Intercom's decisive bet on AI
The clearest outside teardown of how betting the company on an AI support agent took Fin from $1M to $100M+ ARR.
Open growthunhinged.com →LLM agents can now read your help docs, past conversations, and backend data, which is why Intercom's Fin resolves around 56% of conversations on average and Decagon deployments at companies like Bilt report 75% resolution. The 'should' is more nuanced: Klarna famously replaced the work of 700 agents with AI, then publicly rehired humans in 2025 after quality complaints. The emerging consensus is to let AI own the repetitive tier while guaranteeing customers a real human path for complex or emotional cases.
A quick orientation. The real value is below: resources worth your time, from people who've actually done it.
The clearest outside teardown of how betting the company on an AI support agent took Fin from $1M to $100M+ ARR.
Open growthunhinged.com →Decagon's CEO explains exactly which tickets AI can genuinely resolve today and where it still breaks.
Listen on Spotify open.spotify.com →A sharp market-level view of why support became the first big AI agent category and what buyers actually get.
Open a16z.com →The former Salesforce co-CEO's argument for why every brand's AI agent becomes its front door.
Open sierra.ai →Taylor connects AI support resolution to the deeper shift from selling software to selling outcomes.
Open sequoiacap.com →A grounded profile of Decagon with real customer names and numbers, useful for calibrating the hype.
Open forbes.com →The most cited cautionary tale: Klarna's CEO admitting the AI-only cost cutting 'went too far'.
Open forbes.com →Survey of 2,400+ support teams showing real adoption and resolution numbers, not vendor marketing.
Open intercom.com →Concrete numbers from a product founders know: 34% faster resolution and a 3.4% ask-for-human rate.
Open decagon.ai →YC partners explain why support was the first vertical AI agents cracked and what that signals for founders.
Open ycombinator.com →The category-defining AI support agent; its public resolution benchmarks set the bar everyone quotes.
Open fin.ai →An unusually candid founder conversation about what it actually takes to make support agents work in production.
Open upstartsmedia.com →A punchy data-backed post on Fin's jump from 25% to 65% resolution, good for the 'can it really' skeptics.
Open linkedin.com →