resolution is the category benchmark. It's more sophisticated than Breeze (can handle billing disputes, complex refund logic) and so typically resolves a higher-complexity mix of tickets. The math: if you get 40% deflection on 1000 tickets/month, Fin costs $400. That's cheaper than a $2,500+ per-seat plan upgrade but riskier if deflection drops.

Zendesk's pricing post-Forethought is still opaque—the vendor hasn't published a clear per-resolution rate—but sales will likely position it at or above Intercom, anchored by Zendesk's installed base and integration depth.

Gorgias (Shopify integration, e-commerce focused) charges ~$1.00 per automated ticket, with a risk: it can double-charge if a ticket triggers both a ticket fee and an automation workflow. Read the fine print.

For context: Freshdesk AI jumped from ~$0.10 per 100 sessions to $49 per 100 sessions (roughly $0.49 per session) around late 2025–early 2026. Freshdesk calls these "sessions," not resolutions, which is less transparent; ask them to clarify what counts as a session.

The math: when per-resolution is cheaper (and when it isn't)

Per-resolution beats per-seat when your deflection rate is high and stable. Here's the calculation:

Scenario A: 1,000 tickets/month, 35% deflection rate. Intercom Fin at $1.00/resolution = $350/month. A per-seat Freshdesk instance (3 agents) = $19–$89 per agent × 3 = $57–$267/month. Per-resolution wins, even though it costs more per event, because you're only paying for the deflections. But if your deflection rate drops to 15%, Fin costs $150/month—still competitive, but now a seat-based plan at $50/seat per agent × 2 agents = $100/month starts to look reasonable.

Scenario B: 500 tickets/month, 20% deflection rate. Intercom Fin = $100/month. A 2-seat per-agent plan at $40/seat = $80/month. Per-seat wins. You're paying for agents regardless; the AI is an add-on, not the primary lever.

The inflection point is roughly: deflection rate × tickets/month × per-resolution price vs. (agents needed – deflection_reduction) × seat_price. Use the billing model picker to run your numbers.

The hidden gotchas

Per-resolution pricing introduced new gotchas that per-seat pricing never had:

• Better AI = higher bill. If you upgrade your AI and deflection jumps from 30% to 50%, great for customers, worse for your budget. You need to forecast this and get buy-in from finance.
• Billing failures as resolutions. Some platforms (Ada, for instance) charge per interaction, even if the bot fails and escalates. Ask the vendor: does a failed bot attempt count as a resolution or a chargeable event?
• Seasonal variability. If your support volume spikes in Q4, your per-resolution bill spikes too. Per-seat pricing at least spread the cost evenly.
• Opacity around "resolution." Zendesk and Salesforce don't publish their resolution rates or prices; they hide them behind sales. You won't know the cost until you're in the contract.
• Double-charging. Gorgias can charge both a per-ticket fee and an automation workflow fee on the same ticket. Clarify the billing model in writing before signing.

The M&A context and what it means for you

Zendesk's acquisition of Forethought (March 2026) is the signal. Zendesk dominates the support category by install base, but it's losing share to smaller, AI-native competitors (Intercom, Gorgias, niche players). Forethought's value isn't the technology—Zendesk can build that—it's the pricing model and the go-to-market strategy. Expect Zendesk to migrate more customers to per-resolution pricing over the next 12–18 months.

Drift's sunset (March 2026, successor 1mind) signals that per-conversation or per-interaction models are less durable than per-resolution. Drift relied on Slack and email integrations to stay sticky; per-interaction pricing made it hard to scale beyond early adopters. 1mind is betting on per-resolution to compete with Intercom and Zendesk.

For you as a buyer: if your vendor is moving toward per-resolution pricing, that's a signal they're optimizing for deflection, not seats. It's usually good—you pay for value—but it means you need to lock in an SLA or cap on per-resolution pricing, or a blended model (per-seat + per-resolution) to hedge against volatility.

What this shift means for your budget

Per-resolution pricing is a sign that the support AI market is maturing. It aligns vendor incentives with actual deflection, which is good for long-term innovation. But it shifts forecasting risk to you: instead of budgeting a fixed cost per agent, you're betting on a volume metric (deflection rate) that can shift quarter to quarter.

Before you commit, model your costs under both pricing models, confirm the vendor's definition of a resolution in writing, and ask for a 6-month historical benchmark if they have it. If your deflection rate is stable and above 30%, per-resolution pricing is likely cheaper. If it's volatile or under 20%, a blended model or per-seat pricing might be safer.

And if your vendor hasn't published a price for per-resolution, that's a red flag. Push back in sales. Pricing opacity is a sign the vendor doesn't yet have confidence in the model—or they're betting you won't shop around.