In a pioneering move for large-scale customer service platforms, the company formerly known as Intercom has introduced an AI agent specifically designed to manage another AI agent. This new system, named Fin Operator, is engineered for the crucial back-office teams responsible for configuring, monitoring, and refining the company’s customer-facing AI, Fin.
Unlike Fin, which directly engages with customers, Operator targets the growing cadre of support operations professionals. These individuals typically spend their time updating knowledge bases, diagnosing conversation failures, and analyzing performance data. Fin Operator aims to streamline these complex, often tedious, operational tasks.
“Fin is an agent for your customers,” explained Brian Donohue, the company’s VP of Product, in an exclusive interview. “Operator is an agent for your support ops team. This is an agent for the back office team who manages Fin and then manages their human agents.”
This launch follows a significant rebranding, with the 15-year-old company now officially bearing the name Fin, underscoring the central role of its AI agent. Fin has achieved substantial growth, surpassing $100 million in annual recurring revenue (ARR) and growing at a 3.5x rate. The overall company generates $400 million in ARR, indicating that Fin now represents a quarter of total revenue and is the primary driver of its expansion.
Fin Operator is entering early access for Pro-tier users immediately, with general availability anticipated in the summer of 2026.
The Hidden Operational Burden of AI Customer Service
As AI agents like Fin handle an increasing volume of customer interactions—currently resolving over two million customer issues weekly for 8,000 clients, including prominent names like Anthropic and DoorDash—the complexity of managing these systems has escalated dramatically. Ensuring knowledge bases remain current, diagnosing conversational errors, and tracking performance metrics all fall to support operations teams, who are reportedly overwhelmed.
“Almost every support ops team is already doing data analysis and knowledge management—that’s table stakes today,” Donohue noted. “Where teams struggle is the agent builder work. It’s a new skill set, and most don’t have enough time for it. They get their first iteration up and running, and then they get stuck.”
The challenge is systemic. AI customer agents are not static software; they require continuous tuning, akin to training a new employee. Each customer interaction presents potential failure points, necessitating diagnosis, root-cause analysis, configuration adjustments, testing, and ongoing monitoring—a relentless and technically demanding process. Fin Operator is designed to consolidate this entire workflow into a conversational interface.
A Unified AI for Data Analysis, Knowledge Management, and Debugging
Donohue outlined Fin Operator’s capacity to perform three critical functions that typically consume support operations teams’ resources: expert data analysis, expert knowledge management, and expert agent building.
In its role as a data analyst, Operator can respond to queries such as, “How did my team perform last week?” generating real-time charts, trend reports, and in-depth analyses from the data within Fin’s platform. It leverages contextually aware knowledge of customer-specific data attributes to interpret metrics accurately.
As a knowledge manager, Operator can process new product information, such as a detailed PDF on a new feature, and autonomously scan the company’s content repository to identify necessary updates. It can pinpoint content gaps, draft new articles, suggest revisions to existing ones, and present these changes in a clear, diff-style review format. This functionality relies on the advanced semantic search system that Fin has developed over two years.
“On that knowledge management front, you just have such a time compression of something that would take, certainly hours, sometimes days, into the space of about 10 minutes,” Donohue commented.
In its agent-building capacity, Operator introduces a “debugger skill.” Support operations teams can provide a link to a problematic conversation, and Operator will meticulously trace Fin’s reasoning process, pinpoint the root cause (often a piece of guidance that inadvertently creates a loop), propose a revised solution, back-test the change against the original conversation, and recommend implementing a production monitor for similar future issues.
“This is literally what our professional services team does,” Donohue explained. “You’ve written guidance that is unintentionally causing Fin to repeat itself—this happens a lot. You didn’t realize it, but you never gave it an escape hatch.”
The ‘Pull Request’ Safety Net for AI Change Management
A key architectural decision in Fin Operator is its “proposal system,” which mirrors the “pull request” functionality common in software development. All recommended changes—whether edits to help articles, modifications to AI guidance rules, or the creation of new quality assurance monitors—are presented as proposals with a comprehensive diff view. Users can review, edit, and approve each change before it is implemented, ensuring human oversight for all system modifications.
“Right now, we’re taking zero risk on this—Fin cannot make any changes to the system without human approval,” Donohue emphasized. “Nothing goes live until a human clicks apply.”
This deliberate choice to maintain a human approval gate, even in a market increasingly favoring autonomous AI, highlights a commitment to safety and control. Donohue acknowledged that this approach may evolve but stressed the necessity of caution: “It’s too big a leap to just let Operator make changes automatically and then tell the team, ‘Hey, let me tell you about what I did.'” This feature is particularly relevant for enterprise clients focused on compliance, security, and risk management.
Why Fin Operator Leverages Anthropic’s Claude
In a significant technical detail, Donohue revealed that Fin Operator utilizes Anthropic’s Claude models, rather than the company’s proprietary Apex models. The Apex models, which power the customer-facing Fin agent, have been benchmarked against leading models like GPT-5.4 and Claude Sonnet 4.6, demonstrating superior performance in customer service scenarios.
“We’re not using our custom models,” Donohue stated. “Those are designed to directly answer customer questions, whereas these are closer to what frontier models are best suited for. This is really closer to software engineering.”
This distinction underscores the different requirements of the two AI systems. Fin’s Apex models are optimized for accurate and hallucination-free customer conversation resolution. Operator’s tasks—data analysis, complex configuration, and reasoning chain debugging—require a different cognitive profile, one that Anthropic’s Claude models have been specifically tuned for, aligning closely with software engineering capabilities.
