AI Handles Half of Support Cases

This points to support shifting from a labor scaling problem to a software governance problem. When Fin resolves roughly half of inbound volume, companies do not usually fire half the team overnight, they stop backfilling attrition, and repurpose remaining agents toward edge cases, angry customers, and teaching the system through better docs, policies, and escalation rules. That is why AI changes both headcount and the day to day shape of support work.

• The economics are stark. AI support agents are now priced around $0.99 to $1.50 per resolution, versus roughly $10 to $15 for a human handled resolution. In a representative mid market deployment, 45,000 annual resolutions can cost about $63K with AI versus $450K plus with human agents.
  2 sacra 4 sacra
• What replaces the old support role is a tighter loop between human reps, help docs, and the bot. In Intercom's model, a rep answers a new issue, updates the knowledge base or policy, and the next similar case gets handled automatically. The team becomes the source of first answers and the reviewer of last resort.
  3 sacra
• This is also why vertically integrated vendors have an advantage. When the bot, inbox, customer data, and help center live in one system, the handoff from AI to human and back into documentation is much cleaner than stitching together separate tools. That product loop is part of how Intercom competes with Zendesk and newer agent vendors like Sierra and Decagon.
  1 sacra 2 sacra 3 sacra 5 sacra

The next step is from answering questions to taking actions. As agents get better at refunds, cancellations, onboarding, and voice support, support teams will keep shrinking at the repetitive end and grow in importance at the high judgment end. The winners will be the platforms that pair strong resolution rates with strong control, reporting, and workflow tooling.