Intercom is a B2B customer service platform combining an omnichannel inbox, ticketing and help centre with Fin, an AI agent that answers customers directly and is billed per outcome on top of per-seat fees.
Used this tool? Rate it
Used this tool? Rate it
Intercom is a customer service platform for businesses, built around the premise that an AI agent should answer most incoming questions and human agents should handle what is left. The company describes it as the only helpdesk designed for the AI Agent era. In practice it combines four things: an omnichannel inbox spanning email, chat, phone, WhatsApp and social, a ticketing system with AI-assisted categorisation, a help centre that doubles as the knowledge source for the AI, and Fin — the AI agent that talks to customers directly.
One identity fact needs stating up front because it changes how you should read everything else. In May 2026 the corporate entity renamed itself Fin, after its AI agent, and the Intercom name was retained for the customer service software platform itself. All 1,400 employees now work for Fin, and the platform was relaunched as a complete rebuild. So "Intercom" today names the product, not the company. If you are searching for corporate information, contracts or support channels, expect to encounter both names, and the Fin brand also appears on separate domains.
What makes this product commercially distinctive is not the feature list — competitors have comparable inboxes — but the billing model. Alongside conventional per-seat subscriptions, Fin is charged per outcome. That single design choice reshapes the economics of running support: it means your bill moves with how much work the AI does, not just with how many people you employ. It is the most consequential thing to understand before buying, and it is covered in detail under pricing and limitations.
On scale, the company claims more than 30,000 leading brands as customers, naming Anthropic, Clay, Lightspeed and Rocket Money among them. A company profile hosted on a third-party review platform adds further figures: resolving over two million conversations weekly and passing $400M in annual recurring revenue, with more than 1,400 employees across six offices and origins in 2011. These are self-reported and, being a private company, unaudited.
Pricing is published rather than quote-gated, and it has two components that must be added together. The seat component runs Essential at $29 per seat, Advanced at $85 and Expert at $132, monthly. On top of every one of those tiers sits the Fin component: from $0.99 per outcome. There is also a seat-free option — Fin alone, deployed on your existing helpdesk, priced from $0.99 per outcome with sales contact required.
What counts as a billable outcome is documented precisely, and it is worth knowing before signing. There are four types: a resolution at $0.99, a procedure handoff at $0.99, a disqualification at $0.99, and a lead qualification at $9.99. Only one outcome is charged per conversation, even if Fin takes multiple actions. A number of situations are explicitly not billed, including escalations triggered when frustration is detected, procedure failures caused by technical errors, customers who ask for a human, spam, and conversations abandoned after Fin only asked clarifying questions.
Tier differences beyond seat price matter for feature access. Advanced adds multiple team inboxes, the workflow automation builder, round-robin assignment and a private multilingual help centre, and includes 20 free Lite seats. Expert adds SSO and identity management, HIPAA support and service level agreements, plus multibrand messenger and help centre, and includes 50 free Lite seats. Add-ons are priced separately: Pro from $99 monthly including Operator, Copilot at $29 per agent monthly, and Proactive Support Plus at $99 monthly.
Two further cost realities deserve attention. First, channel usage is not all included — live chat, support email, in-app chats, banners and tooltips are unlimited, but email campaigns, SMS, WhatsApp and phone are pay-as-you-go on every plan. Second, discounting is aggressive: a 35% new customer discount on the entry plan and startups get 93% off. Steep introductory discounts mean the price you are quoted in year one is not the price you should plan around in year three.
Per-outcome billing makes costs move in ways buyers do not expect. This is the defining consideration. Because Fin charges per outcome on top of seats, the bill scales with conversation volume rather than with headcount, and it can rise as the AI becomes more effective — better documentation produces more AI-resolved conversations, and more resolved conversations produce more billable outcomes. The vendor's own provision of a cost estimator and an ROI calculator implicitly acknowledges that this model is hard to predict. Anyone who requires a fixed annual figure should treat this as a structural mismatch rather than a negotiable detail.
"Assumed" resolutions are billable, and they are not proof of a happy customer. The official documentation distinguishes a confirmed resolution, where the customer explicitly says the answer helped, from an assumed resolution, where the customer merely exits the conversation without requesting further assistance. Both bill at $0.99. A customer who gave up, got distracted, or went to a competitor's product instead can register the same way as one who was genuinely helped. The vendor does exclude several categories from billing — frustration-triggered escalations, requests for a human, technical procedure failures, spam — which limits the problem, but it does not remove the core ambiguity. Read transcripts, not just counts.
User reviews are polarised and support quality is the recurring complaint. On the main third-party review profile the product scores 2.9 out of 5 across 524 reviews, with 44% five-star against 24% one-star — a genuinely split population rather than a mediocre average. The platform's own summary notes users find it intuitive and feature-rich while also reporting excessively long response times from the company's own support. There is an irony worth naming plainly: the most consistent criticism of a customer service platform concerns its customer service.
Implementation and migration support draw specific criticism. Reviewers describe being passed round from agent to agent on billing, onboarding and technical questions, receiving no onboarding support beyond outdated articles that did not match the current product, and finding that the terms of free seats changed repeatedly. For a platform whose value depends on getting the knowledge base and workflows right, weak implementation support is a material risk rather than a cosmetic one.
