Waitlist Enabled Controlled AI Rollout

The waitlist mattered mainly because it gave Intercom room to turn a flashy demo into an enterprise grade product before exposing it to tens of thousands of paying customers. Fin was launched into a live support stack where bad answers, weak handoffs, or unclear pricing could damage trust fast, so the gated rollout let Intercom control model capacity, test resolution quality, add importers for outside help centers, and learn which customers actually got to 30% to 78% resolution.

• This was startup style product discovery, done inside a late stage company. Intercom used the waitlist to create artificial scarcity so it could work with a narrow initial customer profile, fix failures, and then widen the aperture once Fin worked on existing docs and later on imported docs from Zendesk, Notion, and other public URLs.
  1 sacra
• The real risk was not awareness, it was trust. In customer support, hallucinations are dangerous because the end user may accept a wrong answer and the business may not even know it happened. That made controlled rollout especially important for testing confidence thresholds, human handoff, and custom rules around sensitive answers.
  1 sacra 2 sacra
• That caution fits the economics of the category. Intercom was moving from seat based SaaS toward 99 cents per AI resolved ticket, while trying to protect its core help desk business. Gating Fin helped it tune product and pricing together, which later helped revenue reaccelerate to an estimated $343M in 2024 as AI lifted revenue per seat and added usage revenue.
  3 sacra 4 sacra

This playbook has become standard for serious AI support products. As the market shifts from simple bots to agents expected to resolve 60% to 80% of conversations, the winners will be the vendors that use controlled launches to harden workflows, integrations, and pricing before scaling broadly. Intercom's waitlist was an early version of that discipline.