At the Pioneer 2025 conference, Intercom launched the brand-new generation of AI customer support agent—Fin 3. It not only addresses routine inquiries but also deeply understands complex issues and delivers context-rich support across multiple channels.
Fin 3’s core goal is to help customer service teams boost efficiency and service quality across diverse interaction scenarios. The new version upgrades intelligent comprehension, response depth, and channel adaptability, featuring stronger learning and reasoning capabilities, a refined training system, and quantifiable performance insights.
Meanwhile, Fin 3 expands support for mainstream channels such as Slack and Discord, and significantly upgrades Fin Voice for more natural, fluid voice interactions.
Next, let’s explore Fin 3’s key features and experiential innovations.

A new way to train Fin to solve complex queries

In customer support, not every issue can be solved in a single sentence. Scenarios like return requests, account verification, technical troubleshooting, or fraud investigations often involve multiple steps, system interactions, or cross-team approvals.
To ensure AI can truly handle these complex, time-consuming requests, Fin 3 introduces the new “Procedures” feature—enabling companies to train Fin to manage complex workflows just like onboarding a team member, while preserving a natural conversational experience.
With Procedures, you can configure Fin to strictly follow standard operating procedures (SOPs), showcasing human-like judgment and flexibility while executing automatically. This allows Fin not only to “understand the issue” but also to “follow the rules,” resolving complex queries in a more controlled, efficient way.
Procedures consist of four core capabilities:
01. Natural language instructions
Train Fin as you would a new employee. You can directly paste existing SOP documents or describe the operational logic in natural language. AI can also generate complete workflows from brief outlines, requiring no code or engineering involvement so anyone can get started easily.
02. Deterministic control
When processes demand higher precision or structure, Fin supports adding deterministic logic to ensure consistency and control. Through data connectors, Fin can view information or execute actions directly within your systems, maintaining real-time data awareness throughout automation.
Conditional steps (such as whether to approve a refund) help Fin make clear decisions at critical junctures, stabilizing outcomes and improving predictability. For critical tasks, you can insert code snippets to guarantee identical outputs for identical inputs.
Additionally, you can set checkpoints so Fin pauses for human approval before executing sensitive actions or hands off tasks to teammates when needed. These flexible controls collectively ensure Fin’s reliability and safety, enabling precise execution even in complex business environments.
03. Proactive behavior
Traditional workflows are often linear, yet real conversations constantly shift—customers may interrupt or change direction at any moment. Fin can intelligently interpret conversational context, dynamically adjust its path, jump to the appropriate step, or trigger other workflows when necessary, achieving flexible yet robust conversational automation.
04. AI assistant support
AI assistants make it easier for teams to author and maintain workflows. Starting from short summaries or knowledge documents, they can automatically generate drafts, suggest control conditions, or produce API connector boilerplate code, sharply reducing design complexity and accelerating iteration.
These elements work together so Fin embodies the flexibility of human thinking and the precision of software execution. Teams can begin with no-code or low-code workflows and collaborate with engineering when needed to realize truly controllable intelligent automation.

Ensuring Fin is stable and reliable before launch

In complex support scenarios, even meticulously designed workflows can encounter unexpected branches, logic gaps, or response deviations during real conversations. To make sure Fin performs reliably before going live, the Simulations feature enables teams to validate its behavior in advance, uncovering and fixing potential issues before customers ever experience them.
01. Simulate full conversations
Run complete virtual customer dialogues from beginning to end to thoroughly test Fin’s responses. Through a visual interface, you can clearly observe Fin’s decision paths, reasoning, and success or failure at each step, ensuring the flow executes as intended.
02. AI-assisted test creation
No more writing test cases manually. AI automatically generates new test cases, repairs failed scenarios, and continuously optimizes based on results. Fin will even recommend creating additional simulations to boost coverage and stability.
03. Comprehensive regression testing
All simulations are stored in a central library. When business processes, product logic, or policies change, simply click “Run All” for automatic regression testing to quickly detect regressions or potential errors.

