If your AI support tool bills per resolution, ask yourself a simple question: who decides what counts as a resolution? Not you. The vendor does. And the moment that definition lives inside a black box, your monthly bill becomes a guess. I got tired of explaining surprise invoices to small teams, so I published our billing definition on the pricing page. The awkward conversations stopped.
The short version
- The AI customer service market is about $15 billion in 2026 and growing roughly 23 to 26 percent a year (Solid Market Research, 360iResearch).
- Intercom's Fin charges $0.99 per resolution, and "resolution" is defined by Intercom, not by you.
- Cost per contact runs about $1.84 self-service versus $13.50 agent-assisted. AI is cheaper. The billing model is what hurts.
- A "resolution" can be counted when the customer just stops replying, so you pay for silence.
- We switched to a flat month with a published definition, and overage means a ticket instead of a shutdown.
Confession: I used a per-resolution tool for a year and never once predicted the invoice. I blamed the product. The product was working exactly as designed, the design just was not for me.
Caveat: flat pricing is not always cheaper. If you run 50,000 resolutions a month, a per-outcome model may win. This post is for the team doing 100 to 2,000 a month that just wants the number to not move.
Per-resolution is the default now
Intercom's Fin charges $0.99 per resolution. A resolution, per their help docs, happens "when Fin answers a customer's question using your content or catalog data, the customer doesn't provide negative feedback, and the conversation needs no further follow-up." That sounds reasonable in a sentence. In a month it becomes a variable you cannot see coming, because it scales with how many questions your customers actually ask, which is exactly the thing you are trying to grow.
The definition is the vendor's, not yours
Read that definition again. "No negative feedback." "No further follow-up." Vendors also count a resolution when the customer simply exits without replying, the conversation is assumed solved because nobody objected. You paid for a question that may have gone nowhere. The definer and the biller are the same company, and you are not in the room.
This is the part nobody puts on a pricing page. The model is "pay for value," which is fair in theory. In practice the value metric is defined by the party sending the invoice.
Why this hides the bill
Review aggregates on G2 and Capterra repeatedly flag the same complaint about per-resolution support: the monthly invoice is hard to forecast, and costs climb fast once deflection volume grows. A five-seat team can land an all-in bill near $640 before any add-ons, and the seat cost is the small line. Usage is the one teams underestimate. None of that is fraud. It is just a number that moves with your traffic, and traffic is the goal.
What we did instead: a published billing definition
For ClearDesk (launching this month), the pricing page states the definition in plain language:
- A conversation ends after 30 minutes with no new customer message. That is one unit. Not per question, not per "outcome."
- The monthly price is flat: Indie $9, Growth $29, Scale $79. The bill in March is the bill in August.
- Hit your tier's session cap and you downgrade to a ticket, answered next business day. The widget does not shut off. Nobody loses a customer to a paywall at 2am.
The point is not that flat is morally superior. It is that the customer can read the math before they buy, and the math does not change when they succeed.
What it is not
It is not a per-resolution killer for enterprise volume. At 50k+ resolutions a month, outcome pricing may beat a flat fee, and we say so. It is not "unlimited" with a fair-use asterisk buried in the terms, the caps are on the page. And it is not free of AI limits; the human handoff is a ticket, not a live agent hovering. We promise the handoff, not a person on the line.
FAQ
What does "resolution" mean in most AI support tools?
In per-resolution models like Intercom Fin, a resolution is counted when the AI answers from your content, the customer gives no negative feedback, and no follow-up is needed, and vendors also count a conversation resolved when the customer simply stops replying. The definition sits with the vendor, so the invoice can move with your traffic.
Why did you switch to flat pricing for ClearDesk?
Because the teams we serve (100 to 2,000 conversations a month) kept telling us the surprise bill was the reason they hesitated. A flat month with a published definition removes the guess. It is not always cheaper at enterprise scale, and we say that plainly.
What happens when I hit my session cap on ClearDesk?
You downgrade to ticket mode. The AI stops auto-answering and your customer's message becomes a ticket answered next business day. The widget stays live. We do not shut off support to force an upgrade.
Is multilingual support included in the flat price?
Yes. Six languages including Arabic (right-to-left) are first-class in every paid tier, at no extra charge. That is a separate bet from billing, see the RTL piece, but it is not a per-language upsell.
References
- AI in Customer Service Market 2026 — Market size $15.12B in 2026, ~25.8% CAGR; self-service $1.84/contact vs $13.50 agent-assisted.
- Intercom Help: Set up Fin — Fin bills $0.99 per resolution; the vendor-defined resolution criteria.
- Drift vs Intercom in 2026 — Per-resolution billing makes the invoice hard to forecast; G2/Capterra review aggregates on the same complaint.
- AI Customer Service Market: $15.12B in 2026 — Category composition; pure-play per-resolution spend is ~$4 to 5B of the $15B.
Where these numbers come from
Every market figure above is attributed to a linked source: Solid Market Research, 360iResearch, Polaris (via CRM Curator), Intercom's own help docs, and aistackguides' aggregation of G2/Capterra reviews. The ClearDesk pricing ($9/$29/$79, 30-minute conversation unit, ticket downgrade) is our own product decision, not an industry rate, and the site launches this month at cleardesk.lxsaihub.com.
Top comments (0)