What Is Intercom?
Intercom is a customer messaging and support platform built around Fin, its AI customer-service agent. Where older tools suggested replies for human reps, Fin is the customer-facing layer: it reads a question, pulls from your knowledge, and either answers, takes an action, or hands off to a human.
The company committed to this so fully that in mid-2026 it rebranded itself to Fin. That’s the headline shift: Intercom is no longer a messaging tool with AI bolted on; it’s an AI agent company with a helpdesk attached.
Fin reports resolving 50–70% of conversations autonomously, and runs at major companies including Anthropic (96% of conversations), Lightspeed (99%), and Clay (90%). It’s aimed at mid-market and enterprise support teams ready to invest in real AI deflection.
Two things that reframe a 2026 decision. First, Intercom rebranded the company to Fin in mid-2026. Second, Salesforce is acquiring Fin. Both can affect the roadmap, pricing, and integrations, so check the latest on the acquisition before committing to a contract.
How Does Intercom Fin Work?
Fin isn’t a thin wrapper around a chatbot. It runs on the Fin AI Engine, a retrieval pipeline of custom models handling query refinement, retrieval, reranking, answer generation, and a validation step designed to curb hallucinations, all layered on frontier LLMs. Intercom brands its latest model Fin Apex 1.0.
The core loop: ingest your help content, retrieve the right passage for each question, generate a grounded answer, and escalate when confidence is low. Because it answers from your knowledge rather than the open internet, it stays on-brand and on-fact more reliably than a generic bot.
Crucially, Fin runs across chat, email, WhatsApp, SMS, phone, and Slack, and it doesn’t only run on Intercom’s own helpdesk. It can sit on top of other support platforms too.
Key Intercom Features
- Intercom Fin AI Agent
- Fin is built to resolve complex customer queries across every channel, operating as part of Intercom’s Customer Agent architecture. It reads a customer’s question, pulls from your knowledge base, and either answers directly, takes an action, or hands off to a human — running across chat, email, WhatsApp, SMS, phone, and Slack.
- Intercom Copilot (AI Assist for Support Agents)
- Copilot is Intercom’s separate agent-assist tool; it doesn’t resolve tickets on its own but helps human reps answer faster, distinct from Fin’s autonomous resolution role.
- Multi-Role Customer Agent
- Fin is evolving into specialised roles: service, sales, and more, operating as a single agent that switches between roles based on conversation context.
- Omnichannel Support
- Chat, email, WhatsApp, SMS, phone, and Slack are managed from a single inbox; no separate tools are needed for each channel.
- Intercom API for Custom Workflows
- Developers can extend Intercom’s automation and AI features into their own stack via its public API.
- Outcome-Based AI Pricing
- Fin is priced at $0.99 per outcome across all Intercom plans, covering resolutions, procedure handoffs, and disqualifications, separate from seat pricing.
Who Should Use Intercom?
Best suited for: Mid-market and enterprise CX teams ready to invest in AI deflection, SaaS and e-commerce companies with high support volume, and teams that want autonomous resolution rather than reply suggestions.
Not ideal for: Startups with fewer than 5 seats or cost-sensitive teams, where Crisp, Help Scout, or Zendesk cover the basics for less; and teams needing traditional IT-style ticketing, which Zendesk handles better.
Imperial AI Tools Feedback
Intercom made the AI-first pivot that most support platforms only talk about. Fin’s autonomous resolution rate is the real differentiator: resolving 50–70% of conversations without a human is a measurable result no direct competitor matches, and the retrieval-plus-validation architecture keeps answers grounded in your knowledge rather than hallucinated. For a high-volume support team, that’s transformative.
The honest constraints are about cost and commitment. Seat pricing escalates, per-resolution billing compounds at volume, and the workflow builder needs real CX-ops investment to get right. Migration is genuinely hard once you’re embedded, which is worth weighing given the Salesforce acquisition; the product you buy today may sit inside a different company tomorrow.
Our view: the default choice for AI-native customer service at mid-market and enterprise scale, provided you can absorb the compounding cost and the workflow learning curve. Smaller or budget-tight teams should look at lighter alternatives first.
Suggestions For Improvement
- Offer volume discounts as per-resolution costs compound at scale.
- Make total cost clearer upfront, combining outcomes and seat fees.
- Simplify the workflow builder to lower the CX-ops burden.
- Provide clearer migration paths to reduce lock-in concerns.
- Publish roadmap clarity following the Salesforce acquisition.