Price increases at renewal are a documented grievance. One long-term customer of over five years reports being forced off legacy pricing plans onto new ones at a 25% increase, with discounts simultaneously reduced. Combined with the published discount structure — startups get 93% off — this points to a pattern worth planning for: the entry price and the steady-state price can differ substantially.
Return on investment is contested by some users. At least one reviewer argues that users end up involving human support anyway and that a well-implemented site search achieves comparable results far more cheaply, questioning whether the investment holds up as costs rise. This is a single account rather than a measured finding, but it names the right test: compare Fin's cost per outcome against your actual cost of handling the same ticket, including the cases where it escalates anyway.
Independent verification is incomplete. G2's review pages returned HTTP 403 on both the direct and headless-browser channels during this research, so no rating from that source is reproduced here. A second review profile for the same brand exists with only ten reviews and a 2.3 score; that sample is far too small to support any conclusion and is not used as one. Context also matters for the 2.9 figure: comparable platforms on the same review site sit lower still, which suggests category-wide dissatisfaction with enterprise support software rather than a defect unique to this product.
Privacy documentation is strong on rights but silent on AI training. The policy sets out GDPR rights for EEA, UK and Swiss residents and CCPA rights for Californians in itemised detail, which is more than many competitors provide. However, retention is described mostly by purpose limitation, with a specific period given only for prefilled form data kept for no more than 14 days. And while the policy confirms that AI and machine learning including generative AI may be used, it does not state whether customer data is used to train models — a question worth asking sales directly, particularly given that HIPAA support is sold as a tier feature while no certification is named in the policy text itself.
The rebrand introduces practical ambiguity. With the company now named Fin and the platform still named Intercom, buyers encounter both names across contracts, domains, support channels and review sites. Independent coverage of the change was measured rather than celebratory — one bylined report observed that whether that reads as confidence or catch-up depends on where you sit in the market. Expect some friction in procurement paperwork and in searching for information during the transition period.
Yes, and the distinction is worth getting right. The corporate entity renamed itself Fin on 12 May 2026, taking the name of its AI agent, while the Intercom name was kept for the customer service software platform. All 1,400 employees now work for Fin. So Intercom is the product and Fin is both the company and the AI agent inside the product. You will encounter both names in contracts, on support channels and on review sites.
Two components added together. Seats cost $29 monthly on Essential, $85 on Advanced and $132 on Expert. On top of any of those, Fin bills from $0.99 per outcome. Add-ons are separate again — Copilot at $29 per agent monthly, Pro from $99 monthly, Proactive Support Plus at $99 monthly — and email campaigns, SMS, WhatsApp and phone are pay-as-you-go. There is also a seat-free option to run Fin on your existing helpdesk.
Four types: a resolution, a procedure handoff and a disqualification each at $0.99, and a lead qualification at $9.99. Only one outcome is charged per conversation regardless of how many actions Fin takes. Explicitly not billed are escalations triggered by detected frustration, technical procedure failures, customers asking for a human, spam, and conversations abandoned after Fin only asked clarifying questions.
Possibly, and this is the sharpest edge of the model. Documentation distinguishes confirmed resolutions, where the customer says the answer helped, from assumed resolutions, where the customer exits the conversation without requesting further assistance. Both are billable. Several dissatisfaction signals are excluded from billing, but silence is treated as success. Review transcripts rather than relying on the resolution count as a satisfaction proxy.
Yes. Fin is sold as a standalone deployment on your current helpdesk, including Salesforce, with no seats required and setup described as taking under an hour. It answers email, live chat and phone, allows customisation of tone and answer length, can act on external systems and hands off to agents in your preferred inbox. This is the lowest-commitment way to evaluate it.
It can, and that is the counter-intuitive part of the model. Improving your help centre increases the share of conversations Fin resolves, and each resolution is a billable outcome. Better automation therefore produces a larger AI bill even while human workload falls. Whether that is a good trade depends entirely on your cost per human-handled ticket, which is why modelling it before committing matters.
Advanced adds multiple team inboxes, workflow automation, round-robin assignment and a private multilingual help centre, with 20 free Lite seats. Expert adds SSO and identity management, HIPAA support and service level agreements, plus multibrand messenger and help centre, with 50 free Lite seats. If you need SSO, SLAs or HIPAA support, your effective entry point is the top tier at $132 per seat, not the headline $29.
The most substantial dataset is 2.9 out of 5 across 524 reviews, sharply split at 44% five-star and 24% one-star — informative, but read the distribution rather than the average. A second profile for the same brand has only ten reviews and is too small to draw conclusions from. G2 could not be verified firsthand during this research. Note too that comparable helpdesk platforms score similarly or lower on the same site, so the number reflects category-wide sentiment as much as this product.
Three recur. First, cost: unpredictable per-outcome billing and price increases at renewal, including one long-term customer reporting a forced move off legacy pricing at a 25% increase. Second, the company's own support responsiveness, with reviewers citing excessively long response times. Third, implementation, with reports of being passed round from agent to agent and receiving outdated onboarding material.
The privacy policy does not say. It confirms that AI and machine learning technologies including generative AI may be used by the company, its service providers and third parties, but it does not state whether customer data feeds model training, and it names no security certification in its own text, directing readers to a separate security page instead. If this matters to your procurement or compliance review, get the answer in writing from sales rather than inferring it.
