That single design choice is the reason two companies handling identical ticket volume can pay wildly different amounts. Below is the full 2026 pricing landscape, the hidden costs that do not show up on pricing pages, a worked example, and how to forecast your own bill before you sign anything.
The AI customer support pricing models in 2026
The market has split into a base license plus a metered AI layer. Understanding which model a vendor uses matters more than the headline number, because the models behave very differently as you grow.
| Model | How you are billed | Typical 2026 range | Who it favors |
|---|---|---|---|
| Per resolution | A fee each time AI resolves a conversation | $0.75 to $2.00 per resolution | Low-volume teams; gets expensive at scale |
| Per engaged conversation | A fee whenever AI replies, even if it escalates | ~$0.50 to $0.65 per conversation | Vendor, not buyer (you pay even on failures) |
| Per seat / per agent | Monthly license per human agent (AI added on top) | $19 to $210 per agent per month | Stable, small human teams |
| Per session | A fee per time-windowed AI interaction | ~$0.10 to $0.49 per session | High deflection on simple FAQs |
| Flat rate | One fixed monthly fee, all volume included | Fixed (Corebee: $99/mo, or $79/mo annual) | Predictable budgets; scales for free |
A critical distinction that pricing pages blur: per resolution means you only pay when the AI actually solves something. Per engaged conversation means you pay whenever the AI speaks, even if it immediately hands off to a human. Kustomer and Gladly use the engaged-conversation model at around $0.60 per conversation, so you can be charged for AI that fails to resolve anything.
What the major vendors actually charge
Here is the verified per-vendor breakdown as of mid-2026. Plan prices are annual-billing rates per agent unless noted. AI usage fees are charged on top of the base plan.
| Vendor | Base plan (per agent/mo) | AI billing | AI unit price |
|---|---|---|---|
| Intercom / Fin | Seats from $29 (annual) | Per outcome | $0.99 per resolution; 50-outcome/mo minimum standalone |
| Zendesk | Support Team $19, Suite Pro $115 | Per automated resolution | ~$1.50 to $2.00 (third-party estimate; not published) |
| Freshdesk | Growth $19, Pro $55, Enterprise $89 | Per session | ~$0.10 to $0.49 per session, prepaid packs |
| Help Scout | Standard $25, Plus $45, Pro $75 | Per resolution | $0.75 per resolution (AI Answers) |
| Gorgias | Per ticket: Starter $10 to Advanced $900 | Per resolution | $0.90 (annual) / $1.00 (monthly) per resolution |
| Tidio | Starter $24.17, Growth $49.17+ | Per conversation | ~$0.50 to $0.65 per Lyro conversation |
| Zoho Desk | Express $7 to Enterprise $40 | Mostly bundled | Autonomous resolution gated to $40 tier |
| Gladly | ~$180 to $210 (10-seat minimum) | Per conversation | ~$0.60 per conversation (estimate) |
| Kustomer | Enterprise $89, Ultimate $139 | Per engaged conversation | ~$0.60 per engaged conversation |
| Corebee | None (no seats) | Flat | $99/mo, or $79/mo billed annually |
Note on sourcing: Intercom/Fin, Help Scout, Zendesk, Gorgias, and Tidio base prices are confirmed on the vendors' own pricing pages. Zendesk's exact per-resolution dollar figure, plus Gladly and Kustomer numbers, come from third-party analyses because those vendors use contact-sales pricing. Always get a binding quote.
The hidden costs nobody puts on the pricing page
The sticker price is rarely the real price. Three traps inflate AI support bills in practice.
Double billing on a single interaction
Intercom and Gorgias can charge you twice for one AI interaction. With Intercom you pay for the seat AND the $0.99 Fin outcome. With Gorgias, an AI resolution can also count as a billable helpdesk ticket, so one automated reply incurs both a ticket charge and the $0.90 to $1.00 AI fee. You are paying for the infrastructure and the outcome separately.
Paying for AI that does not resolve
Per-engaged-conversation vendors (Kustomer, Gladly, Tidio outside Premium) bill when the AI engages, regardless of whether it solved anything. If your AI escalates 40% of conversations to humans, you still pay the per-conversation fee on all of them, then pay your human agents to finish the job.