While the company has not ruled out developing custom models for Operator in the future, the immediate focus is on the differentiated capabilities built around Claude, including the proposal system, debugger skill, semantic search integration, data attribution logic, and charting features.
Early Adopters Report Significant Productivity Gains
Fin Operator is currently in beta with approximately 200 customers, a number that has grown rapidly in recent weeks. Early feedback indicates substantial benefits.
Constantina Samara, VP of Customer Support, Enablement & Trust at Synthesia, noted, “Previously, improving how Fin handles a conversation often meant reviewing everything yourself—the conversation, the configuration, the content. With Fin Operator, you just ask. It walks you through what happened and makes improving Fin dramatically easier.”
Jordan Thompson, an AI Conversational Analyst at Raylo, reported daily use of Operator and found its analysis to be highly accurate, matching his own manual assessments. He added, “It’s just as strong at high-level trend analysis as it is at debugging individual conversations. That’s a real limitation when using an LLM connector on its own—you get conversational depth but nothing on reporting or trends.”
Donohue also shared an internal anecdote from Beth, who leads knowledge operations, stating that Operator made her feel as if she had “five more people on my team.” This sentiment, particularly regarding the stark time savings in content auditing—compressing hours or days of work into approximately 10 minutes—consistently elicits strong positive reactions from users.
Evolving Pricing Models Reflect AI’s Impact on Enterprise Software Economics
Fin Operator will be integrated into the company’s Pro add-on tier, which already includes advanced analytics features like CX scoring, topic detection, real-time issue detection, and quality assurance monitoring for both AI and human agents.
The pricing structure introduces a new element for the company: usage-based billing. Historically, Fin has utilized outcome-based pricing, charging per conversation resolved by the AI. Operator’s function, which involves configuration changes rather than direct customer resolution, doesn’t align with this model.
“This has pushed us to a different model, to go more into that usage model for support ops teams,” Donohue explained. “We’ll try to be generous with the usage amounts that come into Pro, but for people who are leaning heavily in, we’ll have the ability to buy more usage blocks.”
This shift is noteworthy. Outcome-based pricing was a distinctive market strategy for Fin, predicated on customers valuing results over volume. Applying this to internal operational work proved less feasible, suggesting that as AI agents assume more varied roles within organizations, their supporting pricing models will also need to diversify.
Fin Operator’s Competitive Positioning in the AI Customer Service Arena
Fin Operator enters a highly competitive market, with established players like Zendesk and Salesforce, as well as numerous AI-native startups, developing AI-powered support operations tools. The broader AI automation market is projected to reach $169 billion by 2026, according to Grand View Research, with a compound annual growth rate of 31.4%.
Donohue highlighted two key differentiators for Operator: breadth and holistic coverage. Unlike single-use-case solutions, Operator integrates across the entire configuration spectrum—data, content, procedures, simulations, guidance, and monitoring. Crucially, it spans both AI and human support operations.
“Most critically, where I think we have the most differentiation is because it’s for your human system and your AI system,” Donohue stated. “That’s really one of the unique spaces we have—to have a first-class AI agent and a first-class help desk, and Operator works across both.”
The company’s recent rebranding to Fin, alongside the launch of the Fin API Platform, reinforces its strong commitment to AI. By opening its proprietary Apex models to third-party developers and offering technology licensing, Fin is positioning itself as a leader in the evolving AI landscape. Industry observers suggest this strategic move could be aimed at strengthening its market position ahead of a potential IPO.
The True Paradigm Shift: AI Agents Handling Complex Cognitive Work
Beyond specific product features, Fin Operator signifies a more profound shift: the emergence of AI agents designed to manage other AI agents. This layered abstraction is fundamentally reshaping how companies approach operational software.
Donohue emphasized that the primary paradigm shift is not merely the adoption of chat interfaces over traditional UIs, but rather the AI’s capacity to perform complex knowledge work—determining what changes are needed, why, and how.
“The UX change is secondary, even though it’s most visible,” Donohue remarked. “The change is that we are identifying and doing the work of support operations. It’s doing the work of what the knowledge manager is doing, so that they just have to approve that. That’s the huge shift.”
This aligns with the transformation seen in software engineering, where AI coding agents have shifted developers’ focus from writing code to reviewing and guiding AI-generated code. Donohue anticipates a similar evolution for support operations professionals.
“Software engineers—three months have upended their world, where their primary job now is managing agents who are actually writing the code,” he observed. “Similarly now, support ops, your job is to manage an agent who’s managing the agent for your customers.”
The ultimate success of this vision at an enterprise scale is yet to be determined. The beta phase of Operator is critical for refining its quality through meticulous, conversation-by-conversation debugging. “We’ve spent three months, conversation by conversation, learning, fixing, learning, fixing, to get it where it’s robust,” Donohue shared.
If early indicators hold true, Fin Operator may offer a glimpse into the future of enterprise software: systems where AI agents perform the core work, with humans providing judgment and approval. For customer service leaders deploying AI, the critical question evolves from “How good is my bot?” to “Who is managing it?” Increasingly, the answer points to another bot.
Business Style Takeaway: The introduction of Fin Operator signifies a crucial evolution in AI deployment, moving beyond customer-facing bots to AI systems managing operational complexity. This development highlights the growing need for sophisticated AI governance and offers businesses a new lens through which to evaluate efficiency gains and the future structure of their support operations teams.
Learn more at : venturebeat.com