Vendor explainer · AI customer agent · Updated May 12, 2026 (post-rebrand)

What Is Fin? The AI Customer Agent Platform Explained

Fin is the AI customer agent platform built by Fin (the company formerly known as Intercom). It runs autonomous conversations across customer support, inbound sales, ecommerce, and customer success — resolving 40-60% of conversations end-to-end without human handoff at typical deployments. The product launched three years ago; the parent company rebranded to Fin on May 12, 2026. There are now two things called Fin: the product (what you deploy) and the company (who you sign with). This explainer covers what Fin does, how it deploys, what it costs, who it competes with, and how to evaluate against Ada / Decagon / Sierra / Salesforce Agentforce.

See if Fin overlaps my stack →Compare Fin vs Ada / Decagon / SierraThe rebrand context

What Fin actually does

Four workflow roles, one shared agent architecture. The architectural choice that distinguishes Fin from most competitors: knowledge and customer context are shared across all four roles, so a prospect asking a product question during a sales conversation gets handled by the same agent that would answer the same question for a support ticket. Most competitors run separate agents per workflow.

RoleWhat it doesCapability depth
Customer support (Fin Customer Agent)Autonomous resolution of inbound support tickets — answers product questions, troubleshoots issues, processes refunds, updates orders, escalates to humans only when the conversation requires human judgment.Reads from your knowledge base, help center, internal docs, past tickets, and product API. Resolves an estimated 40-60% of inbound conversations end-to-end without human handoff (varies by industry and KB quality).
Inbound sales (Fin for Sales, GA April 2026)Engages inbound prospects via chat, qualifies against playbook rules, answers product / pricing questions, books meetings via Chili Piper or Calendly, routes high-intent buyers to sales with full context.Shares knowledge base with the support role, so a prospect can ask "what does pricing look like" and a product question in the same conversation — Fin handles both without handoff. Replaces Drift, Qualified, Default.com, Chili Piper inbound triage layer.
EcommerceOrder tracking, returns, exchanges, shipping updates, product recommendations, abandoned cart recovery. Handles the volume layer of B2C support where conversation volume swamps human teams.Native integrations with Shopify, BigCommerce, WooCommerce, Magento. Fin was originally built for B2C / ecommerce — this is the strongest-fit deployment shape.
Customer successOnboarding nudges, feature discovery prompts, renewal-window conversations, expansion-opportunity surfacing. Operates in the post-sale conversation layer where CSMs are stretched.Shares customer history across the sales / support / success roles. The single-customer-agent vision is structurally distinct from point-solution AI tools that operate in one workflow.

How Fin deploys — five paths

The strongest fit is Fin inside Intercom 2 (native), but Fin sells standalone for non-Intercom-2 deployments. Your existing helpdesk anchor determines the right path.

DeploymentFitWhy
Inside Intercom 2 (native)Strongest fitFin is baked into Intercom 2 — the rebuilt helpdesk product from the same company (now called Fin). Workforce forecasting accounts for Fin resolution volume, Monitors review Fin conversations against custom scorecards, the workflow is purpose-built for AI + human teams operating together. If you are evaluating Intercom 2, Fin comes with it.
On Zendesk (via Fin standalone)Strong fitFin sells standalone for Zendesk customers who do not want to migrate helpdesks. Integration is mature — Fin Connect to Zendesk Support, Fin reads from Zendesk Guide knowledge base, ticket creation / handoff flows live in Zendesk. Common deployment for Zendesk customers who want AI-native resolution without buying Zendesk's own AI add-ons.
On Salesforce Service CloudGood fitFin integrates with Salesforce Service Cloud as the AI layer. Less mature than the Intercom 2 native deployment, but functional. Competes with Salesforce Agentforce (Salesforce's own agent platform) — buyers choose based on whether they want AI-platform independence or Salesforce-native bundling.
Standalone (no helpdesk)Partial fitFin can run on its own without a helpdesk for chat-only deployments, but most teams want ticket-system integration. Fin without a helpdesk is unusual — typically a transitional deployment before adopting Intercom 2 or integrating with an existing helpdesk.
On Help Scout, Front, Freshdesk, Kustomer (via API)VariableFin's API allows integration with most modern helpdesks. Depth varies — native integrations (Intercom 2, Zendesk, Salesforce) get more polish; API-based deployments require integration work. Evaluate based on your team's integration capacity.

