AI Customer Support Cost: The Real Per-Resolution Math

The short answer: AI customer support is billed per resolution, not per seat, and the market rate sits around $0.99 per resolved conversation. SupportWire charges $0.49. At a typical mid-size volume, that one number is the difference between a $6,435 monthly AI bill and a $3,185 one.

But the per-resolution price is only one of three numbers that decide what you actually pay. Let's run all three on a real model.

Jump to a section:

• What does AI customer support actually cost?
• What three numbers set your AI support bill?
• Running the real ROI math
• Why is a free AI agent often the most expensive option?
• The part the math doesn't stop at
• Calculate your own number

---

What does AI customer support actually cost?

Old support pricing was easy to reason about. A seat cost a flat monthly fee, and one human handled some number of conversations a day. AI changed the shape of that bill. Once software can resolve a conversation on its own, the cost of each one falls toward zero, and pricing shifts from per-seat to per-outcome. We dug into what that does to a whole support org in why support is no longer a cost center.

It is also why most modern AI support tools, Intercom's Fin included, charge per resolution. You pay when the AI fully closes a conversation on its own. If it hands off to a human, you do not pay the resolution fee for that one.

So your bill is no longer "seats times price." It is closer to this:

(AI resolutions x price per resolution) + (human resolutions x human cost per resolution)

Two of those four variables are set by your volume and your AI's quality. The third, price per resolution, is set by your vendor. That is the one worth looking at hard.

What three numbers set your AI support bill?

Every AI support cost comes down to three levers:

1. Involvement rate is the share of inbound conversations the AI touches at all.
2. Resolution rate is the share it fully closes without a human. This is the number that actually matters, and the AI playbook for support teams goes deep on how to lift it. A conversation the AI does not resolve still flows to your team and carries the full human cost.
3. Price per resolution is what you pay for each one it does close.

Vendors love to talk about the first two because high resolution rates sell. Almost nobody puts the third one next to a calculator. So let's do that.

Running the real ROI math

Conveniently, Intercom published the ROI model itself, inside its own Fin AI Agent blueprint. It runs on 10,000 monthly conversations, a 65% resolution rate, $0.99 per AI outcome, and a fully loaded human cost of $6.60 per resolution. We did not invent any of these numbers. They are Fin's.

Here is that exact model, with SupportWire's $0.49 dropped into the one cell that changes:

Variable	Fin ($0.99)	SupportWire ($0.49)
Monthly conversation volume	10,000	10,000
AI resolution rate	65%	65%
AI resolutions	6,500	6,500
Price per resolution	$0.99	$0.49
AI cost	$6,435	$3,185
Human resolutions	3,500	3,500
Human cost per resolution	$6.60	$6.60
Human support cost	$23,100	$23,100
Total monthly cost	$29,535	$26,285

Nothing else changed. The volume, the resolution rate, and the human cost are all identical between the two columns. Only the price per resolution dropped, and that alone took $3,250 off the monthly bill. Over a year, on the AI line by itself, it adds up to $39,000 for the exact same outcome.

And that is before Fin's Monitor add-on, which charges another $99 a month to tell you whether those resolutions were any good. SupportWire scores every conversation on every plan, no add-on.

Why is a free AI agent often the most expensive option?

Now, the obvious objection: if cheaper is better, why not use a free AI agent and pay $0 per resolution?

Because the per-resolution price is only worth comparing at the same resolution rate. Intercom makes this exact point in its blueprint, and it is correct: an AI that costs $0 but resolves 30% of conversations is worse than one that costs $0.99 and resolves 50%, because every unresolved conversation still lands on your human team at full cost.

Run it:

• Free AI, 30% resolution: 3,000 AI resolutions at $0, plus 7,000 human resolutions at $6.60 = $46,200/month.
• Paid AI, 50% resolution: 5,000 AI resolutions at $0.49, plus 5,000 human resolutions at $6.60 = $35,450/month.

The "free" option costs $10,750 more every month. What you are actually buying is the resolution, and the sticker price tells you almost nothing about it.

The part the math doesn't stop at

Here is where the blueprint logic quietly works in SupportWire's favor. Once you accept that resolution rate is what matters and you have picked an AI that resolves at a high rate, the per-resolution price becomes pure margin. There is no quality tradeoff left to make. You are paying more for the identical outcome.

So the honest framing is simple. Pick the AI that resolves well, then pay as little as you can for each of those resolutions. SupportWire's AI agent resolves at the same quality tier as Fin and charges half, so you get Fin's own "outcomes over unit cost" argument and the lower unit cost layered on top. Paying double for an identical resolution was never the clever move.

For the full feature-by-feature breakdown, see Intercom vs SupportWire.

Calculate your own number

The 10,000-conversation model is illustrative. Your volume, resolution rate, and human cost are your own. Plug them in:

• AI support pricing calculator runs the per-resolution math for your numbers.
• Annual cost vs SupportWire TCO projects it over a year, seats included.

Or just look at the pricing: per-seat plus $0.49 per resolution, every feature unlocked, and no add-ons waiting to surprise you on the invoice. Cutting that AI line in half is the whole reason we built SupportWire in the first place.