What 2,000 AI-resolved tickets actually cost: Fin vs flat-rate

I build PilotPM, which competes with Intercom's Fin, so read everything below knowing I have a horse in this race. I'll try to earn your trust the boring way: with the billed definitions, a worked example you can recompute yourself, and an FAQ that includes the answers that don't help me.

Here's the short version: $0.99 per AI resolution is one of the best-sounding prices in software. Whether it's a good price for you depends almost entirely on two things most teams don't check until the second invoice — what counts as a "resolution," and what else stacks on top of it.

What "a resolution" means on the invoice

Fin is billed at $0.99 per resolution. The word does a lot of work in that sentence.

Per ClearFeed's teardown of Intercom's pricing — the most careful public breakdown I've found, and the one I cite on my own pricing page — a billed resolution isn't only "the customer said thanks, that fixed it." The billed definition also includes assumed resolutions: the customer stops replying after Fin answers, and the conversation is counted as resolved. Silence is billable. Anyone who has run a support queue knows that silence sometimes means "solved" and sometimes means "gave up and churned" — the invoice doesn't distinguish.

Handoffs are the other edge. A conversation where the AI answers part of the question and a human finishes the job can still register an AI outcome along the way. You end up paying the per-resolution fee and the loaded cost of the human who actually closed it.

None of this is hidden — it's in the docs. But "$0.99 per resolution" and "$0.99 per conversation where the AI answered and the customer didn't reply" are different products, and most buyers price the first while signing up for the second.

The stack: seats, add-ons, channels

The per-resolution fee is also not the bill. It sits on top of:

• Seats. Fin rides on Intercom, and your humans need seats — billed per person, per month, on whichever Intercom plan you're on.
• Add-ons. Proactive support, advanced reporting tiers, copilot for your agents — the useful stuff has a way of living one add-on up.
• Channel fees. Email, WhatsApp, SMS carry their own metering depending on configuration.

Each line is individually reasonable. The stack is what surprises people.

The worked example

Take a concrete, realistic mid-size team: 5 support seats, 2,000 AI outcomes a month. That's not a whale — that's a healthy B2C startup or a mid-market B2B team with decent volume.

Working from ClearFeed's breakdown (as of July 2026), that team on Intercom + Fin lands at roughly:

• ~$1,980/month in per-resolution fees (2,000 × $0.99)
• plus seats for 5 people
• plus the add-ons a real deployment tends to carry

Total: ≈ $2,745/month, or about $33,000 a year — before channel fees.

You should absolutely recompute this for your own seat count and volume; the point isn't the third significant digit. The point is that a team that budgeted "$0.99 a ticket" mentally files this as a ~$2k/month product and gets a ~$33k/year line item.

The part that bothers me more than the number

Per-resolution pricing has a property that I think is genuinely backwards: your bill grows as the AI gets better.

Deflect 40% of tickets this quarter and 55% next quarter — same ticket volume, better AI, bigger invoice. You are, in effect, paying a success tax on the thing you bought the product to do. And because the billed definition includes assumed resolutions, the "success" being taxed is partly just customers going quiet.

Flat pricing inverts that. The vendor absorbs the cost of the AI improving, and improvement shows up as margin for you — fewer escalations, faster queues — instead of as revenue for the vendor.

That's the actual argument, more than any single number: per-resolution pricing misaligns you and your vendor on the exact metric you're both trying to move.

What the same team pays flat

On PilotPM, that 5-seat, 2,000-outcome team fits comfortably in Growth: $2,499/month, flat. That includes 20 seats (not 5), 10,000 conversations, 15,000 AI credits a month, and the AI features — drafting, autopilot, account-grounded answers, the ops layer — without a per-outcome meter running underneath. If your AI handles 2,000 conversations or 3,500, the invoice is the same.

Smaller team? Starter is $149/month for 5 seats and 1,000 credits. A team doing a few hundred AI outcomes a month has no business paying enterprise money, including to me.

Yes: $2,499 vs ≈$2,745 is not a dramatic monthly gap. The gap is in the shape. One number is a ceiling; the other is a floor that rises with your deflection rate.

When per-resolution pricing is actually fine

Honestly: sometimes.

• Low volume. At 200 outcomes a month, Fin's meter is ~$200. If you're already paying for Intercom seats anyway and volume is genuinely small, the meter is the cheaper model. Flat tiers only win once volume is real.
• Spiky or seasonal volume. If ten months a year are quiet, paying per outcome in the two loud months can beat carrying a flat tier year-round.
• You live in Intercom. If your team runs on Intercom's inbox, product tours, and messenger, and you're happy there, switching platforms to save a few hundred dollars a month is rarely worth the migration.

If that's you, keep the meter. Pricing models aren't moral positions; they're fit questions.

Honest FAQ

Is Fin bad? No. Fin is a genuinely capable AI agent and Intercom has real scale behind it. This piece is about the pricing model, not the model quality.

Are you cheaper for everyone? No. Below roughly 1,000 outcomes a month, a metered bill can undercut my Growth tier, and if you only need Fin-style deflection with no ops layer, compare against Starter at $149 and do the math.

Doesn't PilotPM have limits too? Yes — flat tiers include stated allowances (Growth: 10,000 conversations, 15,000 credits). The difference is the failure mode: you see a number approaching a ceiling, not a meter that bills silence.

Where do your numbers come from? The Fin-side math is ClearFeed's published breakdown, as of July 2026. Pricing changes; check current pages before deciding, and recompute the worked example with your own seats and volume.

What should I actually do? Pull your last three months of resolved conversations, apply the billed definition (including assumed resolutions), and price both models. It's an hour of spreadsheet work against a $33k/year decision.