A new way to connect with customers across platforms

Customers want support on the platforms they know—be it a website, Slack, Discord, or the phone. For businesses, delivering a seamless, consistent multichannel experience has become a core competitive advantage. To help teams meet these expectations, Fin 3 significantly expands channel capabilities and fully upgrades Fin Voice.
Fin now supports more channels than any other AI support agent. With this update, Fin 3 not only covers more interaction scenarios but also debuts an all-new voice experience, making conversations more natural and trustworthy.
01. A radically improved phone experience
Traditional phone support often frustrates customers: waiting on hold, navigating intricate IVR menus, or dealing with mechanical bots. The new Fin Voice transforms this experience, enabling customers to receive immediate, natural, context-aware voice responses for real-time, intelligent spoken interactions.
Since launch, Fin Voice has been fully enhanced with better guidance prompts, more flexible customization, stronger testing and deployment tools, plus complete transcripts and summaries of every call. These upgrades make Fin Voice more practical and stable, maintaining consistent, high-quality service at scale.
“Fin Voice has exceeded our expectations, enabling us to offer round-the-clock phone support without staffing agents outside business hours. Customers often tell us they’re surprised it isn’t a human—Fin Voice sounds natural, responds contextually, and doesn’t feel robotic at all.” — Rizwan Sherif, Director of Customer Experience at Credit Repair Cloud

02. Deliver instant support in Slack
Fin embeds directly into Slack communities and dedicated Connect channels, providing instantaneous, threaded responses. Replies flow naturally with clear message streams, making it feel like a native Slack capability that helps teams resolve issues more efficiently.

03. Engage customers on Discord
From crypto to gaming, more brands are building customer communities on Discord. Fin can now answer customer questions directly within Discord channels, interacting with users on the platform they use most.

Gain deeper understanding of Fin’s performance

As Fin takes on more complex queries and automated tasks, companies need more than a basic “is it working?” They need to know why it performs a certain way, what influences the outcomes, and how to improve. Fin’s Insights continue to evolve, offering clearer, more actionable performance analytics and optimization suggestions.
01. CX score drivers
CX scores now not only present results but also explain the underlying drivers—such as product feedback, customer sentiment, or answer quality. Teams can more intuitively understand Fin’s performance sources and optimize customer experience accordingly.

02. Topic trend reports
New trend reports automatically detect critical changes in support topics, such as surges in requests, drops in resolution rates, or emerging issues. Teams can quickly spot anomalies and trends without manually sifting through data.

03. Curated topics
With AI, the Topics Explorer automatically groups conversations into topics and subtopics, giving teams a big-picture view of the factors influencing support volume and quality. Users now have full control over these topics—renaming, merging, moving, or creating new ones for more flexible, in-depth analysis.

04. Smarter recommendations
The upgraded recommendations feature detects duplicates and contradictory information, continuously learns from feedback, and allows one-click updates in Zendesk or Salesforce. It also offers concrete guidance on data, actions, and improvements, helping teams resolve complex issues faster and continuously enhance Fin’s overall performance.

A comprehensively upgraded intelligent support agent—Fin makes complex tasks manageable
Fin 3, the most capable AI agent, is now more powerful than ever. It handles deeper, more complex tasks than any other agent and supports more channels, including voice, Slack, and Discord. With workflow training, simulation-based testing, and richer insights for performance measurement, Fin 3 achieves unprecedented reliability and intelligence.
All of this means Fin can not only automate routine tasks but also tackle the complex, time-consuming queries once reserved for humans. Whether it’s multistep business workflows, cross-system operations, or natural, fluent customer conversations, Fin completes them efficiently in a controllable, predictable way—while persistently providing actionable insights and optimization suggestions.
Thanks to these upgrades, enterprise teams can confidently entrust critical customer service processes to Fin 3, making AI a true cornerstone for boosting efficiency and enhancing experiences.