The math is good. For Intercom.
We sat down and did the numbers at three realistic volumes for a typical six-seat support team. At 500 monthly resolutions, you're looking at $1,005/month minimum. At 2,000 resolutions, $2,490. At 5,000 resolutions, $5,460. That's before any usage spike or seat expansion.
This post walks through each band with the formulas open, compares Fin against HubSpot Breeze and Zendesk's per-resolution model, and shows where flat-fee pricing wins. We'll also concede where Eoghan is right.
TL;DR: Intercom Fin's $0.99 per outcome plus $85/seat creates a per-resolution cost slope that punishes growing SMBs. At 2,000 monthly resolutions, six seats run $2,490/month before overages (Intercom, 2026). For volumes above 1,000 resolutions, flat-fee alternatives produce predictable bills and stop penalising scale.
What does Intercom Fin actually charge?
Intercom Fin charges $0.99 per AI outcome on top of standard seat pricing, with the Advanced plan starting at $85 per seat per month (Intercom, 2026). There's a documented 50-resolution monthly minimum on the agent side (MyAskAI, 2026). That means the cost has two slopes layered together.
How does Intercom define an "outcome"?
Fin charges only when the AI resolves a query without human escalation. Intercom's product data shows Fin reports a 67% resolution rate over the last 30 days across 40 million resolved conversations (Intercom Fin product data, 2026). If Fin escalates to a human, you're not billed for that interaction. That's the steel-man case for the model.
What's the minimum monthly Fin spend?
Independent reviewers document a 50-resolution monthly minimum on Fin's agent product (MyAskAI, 2026). At $0.99 each, that's roughly $49.50/month before any seats. Real teams clear that floor in the first week, so the floor doesn't matter much. The seat math does.
Intercom Fin charges $0.99 per AI outcome plus $85 per seat for the Advanced plan, with a documented 50-resolution monthly minimum (Intercom, 2026; MyAskAI, 2026). Fin only charges when the AI resolves a ticket without human escalation, which the product reports happens 67% of the time across 40M+ conversations.
What does Fin cost a six-seat team at 500 monthly resolutions?
A six-seat team running 500 monthly resolutions on Intercom Fin pays $1,005/month: $510 in seat fees plus $495 in outcomes (Intercom, 2026). That's the entry-level math, and it already exceeds many flat-fee alternatives that include unlimited resolutions in their base tier.
We modelled this against three flat-fee competitors and one per-resolution alternative. The numbers below assume 500 successful Fin outcomes -- not total conversations. Fin's documented 67% resolution rate means 500 outcomes implies roughly 750 attempted conversations.
The 500-resolution formula
Here's the math, line by line:
- 6 seats x $85 = $510 base
- 500 outcomes x $0.99 = $495 in usage
- Total: $1,005/month
Any seat expansion adds $85. Any outcome above 500 adds $0.99. There's no ceiling.
How does that compare?
HubSpot Breeze charges $0.50 per resolution (HubSpot, April 14 2026). At 500 resolutions, Breeze runs $250 in usage plus seat costs. Zendesk's AI agents committed tier sits at $1.50 per resolution with a $2.00 overage (Zendesk, 2026), making it more expensive than Fin at this volume. Corebee's flat $99 plan covers up to 2,000 resolutions with all seats included.
What does Fin cost at 2,000 monthly resolutions?
At 2,000 monthly resolutions, that same six-seat team pays $2,490/month on Intercom Fin: $510 in seat fees plus $1,980 in usage charges (Intercom, 2026). This is the band where the slope problem becomes obvious. You quadrupled your AI workload but your bill more than doubled.
We've watched three Corebee customers migrate off Fin in this volume band. The common pattern: their support volume grew faster than headcount, the Fin bill kept climbing, and finance asked uncomfortable questions about per-ticket cost.
The 2,000-resolution formula
- 6 seats x $85 = $510
- 2,000 outcomes x $0.99 = $1,980
- Total: $2,490/month
That's $1.245 per resolution all-in once you fold seat costs back into per-resolution math. The marketed price is $0.99. The real number is 25% higher.
Why the slope hurts here
Per-resolution pricing rewards low volume. Once you cross roughly 1,000 monthly resolutions, every $0.99 charge stacks against you while the platform's marginal cost stays flat. Intercom isn't paying $0.99 in inference for each resolution -- nobody is. The margin widens as you scale.
Flat vs per-resolution vs per-seat economics
What does Fin cost at 5,000 monthly resolutions?
A six-seat team running 5,000 monthly resolutions on Fin pays $5,460/month: $510 in seats plus $4,950 in usage (Intercom, 2026). That's $65,520 annually for the AI layer alone. At this volume, finance teams start asking whether the AI agent is a product or a tax.
