How Intercom Rose From the Ashes by Betting Everything on AI
Eoghan McCabe
Co-founder & CEO, Intercom
AUG 21 2025
The Bet
Intercom Was Dying. AI Saved It.
"We saw AI coming and had a choice: treat it as a feature, or bet the company on it. We bet the company. It was the right call."
Intercom in 2023: slowing growth, increasing competition, existential pressure
The AI bet: rebuild the entire product around AI-first customer support
Fin — Intercom's AI customer service agent — became the fastest-growing product in company history
The lesson: betting everything on a platform shift, when you can see it coming, is not a gamble — it's a strategy
Framework
Intercom's AI Transformation
40%
of support tickets resolved by Fin
50%
reduction in human agent workload
2×
NPS improvement post-Fin
Fin: AI agent that handles tier-1 support with near-human quality
The pricing bet: charge per resolution, not per seat — aligns incentives with outcomes
Enterprise expansion: AI support unlocks markets Intercom couldn't reach with human-only pricing
The moat: 15+ years of customer conversation data trained into Fin
Eoghan's ruleWhen you can see a platform shift coming, move toward it aggressively. The companies that wait to see if it's real are the ones that miss it.
The AI Customer Service Market
What Intercom Learned
Customer trust: 60% of users prefer AI-resolved tickets when resolution is fast and accurate
Quality bar: AI support must be right 95%+ of the time — accuracy > speed
Human complement: AI handles volume; humans handle complexity and unhappy customers
The data flywheel: Every Fin conversation trains the next version of Fin
The pricing innovation
Charging per resolution aligns Intercom's incentives with customer outcomes. It's the most customer-centric pricing model in SaaS.
The moat realization
15 years of support conversations is not a liability (legacy data) — it's an asset (training gold). Perspective shift was crucial.
Playbook
Bet on Platform Shifts
When you see a platform shift, don't wait for competitors to validate it — move first
Rebuild the core product, not just add features — incremental AI integration loses to AI-native competitors
Design pricing around outcomes, not inputs: per resolution > per seat for AI products
Invest in the data flywheel: every customer interaction should train your AI
The rebirth lessonIntercom's near-death experience was the forcing function for an AI transformation that would have taken 5 years without existential pressure.
Contrarian
SaaS + AI Myths
✗Add AI features to the existing productINSTEAD →✓ Rebuild the product around AI. Feature additions lose to native rebuilds in every platform transition.
✗Customers won't accept AI supportINSTEAD →✓ Customers will accept AI support when it's fast, accurate, and available at 3am. They prefer it.
✗Per-seat pricing is sacrosanctINSTEAD →✓ Per-resolution pricing aligns with AI value delivery. Intercom's fastest growth came after changing the pricing model.
✗Your legacy data is a burdenINSTEAD →✓ Your legacy data is your training set. The company with the most domain-specific data wins the AI transition.