Salesforce Buys Fin for $3.6B to Boost Agentforce
Salesforce agrees to pay $3.6 billion for Fin, the AI customer service agent formerly known as Intercom, adding a proprietary model and 30,000 customers to Agentforce.
6 min read
Salesforce agreed Monday to pay $3.6 billion for Fin, the AI customer service platform formerly known as Intercom, in what both companies are calling the largest agentic customer experience deal on record. The transaction brings Salesforce a proprietary language model, 30,000 existing customers, and a team that shipped something Agentforce still can't deliver out of the box: a working AI agent that SMBs can deploy in hours.
TL;DR
- $3.6B deal for Fin (formerly Intercom), expected to close in Q4 of Salesforce's fiscal year 2027
- Fin's proprietary Apex model resolves 76% of customer support queries autonomously - independent estimates from late 2025 put the real number closer to 67%
- Agentforce reached $1.2B ARR in Q1 FY27, up 205% year-over-year - yet Salesforce still needed to buy this
- Fin CEO Eoghan McCabe and R&D co-founder Des Traynor both stay on post-close
- Zendesk, ServiceNow, and Freshworks now compete against a Salesforce that owns both the dominant CRM and a leading AI service agent
What Salesforce Is Paying For
The Apex Model
Fin's core product runs on Apex, a proprietary language model built specifically for customer support. Unlike general-purpose models from OpenAI or Anthropic, Apex was trained on support conversations at scale and designed to resolve queries without escalating to a human agent. Salesforce says the model handles an average of 76% of support volume end-to-end across all channels.
That number deserves scrutiny. Independent assessments from late 2025 put Fin's actual resolution rate closer to 67%, with substantial variation depending on knowledge base quality and how aggressively the system is configured to attempt resolution rather than defer to a human. Fin says performance has improved since then. Salesforce's announcement uses 76% without qualification. Operators deploying Fin on the $0.99-per-resolution billing model should benchmark their own rates rather than assuming the headline figure.
30,000 Customers
The existing customer base may be the most concrete asset in this deal. Fin serves over 30,000 companies - mainly small and mid-market businesses that found Agentforce too complex to deploy. Those customers represent immediate revenue and a large cross-sell surface for other Salesforce products, with no sales cycle required.
The Engineering Team
Fin's engineers built something Salesforce repeatedly said it wanted but hadn't yet delivered: a customer service agent that organizations can activate in hours rather than quarters. Salesforce CEO Marc Benioff acknowledged this in Monday's announcement: "Fin brings proven agent technology, a deep commitment to customer success, and an incredible AI team that will complement Agentforce."
Read plainly, that's an admission that Agentforce needs help in the speed-to-value category.
The Agentforce Paradox
Agentforce is growing fast. It reached $1.2 billion in annual recurring revenue in Q1 of Salesforce's fiscal year 2027, up 205% year-over-year. Those are real paid deployments at scale. But Agentforce is built for enterprise transformation - deep integrations, custom workflows, dedicated implementation partners - and months of configuration before the first agent handles a live conversation.
Fin's inbox-style interface, shown in a 2025 product update, consolidates chat, Slack, Discord, and other channels so the Apex agent can work across all of them from a single workflow.
Source: intercom.com
That approach works for large enterprises with the budget and headcount to support it. It leaves out the rest of the market. A company with 200 employees and 3,000 support tickets a month doesn't have a Salesforce implementation team. It has a support manager who needs something working by next week.
Fin solves exactly that problem. Connect a knowledge base, configure escalation rules, and the agent handles inbound queries without further setup. The $0.99-per-resolution pricing means costs track directly with value delivered. Salesforce plans to position Fin for organizations needing speed and Agentforce Contact Center for enterprises needing customization - a two-tier strategy that's coherent on paper. Whether customers experience it as a clear product choice or as portfolio confusion remains to be tested.
Fin's Path to This Deal
From Live Chat to AI Agent
Intercom launched around 2011 as a live chat widget that any website could embed. For most of its history, it competed on the strength of its messaging UI and a strong integration ecosystem. The company rebranded to Fin earlier in 2026, marking a deliberate break from its customer service software identity toward a product built completely around autonomous agents.
The Apex model wasn't a bolt-on. Fin's team says the architecture was designed from the ground up to match the patterns of support conversations: high query repetition, inconsistent user phrasing, the need to know when not to attempt a resolution and hand off instead. Fifteen years of Intercom conversation data informed that training. That provenance is part of what Salesforce is buying.
The 76% Number
Fin reports resolving over 40 million conversations with agents and claims resolution rates that exceed those of off-the-shelf frontier models on standard support workloads. The 76% figure is self-reported. Third-party analysis from late 2025 found closer to 67%, with heavy dependency on the quality of the content the agent can draw on.
The gap between those numbers isn't necessarily dishonest. Resolution rates truly vary with knowledge base quality. An Apex model fed well-maintained documentation will outperform the same model drawing from a neglected one. Buyers should treat 76% as an upper bound, not a guaranteed baseline.
What Competitors Face
Zendesk, ServiceNow, and Freshworks each have AI agent products in this category. None of them operate behind the Salesforce distribution network. That distribution advantage is the real structural change this deal creates - Fin can now land in every account where Salesforce already runs the CRM, creating an upsell path that competitors can't reproduce without building a CRM from scratch.
AI agents like Fin are taking on the volume of queries that once required large teams of human support agents across industries.
Source: pexels.com
The AI agent market reached $7.6 billion in 2025 and is projected to grow at nearly 50% annually. In that context, $3.6 billion for a product already deployed at 30,000 companies starts to look less like a premium and more like a market-rate bet on category consolidation before the segment gets any more crowded.
The deal is expected to close in Q4 of Salesforce's fiscal year 2027, subject to regulatory approval. Eoghan McCabe remains CEO of Fin, with Des Traynor continuing to lead R&D.
Sources:
