Fin AI resolves two-thirds of customer queries autonomously. Here's what the other third looks like — and why the setup matters more than the product.
The headline resolution rate Intercom claims is 86%. Our testing across three real businesses found 54-71%. Both numbers tell part of the truth. Here's the full picture.
We tested Fin AI across three real businesses: a mid-market SaaS product (1,200 support tickets/month), an e-commerce brand (3,400 tickets/month), and a professional services firm (380 tickets/month). The conditions were different, the results were different, and both those differences matter for understanding what you're actually buying.
Resolution rates in our tests: SaaS product 71%, e-commerce brand 67%, professional services firm 54%. Intercom's marketed maximum is 86%. The gap is explained by query complexity and knowledge base quality — not by the AI being worse than advertised, but by the real-world conditions being harder than optimal demo conditions.
Resolution rates: what our numbers mean
A 67% resolution rate means two in three customer tickets are resolved without a human agent touching them. At that rate, a 10-person support team handles the same volume with 7 people. The ROI arithmetic is genuinely compelling for businesses with significant support volume — the math works even at numbers meaningfully below the 86% headline.
What the 54% result at the professional services firm tells you: query complexity matters more than volume. The firm's queries involved nuanced professional judgements, exception handling, and contextual interpretation that Fin handled poorly. If your support queries tend toward complexity and nuance rather than repetition and pattern-matching, adjust your expectations accordingly.
The knowledge base preparation is the most important variable in Fin's performance. Better than the AI, better than the integration, better than the prompting. Get the knowledge base right and the resolution rates follow.Our conclusion after deployment testing
Knowledge base preparation: the factor nobody talks about enough
Fin AI learns from your existing Intercom knowledge base, help articles, and past conversations. The initial training requires significant knowledge base preparation — poorly structured or outdated documentation produces poor AI responses. This is not Intercom's fault; it's the nature of retrieval-augmented AI. But it is the primary determinant of your resolution rate.
Our practical recommendation: budget 40-80 hours of knowledge base cleanup before deploying Fin, regardless of how good your existing documentation looks. Specifically: audit for outdated articles (anything over 6 months deserves review), add explicit exception handling for common edge cases, and structure articles as question-answer pairs rather than narrative prose. These changes typically improve resolution rates by 10-15 percentage points.
Handoff quality: where Fin genuinely excels
When Fin can't resolve a query — that 33% — the handoff to a human agent is excellent. The agent receives the conversation history, Fin's attempted responses, the reason for escalation, suggested knowledge articles, and Fin's draft of a potential response. The context transfer is the best we've tested among AI support tools, and it meaningfully reduces the human agent's time-to-resolution on escalated tickets.
Pricing
- Basic Fin AI
- Limited automations
- Core integrations
- Email and chat
- Full Fin AI
- Advanced automations
- All integrations
- AI summaries
- Full reporting
- Everything in Advanced
- Workload management
- SLA management
- Custom roles
- Advanced security
Who should use Intercom Fin AI?
- SaaS and e-commerce businesses with 500+ monthly support tickets and well-maintained knowledge bases where query patterns are predictable
- Companies where support cost is a clear business priority and resolution rate improvements have a measurable commercial impact
- Teams that can invest the upfront knowledge base preparation that makes Fin's resolution rates approach their potential
- Small businesses with fewer than 300 monthly tickets where the Intercom pricing is hard to justify against the ROI
- Professional services firms with highly complex or contextual support queries where Fin's pattern-matching approach struggles
- Teams without the resources to prepare and maintain the knowledge base that Fin's performance depends on