How we review: this assessment is based on the product’s official documentation, current pricing and published independent evaluations; it is not a hands-on test. Read our review methodology.
Intercom bet early that AI agents would handle a large share of customer-service conversations, and Fin is the result. Its most distinctive feature is commercial rather than technical: instead of charging per seat, Intercom charges per successful outcome, so a business pays when Fin actually resolves a conversation. In 2026, Fin runs on Intercom’s own Apex models, works across chat, email, voice and messaging channels, and can be deployed on top of rival helpdesks. This review assesses Fin as documented on Intercom’s official pages.
What it is
Fin is an AI customer-service agent. It answers customer questions using a company’s help content and data, carries out multi-step tasks through workflows Intercom calls Procedures, and hands conversations to human agents when needed. It is natively integrated with Intercom’s helpdesk but can also be used with other helpdesks, which Intercom names as including Salesforce, Zendesk, HubSpot and Freshdesk. Intercom says Fin is used by more than 12,000 customers. See our Intercom Fin directory entry for a short summary.
Key features
- Multichannel support. Intercom says Fin works across voice, chat, email, Slack and social channels, in dozens of languages, around the clock.
- Procedures. Multi-step workflows in which Fin reads from and writes to other systems — updating accounts, processing refunds or troubleshooting technical issues.
- Apex models. Intercom describes proprietary models trained on customer-experience interactions, including a faster Apex Flash variant for voice.
- Testing suite. Simulations, regression testing and manual review before changes go live.
- Observability. AI-powered insights and QA, metric alerts, trend detection and improvement recommendations.
- Cost controls. Configurable alerts, limits and real-time reporting on outcomes.
- Helpdesk flexibility. Deploy with Intercom or on top of an existing helpdesk; Intercom says setup can take under an hour.
Pricing
At the time of writing (October 2026), Intercom’s official Fin pricing page lists:
| Option | Price |
|---|---|
| Fin with your current helpdesk | $0.99 per outcome, 50-outcome monthly minimum |
| Fin with Intercom | From $0.99 per outcome plus Intercom helpdesk seats |
| Fin Voice | Custom pricing, select customers via sales |
Billable outcomes are resolutions, Procedure handoffs, disqualifications and self-serve routings at $0.99 each, and sales qualifications at $9.99. Intercom charges at most one outcome per conversation and does not charge for escalations to a human triggered by Fin’s default rules, failed Procedures, spam or abandoned conversations. A resolution is counted when the customer confirms the answer helped or leaves without asking for more help, and is deducted if they return to the same conversation for help. There are no setup, integration or platform fees, and a 14-day free trial needs no credit card. Intercom’s helpdesk seats are listed from $29 per seat per month on monthly billing for the Essential plan, rising to $85 for Advanced and $132 for Expert.
Strengths
Outcome-based pricing aligns cost with value. Paying only when Fin resolves or routes a conversation, rather than for every seat or message, makes the business case easy to model: if a resolved conversation costs less than $0.99 of human time, Fin pays for itself. The detailed rules on what counts — and what does not — are clearer than many usage-based schemes.
Helpdesk flexibility lowers the switching barrier. Being able to add Fin on top of Zendesk, Salesforce or HubSpot means companies do not have to migrate their entire support operation to try it.
The tooling around the agent is mature. Procedures let Fin complete real tasks rather than just answer questions, while simulations, regression tests and QA tooling address the main risk of AI agents: unexpected behaviour after a content or configuration change. Our guide on how to launch an AI customer support chatbot explains why testing before launch matters.
Intercom reports strong results. It cites an average resolution rate of 76% across customers, with many above 85%. These are vendor figures, but they come from a large installed base.
Limitations
Costs scale with success. Every resolved conversation costs money, so a busy support operation resolving tens of thousands of conversations a month faces a large, volume-driven bill. Intercom provides alerts and limits, and buyers should set them.
The resolution metric needs scrutiny. Counting a resolution when a customer leaves without asking for more help can include customers who simply gave up. Intercom deducts resolutions when customers return, but teams should track satisfaction and repeat contacts alongside Fin’s own numbers.
Some options are not self-serve. Voice pricing is custom and offered only to select customers, and standalone use requires a 50-outcome monthly minimum.
Performance depends on your content. Like any AI agent, Fin is only as good as the help articles, policies and data it can access; gaps or outdated content will surface as wrong or incomplete answers. Our overview of how AI is transforming customer service covers the organizational work this requires.
Who it’s for
- Support teams with high volumes of repetitive questions.
- Companies on Zendesk, Salesforce or HubSpot wanting an AI agent without migrating.
- Existing Intercom customers looking to automate first-line support.
- SaaS and e-commerce businesses that need multi-step actions such as refunds and account changes.
Alternatives
- Zendesk AI — the AI features built into Zendesk’s own customer-service platform.
- Salesforce Agentforce — AI agents for organizations standardized on Salesforce.
- Building a custom agent on a model API — maximum control, but you own the engineering.
Verdict
Intercom Fin earns 4 out of 5. Its documented strengths — outcome-based pricing with clear billing rules, support for rival helpdesks, Procedures for real tasks, and testing and observability tools that take agent risk seriously — make it one of the most complete AI support agents available. We deduct a point because costs rise directly with volume, voice and standalone use come with conditions, and its headline resolution rates are vendor-reported and depend on a definition of “resolved” that buyers should verify against their own customer-satisfaction data.
