Intercom Showcases How Monitors Helps Teams Scale AI Support to 100% Coverage.

Intercom showcased real-world results from its AI-first customer support approach at Fin Labs in Paris, where representatives from healthcare startup Numan and enterprise search company Glean shared how they've scaled support operations using Fin, Intercom's AI agent. Both companies reported resolution rates exceeding 80 percent while maintaining quality standards, according to a panel discussion shared by Intercom.

The session highlighted Monitors, a feature Intercom launched to address one of the biggest challenges in AI-powered support: visibility into automated conversations. Monitors allows support teams to review 100 percent of conversations handled by AI agents, creating feedback loops that help improve both the AI's performance and the underlying knowledge base.

Rhidian Boobier from Numan and Kat Crichton from Glean joined Intercom's Ruth O'Brien to discuss how their teams approach AI implementation. The conversation focused on practical aspects of reaching what Intercom calls "100% involvement" without sacrificing the quality customers expect from human support teams.

Monitors represents Intercom's answer to a common objection support leaders raise about AI agents: the black box problem. Traditional AI implementations often leave teams unable to see what's happening in automated conversations unless a customer escalates or complains.

The feature integrates with Intercom's broader analytics suite, which the company has been expanding to give agents more visibility into AI performance. Teams can now set up automated reviews of conversations based on specific triggers, sentiment scores, or resolution outcomes, then use those insights to refine their knowledge base and improve future interactions.

For support organizations, this addresses a practical bottleneck. Reviewing every conversation manually becomes impossible as volumes scale, but spot-checking only catches issues after they've already affected customers. Monitors sits between these extremes, using AI to surface the conversations most likely to need human attention.

Intercom has been transparent about its own use of Fin. The company previously disclosed that Fin handles 81 percent of its support volume, saving an estimated $9 million annually. The Paris event suggests the company and its customers are now pushing beyond those benchmarks.

The 80 percent resolution rate mentioned by Numan and Glean reflects successful first-contact resolutions without human intervention. This metric has become a key benchmark in AI support, distinct from traditional metrics like first response time or customer satisfaction scores.

Support teams using platforms like Crisp or traditional helpdesk software typically see human agents handle the majority of conversations, with automation limited to basic routing and canned responses. Intercom's pitch centers on flipping that ratio, with AI handling routine inquiries and humans focusing on complex issues or conversations flagged by Monitors.

The Paris showcase reflects a broader shift in how companies approach customer support infrastructure. Rather than viewing AI as a tool for deflection, teams are treating it as a first-line agent that needs ongoing training and oversight.

Monitors makes this operational model more feasible by automating the oversight itself. Instead of manually sampling conversations, support leaders can rely on the system to identify patterns, surface edge cases, and recommend knowledge base updates. This creates what Intercom calls a feedback loop: AI handles conversations, Monitors reviews outcomes, teams update documentation, and AI improves.

For companies evaluating AI support tools, the results from Numan and Glean provide concrete benchmarks beyond vendor claims. An 80 percent resolution rate while maintaining quality suggests these systems have moved past the early adoption phase where AI primarily handled the simplest inquiries.

The full panel discussion is available on Intercom's LinkedIn post, offering additional detail on implementation strategies and lessons learned from teams running AI-first support operations at scale.