Decagon

Decagon is an AI agent platform for customer support. Support and product teams define what the agent should do using what Decagon calls Agent Operating Procedures, written in plain language rather than a visual tool or code. The same agent logic deploys across voice, chat, and email at once, so a team updates one set of instructions rather than three. A tool called Watchtower monitors conversation quality on an ongoing basis, and built-in A/B testing lets a team compare two versions of agent behavior before fully switching over. Companies like American Airlines, Delta, Duolingo, and Deutsche Telekom use it in production. It integrates with Salesforce, Zendesk, Intercom, and Stripe and reports deflection rates as high as 80 percent in some deployments.

This fits a company that wants to adjust its AI agent quickly using support and operations staff rather than developers, with no engineering sprint for every change. Pricing is not public and does not appear on independent listings either, and the platform is generally positioned toward larger support operations, so a smaller team should ask directly about minimum commitments and typical contract size before assuming it fits. A reasonable pilot tests one Agent Operating Procedure on a single high-volume request type in chat first, using the A/B testing tools to compare it against current human-handled results. Because most public case studies come from very large, high-volume brands, ask for a reference customer closer to your own size before committing.