The 5,000-resolution band is where most pricing comparisons stop being academic. It's also where the Klarna case study becomes relevant -- and where outcome pricing's fairness argument starts to wobble.
The 5,000-resolution formula
- 6 seats x $85 = $510
- 5,000 outcomes x $0.99 = $4,950
- Total: $5,460/month
Independent benchmarking from MyAskAI's 2026 guide put Fin's resolution rate at 73% versus Decagon's 49% in a head-to-head test (MyAskAI, 2026). Fin is genuinely good at the AI part. The pricing is the issue, not the product.
What flat pricing looks like at this volume
Corebee's flat $99/month plan caps usage but most teams in this band sit on a custom tier closer to $499-$999 with unlimited resolutions and unlimited seats. Even at the high end, that's an 80%+ saving against Fin. The slope difference compounds every month volume grows.
At what volume does Fin actually become cheaper than flat?
For an SMB running 6-12 seats, Intercom Fin never becomes cheaper than well-designed flat-fee alternatives within realistic volume ranges (Corebee analysis, 2026). The breakeven sits below the practical floor of either tool. If you're under 100 resolutions a month, you probably don't need an AI agent layer at all.
The full breakeven table across five volume bands
We modelled four pricing structures against five volume bands, holding team size at six seats. Each row shows the all-in monthly cost for that volume. The Fin column folds in the $510 seat floor, the Zendesk column folds in the platform seat cost ($115 per seat for the Suite Professional tier), and the flat column stays put at $99 regardless.
| Volume | Fin ($0.99/res) | Breeze ($0.50/res) | Zendesk ($1.50/res + $115 platform x 6) | Corebee Flat |
|---|---|---|---|---|
| 100 res/mo | $510 + $99 = $609 | $50 | $1,050 | $99 |
| 500 res/mo | $510 + $495 = $1,005 | $250 | $1,440 | $99 |
| 2,000 res/mo | $510 + $1,980 = $2,490 | $1,000 | $3,690 | $99 |
| 5,000 res/mo | $510 + $4,950 = $5,460 | $2,500 | $8,190 | $99 |
| 10,000 res/mo | $510 + $9,900 = $10,410 | $5,000 | $15,690 | $99 |
[ORIGINAL DATA] These figures assume a six-seat Advanced Intercom team, HubSpot Breeze's announced $0.50 rate, and Zendesk's per-resolution committed tier added to a Suite Professional six-seat platform cost. The Corebee column reflects the published flat $99 plan with seats included.
What the curves actually tell you
Flat pricing wins at essentially every volume band a real SMB encounters. At 100 resolutions per month, only Breeze beats flat in absolute terms ($50 versus $99), and only because Breeze's seat costs aren't folded in. Cross 500 resolutions and the picture inverts: Fin is ten times the cost of flat, Zendesk is fifteen times, and even Breeze loses its lead once you account for HubSpot's required hub seats.
Where Breeze beats Fin and where it doesn't
Breeze beats Fin on raw per-resolution cost across every band ($0.50 versus $0.99). That gap matters most at high volume: at 10,000 resolutions, Breeze's $5,000 monthly bill is half of Fin's $10,410. At low volume (around 100 resolutions), Breeze actually beats both flat and Fin if you ignore platform seat costs. The trade-off is feature breadth, not price.
The structural point about meters
Every metered model scales with success. The more tickets your AI resolves, the more you pay. Flat pricing breaks that link. A team that grows from 500 to 5,000 resolutions on Fin watches its bill jump from $1,005 to $5,460, a 5.4x increase. The same team on flat pricing pays $99 either way. [UNIQUE INSIGHT] Outcome pricing markets itself as buyer-aligned but actually penalises the buyer's growth. Flat pricing is the only structure where the vendor's incentive to ship a good product isn't tied to your invoice.
Where outcome pricing does win
Outcome pricing wins on perceived risk, not actual cost. Buyers feel safer paying only for resolutions, even when the math is worse. Eoghan's Bloomberg argument leans on this: alignment of incentives is psychologically valuable even when the spreadsheet disagrees (Bloomberg, 2026).
That's a fair point. We're going to give it more space than most competitors would.
Where is Eoghan McCabe right about outcome pricing?
Eoghan McCabe's defence of outcome pricing is genuinely valid for enterprise customers who can attribute revenue per resolution (Bloomberg, 2026). When a Shopify or a DoorDash can model the dollar value of a resolved ticket, paying $0.99 to capture, say, $14 in retained customer lifetime value is excellent unit economics.
The enterprise attribution loop
At enterprise scale, support teams measure Customer Effort Score, downstream retention, and NPS movement per channel. They can run experiments. They know that a resolved billing dispute is worth more than a resolved password reset. Outcome pricing lets them weight that.
Why SMBs can't run that loop
SMBs don't have attribution infrastructure. They have a six-seat team, a Stripe dashboard, and a monthly P&L. They need predictable cost lines, not variable ones tied to support volume they can't forecast. Gartner expects GenAI cost-per-resolution to exceed $3 by 2030 in some categories (Gartner, Jan 26 2026). For an SMB, that volatility is unmanageable.
Outcome pricing assumes the buyer can do the attribution math. Most can't.
What does the Klarna AI walkback teach us about outcome pricing?
In February 2024, Klarna announced its OpenAI-powered support agent had absorbed work equivalent to 700 full-time human agents, with a projected $40 million profit improvement that year (OpenAI/Klarna case study, 2024). Fourteen months later, the company quietly reversed course and started rehiring humans. The cost-per-resolution metric was excellent. The customer experience metric was not.
The 14-month reversal in CEO Sebastian Siemiatkowski's own words
In May 2025, Klarna CEO Sebastian Siemiatkowski told Bloomberg the company was rebuilding its human support function: "Cost, unfortunately, seems to have been a too predominant evaluation factor when organising this. What you end up having is lower quality" (Bloomberg, May 8 2025). The $40M profit improvement that headlined the original case study turned into a measurable CSAT decline that no resolution-rate dashboard caught in time.
The OpenAI announcement claimed the AI handled two-thirds of customer service chats in its first month. The 2025 reversal didn't dispute that volume number. It disputed what the volume meant.
Why the metric chose against the customer
Outcome pricing optimises for resolution count. It doesn't optimise for resolution quality. A Fin agent that resolves a ticket by closing it with a generic answer is billed identically to one that actually solved the problem. The pricing model can't see the difference. Klarna's deployment showed the same dynamic at the operational layer: tickets were resolved on paper but the customers weren't satisfied.
The cost-only optimisation trap
[UNIQUE INSIGHT] When the price tag is per-outcome, every product decision tilts toward maximising outcome count. That's the opposite of what good support requires, which is fewer escalations through better knowledge. The lesson from Klarna is that optimising only on cost-per-resolution misses the long-tail quality cost: the churn, the brand damage, the eventual rehiring bill. None of those show up on the AI invoice.
It's worth asking whether outcome pricing structurally rewards the wrong behaviour, even when the headline ROI looks great. Klarna learned this in public so the rest of us don't have to.
Best Intercom alternatives in 2026
When should you pick Corebee over Fin?
If your monthly resolution volume sits between 500 and 5,000 and you're running a small support team, Corebee's flat-fee model produces a more predictable bill than Intercom Fin (Corebee pricing, 2026). The break in the curve happens early and stays in your favour as volume scales.
We're not telling you Fin is a bad product. The 73% resolution rate in head-to-head benchmarking is genuine (MyAskAI, 2026). It's a pricing-fit problem, not a product-fit problem.
Where Fin still makes sense
Fin makes sense if you're enterprise, you have attribution infrastructure, and you want vendor incentives aligned with measurable outcomes. It also makes sense if your volume is genuinely below 100 resolutions a month and likely to stay there.
Where Corebee fits better
Corebee fits the 500-5,000 monthly resolution band where Fin's slope punishes growth. The bill is the bill. You can plan around it.
Frequently asked questions
(See FAQ block below.)
Conclusion
Intercom Fin's $0.99 per outcome model is well-engineered for Intercom. The pricing aligns with the company's enterprise positioning and produces predictable revenue growth tied to customer success metrics. Eoghan McCabe's Bloomberg defence (Bloomberg, 2026) is internally consistent.
But the math at 500, 2,000, and 5,000 resolutions tells a different story for SMBs. $1,005, $2,490, and $5,460 monthly bills aren't catastrophic -- they're just steeper than they need to be. Flat-fee alternatives cap the slope. The Klarna walkback reminds us that optimising on cost-per-resolution alone is a trap.
The definition risk is real too. SaaStr made the same point about HubSpot's resolution-based model: "the vendor decides what resolved means, and that definition will get more generous to the vendor over time" (SaaStr, 2026). The same dynamic applies to any per-outcome model, including Fin's.
"The most predictable support budget is one where the line item doesn't move when your customers start asking more questions. Per-resolution pricing is a success tax — you pay more precisely when your AI is working best." — Jonathan Bar, Founder, Corebee
If you're sizing AI support tooling for a six-seat team this quarter, run the math at three volume bands before committing. The vendor that wins at 500 may not be the one that wins at 5,000.