Pricing — per resolution, not per seat

Fin charges per resolution — a conversation that Fin closes end-to-end without human handoff. Roughly $0.99 per resolution at typical volumes (negotiable at scale). The structural advantage of per-resolution pricing: you pay only when Fin actually closes conversations. If Fin escalates to a human, no resolution charge.

The math: if your team handles 10K conversations / month and Fin resolves 50% (5K resolutions), the Fin bill is ~$5K/mo. Compare against the loaded cost of human-only agent capacity for those 5K conversations: at $50/hr fully-loaded agent cost and ~10 minutes per conversation, that's ~833 hours = ~$42K/mo. Even at 30% resolution (3K conversations), the math meaningfully clears the human-cost line. The per-resolution model also caps your downside — Fin resolution quality determines the bill, not seat count.

What is not in the per-resolution price: the Intercom 2 helpdesk seat costs if you deploy Fin native (Intercom 2 is priced separately, per-seat). Fin standalone on Zendesk: you pay Fin per resolution + Zendesk per seat as separate line items.

Customer proof — what we know

Three named references plus the aggregate claim:

CustomerResultContext
Attio1,600+ conversations handled, 50+ qualified leads, one prospect converted at 6x ACVFin for Sales role, early-customer launch cohort (April 2026)
Fellow18 meetings booked in January, ~48% conversion rateFin for Sales role, early-customer launch cohort
Anthropic (customer support)Reported as a Fin customer — handles Claude support volumeFin Customer Agent on AI-native company support, public reference
Aggregate claim (Intercom announcement)"Close/win rates of nearly 50% in the first month" for Fin for Sales early customersLaunch-cohort numbers, almost certainly best-case — verify in pilot

These are largely launch-cohort numbers. Public Fin customer count is reported as growing rapidly — over 5,000 customers on Fin as of early 2026, per Intercom (now Fin) disclosures. Run your own pilot before committing — the 40-60% resolution rate is real but varies meaningfully by industry, knowledge base quality, and ticket complexity.

The competitive set — Fin vs Ada, Decagon, Sierra, Forethought

Five named competitors in the AI customer agent category. The category got crowded fast in 2024-2026 — most CX AI vendors are now positioning around "agent" terminology. Genuine differentiation is narrower than the marketing suggests.

Fin's differentiation: the cross-workflow agent (support + sales + ecom + success) where most competitors focus on one workflow. The Intercom 2 bundle is also a real differentiator for teams willing to adopt the helpdesk along with the agent. Vendor-neutral comparison at /best-ai-customer-agents-2026.

Sources

FAQ

Fin is the AI customer agent platform built by Fin (the company formerly known as Intercom). It runs autonomous conversations across customer support, inbound sales, ecommerce, and customer success workflows. As of May 12, 2026, "Fin" also refers to the company — the corporate brand was renamed to match the flagship product. Fin (the product) was launched three years ago, predates the company rename, and continues to be sold standalone or bundled with the Intercom 2 helpdesk.

No. Fin is a customer-facing AI agent — designed to resolve customer service conversations end-to-end without human intervention. ChatGPT / Claude / Gemini are general-purpose LLMs. Fin is built on top of LLM foundation models (the specific model is proprietary, likely a mix of Anthropic Claude + OpenAI GPT-4-class models routed by task) but the product surface is purpose-built for customer service workflows: ticket resolution, refund processing, order updates, knowledge-base reading, escalation logic. The distinction matters: Fin is an agent (takes actions, reads/writes to systems) where ChatGPT is a chat interface.

Traditional chatbots (HubSpot Chat, ManyChat, Tidio, Intercom's pre-Fin bots) follow scripted decision trees — you write rules, the bot routes based on keyword matches, conversations break when prospects say something the rule library doesn't cover. Fin operates on LLM understanding — reads natural language, accesses your knowledge base, takes actions via API integrations, and resolves conversations end-to-end without escalating until human judgment is genuinely required. The output quality difference is meaningful — chatbots resolve ~5-15% of conversations end-to-end; Fin reports 40-60% resolution rates (varies by industry + KB quality).