Success becomes a tax
The cruelest dynamic of per-resolution pricing: the better your AI performs, the more you pay. A vendor that resolves 76% of volume end-to-end, as Fin reports, bills you for every one of those resolutions. Higher deflection should lower your cost. Under per-resolution pricing it raises your bill. Expiring prepaid AI packs (Freshdesk) and per-agent AI copilot add-ons ($29 to $50 per agent per month at Zendesk and Freshdesk) pile on further.
A worked cost example: 5,000 resolved tickets a month
Take a small but real scenario: a support team with 5 human agents and 5,000 conversations a month that AI resolves without escalation. Here is the rough annual math.
| Vendor | Base (5 agents, annual) | AI resolutions (5,000/mo) | Estimated annual total |
|---|---|---|---|
| Help Scout (Plus $45) | $2,700/yr | $0.75 x 5,000 x 12 = $45,000 | ~$47,700/yr |
| Intercom / Fin (5 seats $29) | $1,740/yr | $0.99 x 5,000 x 12 = $59,400 | ~$61,140/yr |
| Gorgias (Pro $360/mo flat) | $4,320/yr | $0.90 x 5,000 x 12 = $54,000 | ~$58,320/yr |
| Zendesk (Suite Pro $115) | $6,900/yr | ~$1.75 x 5,000 x 12 = $105,000 | ~$111,900/yr |
| Corebee | $0 (no seats) | $0 (flat) | $948/yr (annual) |
These are directional estimates using the documented unit prices, not quotes, and they ignore add-ons and overage tiers that would push the metered vendors higher. The point is structural, not precise to the dollar: at moderate volume, per-resolution pricing crosses into five figures fast, while a flat plan stays flat. The more your AI succeeds, the wider the gap grows.
For a head-to-head on one vendor pair, the dynamics in the Freshdesk vs Zendesk pricing breakdown show how even two "similar" platforms diverge once AI usage is included.
How to forecast your real AI support bill
Do not trust the per-unit number in isolation. Run this checklist before signing.
- Count your monthly resolvable volume. Pull the count of conversations your AI could realistically close (FAQ, order status, returns, account questions). This is your billable base under per-resolution and per-conversation models.
- Find your true unit. Confirm whether the vendor charges per resolution, per engaged conversation, or per session. The same headline price means very different bills. Per engaged conversation almost always costs more because it includes failures.
- Add the base license. Per-seat vendors charge the monthly license on top of every AI fee. Multiply by agent count and by 12.
- Add the add-ons. Copilot or agent-assist seats ($29 to $50 per agent per month), channel fees, and prepaid pack expiry all stack.
- Model your growth case, not today. Project 12 to 24 months out. Per-resolution bills scale linearly with both volume and AI quality. A flat plan does not move.
- Watch for double billing. If a resolution also counts as a ticket or requires a seat, count it twice.
When you run this on a metered vendor, the bill is a moving target tied to your busiest months and your best-performing AI. That unpredictability is the actual cost, separate from the dollar amount.
Where flat-rate fits
Flat-rate pricing inverts the incentive. Corebee charges $99/mo (or $79/mo billed annually, which is $948/year) with no per-seat, per-agent, or per-resolution fees. Its 3-tier AI (Instant, Pattern, and Deep) reaches up to 86% average auto-resolution across web chat, WhatsApp, and email, with a RAG knowledge base that auto-learns from your docs. Because nothing is metered, doubling your ticket volume or improving your resolution rate does not change the bill.
The trade-off is real and worth stating plainly: flat-rate makes the most sense when your volume is high enough that per-resolution fees would exceed a fixed plan, or when budget predictability matters more than paying exactly per unit at very low volume. If you handle a handful of tickets a month, a metered plan with a small minimum may be cheaper in absolute terms. The crossover comes quickly, usually within the first few thousand monthly resolutions.
For the broader market context behind this shift, see why Zendesk acquired Forethought and how ticket deflection changes the unit economics of support. You can compare full feature sets on the features page and the pricing breakdown.
The bottom line
In 2026, the question is not just "how much does AI customer support cost" but "what happens to that cost when volume grows and AI gets better." Under per-resolution and per-seat models, both growth and success raise your bill. Under flat-rate, neither does. Map your resolvable volume against each model's true unit, project forward, and the right structure for your team becomes obvious.
If predictable beats metered for your team, you can try Corebee free for 14 days, no credit card required, with a 30-day money-back guarantee.