Decagon AI features and pricing (October 2026)

Decagon is a conversational AI platform from San Francisco for building customer service agents that answer on chat, email, voice and SMS. It is aimed at enterprises and lets support teams define agent behavior in natural language.

Teams describe workflows as Agent Operating Procedures (AOPs), which combine natural-language instructions with code-level control and Git-based versioning. The platform connects to ticketing systems, CRMs, knowledge bases and contact center software and includes testing, experiments, insights and reporting. Watchtower adds quality monitoring for AI and human conversations, and Suggestions keeps the help center current. Decagon AI, Inc. is based in San Francisco and backed by investors including a16z, Accel and Index Ventures. The vendor names enterprises such as Avis Budget Group, Chime and Oura Health as customers, so the product suits companies with their own support organization and high contact volume.

• Resolves customer conversationsAI agents follow the AOPs a team writes and act in connected systems, for example to verify identity or process refunds under strict controls. A supervisor model checks responses for hallucinations before they are sent. Support teams handle fewer repetitive tickets across chat, email and voice.
• Builds and improves agent proceduresDuet analyzes past customer interactions to generate AOPs, finds gaps in guardrails and creates tests for edge cases. Updates are staged for human review with a health report. Customer experience teams change agent behavior without waiting for engineers.
• Reviews conversation qualityWatchtower checks AI and human conversations against criteria written in plain language, such as signs of frustration or privacy violations, and can score them. Findings are grouped into clusters and tracked over time. Quality teams see risks without sampling transcripts by hand.
• Keeps the knowledge base currentSuggestions detects topics where the AI agent falls short and drafts new help articles based on how human agents resolved the same issues. The team reviews drafts before publishing, so documentation keeps pace with product changes.

• Choice between per-conversation and per-resolution pricing, with no charge for escalations under per-resolution pricing.
• Zero data retention with AI providers such as OpenAI and Anthropic, and automatic redaction of sensitive data using Google's DLP service.
• AOPs mix natural-language instructions with code, and technical teams can track versions in Git.

According to its privacy policy, Decagon AI, Inc. is headquartered in the United States, with an address on Market Street in San Francisco.

Choose between per-conversation and per-resolution pricing and ask how a resolution is defined. At our September 2026 review, no hosting region was published; production runs on Google Cloud, so request data location and the DPA. Watchtower can score conversations handled by human agents, so check workplace monitoring and employee consultation rules wherever your support staff work before switching this on.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews