Intercom

Where Intercom sits

Intercom sells a customer service platform — helpdesk, messenger, ticketing, help center — with an AI agent named Fin as its headline. The reason Fin gets top billing is its pricing: a flat, published $0.99 per outcome, billed only when the AI actually resolves the conversation, with no Intercom seats required if you run it on top of Salesforce, Zendesk or another helpdesk. Intercom essentially turned competitor helpdesks into distribution surfaces for its AI, and in doing so set the per-resolution price anchor that the rest of the customer-service AI market is now measured against. HubSpot's Customer Agent charges $0.50 per resolution, Salesforce's Agentforce Help Agent $2, Zendesk layers ~$1.50 per verified resolution on top of suite seats.

The company's own numbers are eye-catching: Fin resolves an average 76% of conversations across 12,000+ customers (vendor figure, improving roughly 1% per month according to Intercom), powered by purpose-built models — a custom retrieval and reranking stack the company calls fin-cx-retrieval and fin-cx-reranker, and a newer "Fin Apex 1.0" engine. Treat those as vendor claims; the independent evidence is mixed and I come back to it below.

What it actually does

On the platform side, Intercom is a modern support workspace: a shared inbox across chat, email, SMS and social channels, ticketing, a public or private help center, workflow automation, round-robin routing, and reporting. Seat tiers at time of writing: Essential at $29 per seat per month billed annually ($39 monthly), Advanced at $85 ($99 monthly), Expert at $132 ($139 monthly). All three include Fin and bill outcomes at the same $0.99. Add-ons stack on any tier: Copilot, the agent-side AI assistant in the inbox, at $29 per agent per month; Pro Analytics (CX Score, Topics, Recommendations) at $99 per month including 1,000 analyzed conversations; Proactive Support Plus at $99 per month including 500 outbound messages; and pay-as-you-go channel costs for SMS, WhatsApp and phone.

On the AI side, Fin is the autonomous frontline agent. It answers across chat, email, SMS and social, executes multi-step Procedures (processing refunds, checking order status, troubleshooting technical issues), and hands off cleanly when it should. An "outcome" is billed when Fin resolves the issue end to end, or successfully completes a Procedure you configured to end in a handoff to a human or workflow; conversations Fin cannot resolve are not charged. There is also a standalone mode: Fin on your existing helpdesk with no Intercom seats at $0.99 per outcome with a 50-outcome monthly minimum. Spend caps let you set hard monthly limits, and Intercom headlines a million-dollar guarantee around Fin's performance. A 14-day free trial is available without a credit card.

My testing notes

The first thing I scrutinised was the outcome meter, because "resolution" is defined by the vendor's own system and every budgeting number flows from it. Intercom's definition has three triggers: Fin completes its workflow, the customer confirms resolution, or the customer does not return to the same topic within a defined window. The third trigger is the soft one — a customer who gives up can count as a resolution. My practical takeaway from reviewing billing behaviour: the $0.99 price is honest and flat, but your effective cost per "real" resolution depends on your traffic mix. A third-party 60-day test on 500 tickets reported a 38% real resolution rate against a marketed 50%, which at $0.99 puts the effective cost closer to $1.40 per genuinely solved ticket. Your mileage will vary with knowledge base quality, but do the pilot math before committing volume.

Second, the standalone economics are the strongest thing in the category. I modelled a 20-agent team deflecting 3,000 of 5,000 monthly conversations: Fin standalone costs $2,970 with zero platform fees, versus $4,670 for Fin plus 20 Advanced seats. The ability to add an AI resolution layer without migrating your helpdesk is genuinely unusual, and the flat public rate is the rare case where a vendor publishes the number enterprise CX buyers usually have to earn through a sales call. Spend caps close the loop — you can hard-limit the monthly bill.

Third, watch the double meter. If your humans live in Intercom, you pay per seat ($29–$132) and per Fin outcome ($0.99) and, if you want the agent-side assistant, again for Copilot ($29/agent/month). A 10-seat Advanced team with 500 Fin outcomes and Copilot lands around $1,635 per month by third-party reckoning before channel costs. Each meter is defensible; stacked, they compound faster than team heads expect.

Limits

Fin's quality ceiling is your knowledge base. It is trained and grounded on the content you give it plus its own retrieval stack — it does not know your product until your help center does, and reviewers report resolution rates climbing only after sustained content work. Historical conversation data helps the assistive features (suggested replies, summarisation) but the autonomous agent lives or dies on documentation quality.

On the platform side, the constraints are cost and fit. Pricing scales with both seats and volume: a support org with 25 Expert seats and 1,500 monthly outcomes runs roughly $5,600 per month before add-ons by published rates. Setup and integration with existing CRM systems takes technical effort despite the polished surface, monthly billing carries a large premium over annual at the entry tier ($39 vs $29 per seat), and there is no free plan at all — only the 14-day trial. Advanced features like SSO, HIPAA support and SLAs sit on the $132 Expert tier.

Finally, a definitional risk specific to outcome pricing: because billing fires on Intercom's classification of "resolved", you inherit its edge cases. If your reporting disagrees with the billing meter, you have two numbers and one invoice. Intercom's spend caps and guarantee soften this, but they do not replace your own deflection measurement.

What deployment actually involves, for anyone budgeting time as well as money: Fin's knowledge setup is the whole job. You connect your help center, docs and past content, configure Procedures for the multi-step actions you want automated (refunds, order lookups), define the handoff rules and escalation paths, and then run topic-by-topic quality checks before letting it loose on live traffic. In configurations I reviewed, teams that did the topic-level testing phase — a week or two of shadow mode comparing Fin's answers against what an agent would have sent — recovered the setup cost quickly, because every knowledge gap found in shadow mode is a resolution you do not bill at $0.99 and a customer you do not hand to a human. The 14-day trial is enough to stand up one channel and test it honestly. One operational detail worth knowing before you commit: because unresolved conversations are free, your early months include a natural testing period at low risk — you pay only for what already works, which is the rare pricing model that makes cautious pilots cheap.

Verdict

For support teams with real ticket volume and a decent knowledge base, Intercom with Fin is the most mature outcome-priced AI support stack on the market, and I rated it 4.2 — roughly 40% for core functionality depth (Procedures, multi-channel autonomous resolution, standalone deployment, helpdesk in one workspace), 20% for value (the $0.99 flat outcome rate is transparent and success-aligned; seat-plus-outcome stacking gets expensive), 20% for usability (polished inbox, low-friction Fin setup, spend caps), 10% for support and ecosystem (works on competitor helpdesks, huge integration surface, published rates), and 10% for reliability (outcome pricing held at $0.99 since 2023 launch, spend caps, million-dollar guarantee; but independent resolution-rate tests land well below vendor averages). Buy it when AI deflection replaces measurable human hours and you can audit what "resolved" means. Skip it if your volume is small or your documentation is thin — the meter will still run, just against weaker numbers.

Sources

• https://intercom.com/learning-center/intercom-vs-front
• https://usagepricing.com/blueprint/intercom-fin
• https://intercom.com/learning-center/ai-customer-service-agent-pricing-comparison
• https://selvo.co/blog/intercom-fin-pricing
• https://selvo.co/blog/intercom-pricing
• https://weekcrm.com/pricing/intercom
• https://gravity.fast/data/ai-agent-pricing-changes/
• https://fin.ai/pricing
• https://intercom.com/pricing