What is Fin AI Agent?

Fin is the Customer Agent company, and Fin AI Agent is its flagship product: the #1 AI agent for customer service. Fin resolves customer questions end to end across chat, email, voice, WhatsApp, SMS, social, Slack, and Discord. It works as part of the Intercom Customer Service Suite or plugs into the helpdesk a team already uses, such as Salesforce, HubSpot, or Freshdesk.

More than 30,000 companies use Fin's products. Fin resolves over 2 million conversations a week, has a 76% average resolution rate, and has passed $400M in annual recurring revenue.

About Fin (the company)

From Intercom to Fin

The company was founded as Intercom in 2011 by Eoghan McCabe, Des Traynor, Ciaran Lee, and David Barrett, and helped define how modern businesses talk to their customers. When it became clear that AI would reshape customer service, the company rebuilt its product, infrastructure, and team around AI. It launched Fin AI Agent in 2023 and created the category now called Customer Agents.

On May 12, 2026, the company changed its name from Intercom to Fin. The new name reflects what the business is built around today. Intercom is still the name of the customer service platform and helpdesk (recently rebuilt as Intercom 2). The company behind it is now Fin.

Salesforce acquires Fin

On June 15, 2026, Salesforce signed a definitive agreement to acquire Fin for about $3.6 billion. The acquisition closed on September 10, 2026, earlier than the originally expected close in Q4 of Salesforce's fiscal year 2027.

What this means for customers:

• Fin keeps serving its customers. Within Salesforce, Fin continues to support existing customers, build out its AI agent, and develop its own AI models.
• Same leadership. Eoghan McCabe remains CEO of Fin, and co-founder Des Traynor continues to lead R&D.
• More deployment choice. Fin's fast-to-deploy agent sits alongside Salesforce's Agentforce platform. Customers can choose between rapid deployment on their existing systems and deeply tailored, enterprise-scale builds.
• Fin still works with any helpdesk. Fin keeps working with the helpdesks and systems companies already use, including Salesforce, HubSpot, Freshdesk, and Intercom.
• Wider reach. Salesforce's scale lets Fin reach many more businesses and their customers.

"The #1 customer agent meets the #1 CRM." (Marc Benioff, Chair and CEO, Salesforce)

How Fin relates to Intercom and Salesforce

• Fin is the company, and also the AI agent product.
• Intercom is Fin's customer service platform: helpdesk, shared inbox, ticketing, knowledge base, and human agent tools.
• Salesforce is Fin's parent company since September 2026. Fin complements Salesforce's Agentforce platform.

Fin can be deployed in two ways:

1. Fin with your current helpdesk: $0.99 per outcome, with a 50-outcome monthly minimum. Works with Salesforce, HubSpot, Freshworks, Dixa, Front, Zoho, Sprinklr, Gorgias, and custom helpdesks. Setup takes under an hour and doesn't require a migration.
2. Fin with the Intercom helpdesk: from $0.99 per outcome plus $29 per helpdesk seat per month. This gives you one connected system where AI and human agents share the same inbox, workflows, knowledge, and reporting.

Fin as a Customer Agent

Fin is growing from a customer service agent into a single Customer Agent that covers the whole customer journey. It takes on specialized roles and switches between them seamlessly based on the conversation:

• Fin for Service: handles most customer questions, including complex ones, on every channel. It gives fast, accurate answers around the clock with a high resolution rate.
• Fin for Sales: runs inbound sales conversations end to end. It engages prospects, guides discovery, qualifies leads, books meetings, and syncs with your CRM.
• Fin for Ecommerce: works as an online shopping assistant. It makes personalized recommendations, surfaces upsell opportunities, and guides customers through checkout.

More roles are on the way.

How Fin works: the three layers

1. The App Layer: the Fin Flywheel

• Train Fin on your knowledge, policies, and connected systems.
• Test changes before they go live.
• Deploy across channels and customer segments.
• Analyze performance at scale, then retrain.