Per-resolution pricing model. Fin charges per "resolution" — a conversation that Fin closes without human handoff. Pricing is roughly $0.99 per resolution at typical volumes (negotiable at scale). The math: if your team handles 10K conversations / month and Fin resolves 50%, that's 5K resolutions at ~$0.99 = ~$5K/mo. Compare against the loaded cost of human-only agent capacity for those 5K conversations ($50K-$150K/mo at typical CS team math). The pricing model rewards Fin for resolution quality — Fin gets paid only when it actually closes conversations, not for every message.

Fin the product launched three years ago as the AI customer agent inside Intercom. Fin the company is the renamed corporate entity (May 12, 2026) — the parent company formerly known as Intercom. Same name, different referents. The product is what you buy and deploy. The company is who you sign a contract with. The reason they share the name: the company renamed itself after the product it is betting the future on, matching the corporate brand to the flagship product. The previous setup ("Intercom the company sells Fin the product") created mixed signals about which was the strategic priority.

Yes. Fin sells standalone for deployment on Zendesk, Salesforce Service Cloud, and other helpdesks via API. The most common standalone deployment is Zendesk + Fin — Zendesk customers who want AI-native resolution without buying Zendesk's own AI add-ons (Zendesk AI). Standalone Fin retains most capabilities; the integration depth (workforce forecasting, Monitors, knowledge-base sharing) is strongest in the Intercom 2 native deployment. Evaluate based on whether you want to migrate helpdesks (then Intercom 2 + Fin bundled) or keep your existing helpdesk (then Fin standalone).

Five named competitors in the AI customer agent category. (1) Ada — established CX AI platform, deepest enterprise track record, strong in B2C / financial services / telco. (2) Decagon — hot AI-agent startup, raised significant funding, strong technical pitch on action-taking agents. (3) Sierra — Bret Taylor's company (former Salesforce co-CEO), strong enterprise team, growing fast. (4) Forethought — mature CX AI vendor, strong on ticket triage + agent assist. (5) Salesforce Agentforce — Salesforce-native agent platform, wins on bundle for Salesforce-anchored teams. The category is competitive; Fin's differentiation is the cross-workflow agent (support + sales + ecom + success) where most competitors focus on one workflow.

Genuinely autonomous for the resolution rates Fin reports (40-60% end-to-end resolution at typical deployments). Fin reads from your knowledge base, takes actions via API (issue refunds, update orders, change subscriptions, look up account data), and closes conversations without human handoff. The escalation logic is configurable — you set the boundary where Fin transfers to humans (sentiment threshold, complexity score, customer tier rules, specific intents that always escalate). The "just faster routing" framing fits chatbots, not Fin — the product class is meaningfully different.

Five sources, ranked by Fin's preference order. (1) Your help center / public knowledge base (Intercom Articles, Zendesk Guide, custom KB sites). (2) Internal documentation (Notion, Confluence, Google Docs via integration). (3) Past resolved tickets (Fin learns from how humans previously resolved similar issues). (4) Product API responses (live account data, order status, subscription state). (5) Custom playbook rules (your team writes rules for specific intents — refund eligibility logic, escalation triggers, account-tier behaviors). Fin's resolution quality scales with knowledge-base quality — teams with thin or stale KBs see worse results.

StackSwap models GTM stacks against synthetic stack patterns. For Fin specifically: (1) If your stack includes Intercom + Drift + Ada (or any combination of chat / inbound / AI tools), our overlap engine flags the redundancy — Fin consolidates multiple line items into one platform. (2) If you are evaluating Fin vs Ada vs Decagon vs Sierra, the decision turns on workflow scope (single workflow vs cross-workflow), helpdesk anchor (Intercom 2 vs Zendesk vs SFDC vs standalone), and customer-base proof (B2C / B2B / vertical). (3) For existing Intercom customers, Fin is the default AI agent layer — evaluate against staying on a legacy bot layer or buying a different vendor. Run StackScan to see modeled redundancy + recoverable spend for your specific stack.

Related reading

Canonical URL: https://stackswap.ai/what-is-fin-ai-agent. Disclosure: StackSwap has no commercial relationship with Fin (formerly Intercom). Sourced from publicly available announcements, vendor documentation, and third-party coverage.