
Intercom closed 2025 with a flurry of product releases, shipping more than 200 new products and features across its Fin AI agent and Helpdesk platforms. The company announced Fin 3, the latest version of its AI customer service agent designed to handle more complex queries across expanded channels.
The headline metric: Fin now resolves an average of 67% of customer inquiries across all Intercom customers, up from previous benchmarks. The company announced on LinkedIn that this increase came alongside improvements to the Fin Flywheel, its workflow system for training, testing, deploying, and analyzing AI agent performance.
What's New in Fin 3
Fin 3 extends the AI agent's capabilities to handle more sophisticated customer requests. While Intercom didn't detail every enhancement in the LinkedIn announcement, the release follows a pattern the company established earlier in 2025 with infrastructure updates like the Fin Flywheel continuous improvement system.
The new version also expands channel support. Intercom has been pushing multi-channel AI coverage throughout the year, demonstrated by customer deployments like IG Group's 70% deflection rate across financial services touchpoints.
Intercom's 67% average resolution rate represents a significant benchmark in the AI customer service space. For context, traditional chatbots often struggle to break 30-40% resolution without heavy human intervention. The metric suggests Fin is handling two out of every three inquiries end-to-end, though resolution rates vary widely by industry, query complexity, and how companies configure their knowledge bases.

The Fin Flywheel in Practice
The Fin Flywheel workflow gives support teams structured methods to improve AI performance over time. Teams can identify where Fin struggles, add training data, test changes before deployment, and measure impact through built-in analytics. This closed-loop system addresses one of the biggest pain points in AI customer service: knowing what to fix and whether fixes actually work.
Intercom has been emphasizing speed-to-value in recent releases. The company reduced integration time dramatically with products like Fin for Salesforce, which cut Freepik's implementation from six months to days. That focus on deployment velocity shows up in the Flywheel's design, which automates much of the testing and validation work that previously required manual QA.
The 200+ releases span both major features and incremental improvements. Intercom has been shipping quality controls like Monitors for AI quality control alongside foundational model updates. The company also opened its vertical customer service models to competitors, signaling confidence in its platform approach.
What This Means for Support Teams
A 67% resolution rate changes support economics. If an AI agent resolves two-thirds of inquiries, human agents can focus on complex cases that actually require judgment, empathy, or creative problem-solving. The math works in Intercom's favor: higher resolution rates mean lower cost per ticket and faster response times without hiring more support staff.
But averages hide variation. Some Intercom customers likely see resolution rates above 80% for straightforward product questions, while others handling technical troubleshooting or emotional customer issues may see rates below 50%. The Flywheel workflow exists precisely to help teams identify and close those gaps.
Intercom hasn't disclosed pricing changes tied to Fin 3 or whether the 200+ releases include any breaking changes to existing Fin implementations. Companies running earlier versions of Fin will want to review migration paths and whether new capabilities require configuration updates.
The company's pace of 200+ releases in a year works out to roughly four product updates per week, assuming consistent velocity. That cadence suggests Intercom is treating AI customer service as a rapidly iterating space rather than a set-and-forget deployment, which aligns with how most teams experience AI products in practice.
Sources
1 checkedHow we cover tool news: Create With's tool desk drafts these reports with AI from the sources listed above and checks them against those sources before publishing.