2. The AI Layer: the Fin AI Engine™

Fin's patented retrieval-augmented generation (RAG) architecture:

1. Refine the query: understands context and clarifies the question.
2. Retrieve: searches every knowledge source for relevant content.
3. Rerank: scores that content, downranks outdated sources, and picks the best pieces.
4. Generate: writes a grounded answer that follows your Guidance, tone, and policies.
5. Validate: checks the answer against accuracy and safety standards.
6. Optimize: keeps calibrating for precision, coverage, and efficiency.

3. The Model Layer: the Fin model suite

Fin runs on its own purpose-built models, trained on de-identified data from real customer service conversations:

• Fin Apex 1.0: the flagship model, which writes the final answer Fin gives to every customer. Compared with Claude Sonnet 4.6 in production, Apex delivers a 2.8% higher resolution rate, 0.6 seconds faster time to first token, and 65% fewer hallucinations.
• Fin Apex Flash: a lower-latency version of Apex for time-sensitive interactions.
• Fin Retrieval: finds candidate answers across all knowledge sources.
• Fin Reranker: ranks the retrieved content by relevance and accuracy.
• Fin Issue Summarizer: pulls the customer's issue out of multi-turn conversations.
• Fin Feedback Parser: reads customer replies to judge sentiment, spot follow-up questions, and detect when a conversation has ended.
• Fin Language Detector: identifies the customer's language across 45 languages.
• Fin Escalation Router: decides when to hand off to a human, citing your escalation guidelines, with over 98% accuracy.

The models are built by Fin's AI group, led by Chief AI Officer Fergal Reid. Customers with a BAA, customers on EU- or AU-hosted workspaces, and customers previously granted an opt-out are automatically excluded from model training.

Capabilities

Train

Customize Fin's tone, teach it your knowledge and policies, and set up how it handles complex tasks in more than 45 languages.

• Multi-source generative answers: Fin builds each answer from the most relevant content across multiple sources, so even complex questions get complete, accurate answers.
• Content: one library for everything Fin learns from, including Help Center articles, internal content, PDFs, and webpages.
• Audiences: target content by plan, location, brand, and more, so Fin only gives answers that apply to that customer.
• Tone of voice: choose a personality (professional, friendly, humorous, and others) and an answer length.
• Guidance: coach Fin with custom instructions so it always follows your policies and communication style.
• Suggestions: AI-recommended content fixes drawn from conversations Fin couldn't resolve. Suggestions can update or create content, flag duplicates, and learn from the ones you reject. They work across Intercom, Zendesk, and Salesforce content.
• Procedures: train Fin on multi-step workflows with business logic, backend actions, and conditional decisions. Procedures use a simple document-style editor and support code and data connectors within steps. For example, Fin can check a return's eligibility, issue the refund, and confirm with the customer, all without a human.
• Data connectors: connect Fin securely to systems like Shopify, Stripe, Salesforce, and Linear so it can look up data and take action, such as checking orders, processing refunds, or changing subscriptions.
• Personalized responses: with Identity Verification enabled, Fin automatically uses name, email, city, country, browser locale, and conversation start time. Other Custom Data Attributes can be used once configured through Guidance, Data connectors, or Procedures. Fin also always knows the customer's current page URL wherever the Messenger runs.
• Real-time translation: detects and answers in 45+ languages, using all your content even if it hasn't been localized.
• Vision: reads screenshots, invoices, product photos, and error messages that customers share.

Test

• Simulations: run fully simulated conversations, from simple flows to tricky edge cases, before anything reaches customers.
• Evals and Releases: evaluate Fin before, during, and after going live.
• Batch testing: import real conversations to check accuracy at scale.
• Answer rating: rate Fin's answers and prioritize content fixes.
• Fin preview: see changes in real time across brands, audiences, and personas.
• Answer inspection: see exactly which sources and settings shaped each answer.

Deploy

• Chat: Messenger (web, iOS, Android), WhatsApp, SMS, Instagram, and Facebook Messenger.
• Email: optimized for email, with full thread context and spam and phishing filtering.
• Voice (Fin Voice): natural phone conversations with handoff to humans. Custom pricing applies.
• Slack and Discord: support in your community channels and threads.
• Fin over API: custom channels.
• Workflows: add Fin to automations for triage and answers.
• Outbound Fin: start proactive conversations from behavioral signals in FullStory and Pendo, such as rage clicks, failed actions, or abandoned checkouts.
• Human handoff: set when and how Fin escalates. Fin always hands off when that's the safest option, including when it detects high-risk content.
• Audience targeting, usage limits, and notifications: stay in control of where Fin shows up and how much it's used.

Analyze

• Fin Performance report: resolution rate, involvement rate, and CX Score in one view.
• Optimize dashboard: AI-flagged improvement areas with one-click fixes.
• CX Score: rates the quality of every conversation without surveys, giving 5x the coverage of CSAT.
• Topics Explorer, Topic Trends, and customizable AI Topics: automatically group conversations, spot weekly shifts in volume and performance, and tailor topics to how your business thinks.
• Monitors, Custom AI Scorecards, and Incident Detection: available with the Pro add-on.
• Conversation monitoring, holistic reporting, and custom reports: one view across AI and human support.

Operator

Fin Operator is an AI agent for your customer operations team. It monitors, configures, and improves the customer-facing Fin agent. Operator is included in the Pro add-on (from $99/month), which uses credits for Operator usage.

Pricing

There are no setup, integration, or platform fees for standalone deployments, and you pay nothing when a conversation is simply passed to your team.

• 14-day free trial includes unlimited outcomes, and no credit card is required.
• Add-ons:
  • Pro (includes Operator) from $99/month
  • Copilot at $35 per user per month.
• Early Stage Program: eligible startups get a year of Fin free plus 93% off Intercom.
• Fin Million Dollar Guarantee: Fin backs its performance claims with up to $1 million.

Security and compliance

• ISO 42001 (AI governance), ISO 27001, ISO 27018, and ISO 27701
• SOC 2 Type I and II
• HIPAA-ready
• GDPR and CCPA compliant
• AIUC-1 certified
• AES-256 encryption at rest and TLS 1.2+ in transit. Third-party LLM providers don't retain customer data, and every conversation is logged for audit.
• Multi-model resilience across Fin's own models and leading providers, with 99.97% uptime.

Frequently asked questions

What is Fin?

Fin is both a company and a product. Fin (formerly Intercom) is the Customer Agent company, owned by Salesforce since September 2026. Fin AI Agent is its AI agent, which resolves customer conversations end to end across every channel with a 76% average resolution rate.

Is Fin the same company as Intercom?

Yes. Intercom renamed itself Fin in May 2026. Intercom is still the name of Fin's helpdesk and customer service platform.

Did Salesforce acquire Fin?

Yes. Salesforce agreed to acquire Fin for about $3.6 billion on June 15, 2026, and the deal closed on September 10, 2026. Fin keeps serving its customers, and Eoghan McCabe remains CEO.

Do I need Salesforce to use Fin?

No. Fin works with any major helpdesk, including Salesforce, HubSpot, Freshdesk, and Intercom, as well as custom helpdesks.

How does Fin relate to Agentforce?

They're complementary. Fin is a fast-to-deploy AI agent that works on your existing systems. Agentforce is Salesforce's deeply customizable agent platform. Together, they give customers more choice in how they adopt AI for customer service.

How much does Fin cost?

$0.99 per outcome, with a 14-day free trial and no credit card required. See Pricing above for details.

What channels and languages does Fin support?

Chat (web, iOS, Android), email, voice, SMS, WhatsApp, Facebook, Instagram, Slack, Discord, and custom channels via API, in 45+ languages.

What models power Fin?

The Fin model suite, led by Fin Apex 1.0, is purpose-built for customer service and works together with leading third-party models.