Decagon review

Decagon builds AI agents that resolve customer service conversations across chat, email and voice ¹. Jesse Zhang and Ashwin Sreenivas founded it in August 2023 ²⁰. Customers include Chime, Duolingo, American Airlines, Delta and Oura ¹, and news reports also name Avis Budget Group, Hertz, Block and Deutsche Telekom ¹⁴.

Decagon launched in 2024 with $35 million from a16z and Accel ¹⁶. In January 2026 it raised $250 million at a $4.5 billion valuation, led by Coatue and Index Ventures ¹⁴, and in March it ran an employee tender offer at the same valuation ¹⁵. Its main rivals include Sierra, Intercom and Parloa ¹⁵.

Teams write Agent Operating Procedures (AOPs): workflows in natural language that tell the agent what to do ³. Duet, an AI assistant, drafts AOPs from real transcripts, generates tests and flags likely failures ¹². Guardrails cover sensitive steps such as identity checks and refunds ³. On email, Decagon connects to Zendesk, Salesforce, Intercom, Freshdesk, Gorgias and Kustomer ⁶. The voice agent handles interruptions and can also place outbound calls ⁴.

Simulations run mock conversations across channels before go-live and before major changes ¹⁰. Scheduled test runs and alerts to tools like PagerDuty catch regressions over time ⁹. Agent Versioning adds Git-based tracking, isolated workspaces, approval workflows, audit logs and rollback, and live A/B tests compare workflow changes ^11,3. In production, Watchtower monitors conversations around the clock, and handoffs pass a summary to the human agent ^3,4.

Decagon offers per-conversation or per-resolution pricing and says most customers choose per conversation for predictable bills ². It publishes no rates. Third-party estimates suggest a platform fee near $50,000 a year plus about $0.99 per conversation, with annual contracts from roughly $95,000 to $590,000 ¹⁸; Decagon does not confirm these figures. Plan for a quoted annual contract.

Decagon lists SOC 2 Type II, HIPAA, GDPR, PCI and ISO 27001 ^8,7. It names OpenAI and Anthropic as model providers and enforces zero data retention with both, so conversation data is not stored or used for training ⁷. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit, and PII is redacted with Google's DLP service after each conversation ⁷. Data residency options and BAA terms are not published.

Decagon rates 4.8 on G2, but from only 30 reviews ^17,18. Reviewers flag edge cases that still need a human, setup that needed dedicated staff in earlier versions, and disputes over what counts as a resolution ¹⁸. Named customers with published results include Chime, Duolingo, ClassPass, Oura, Curology and Hunter Douglas ¹.

Decagon declined to take part in G2's AI agent evaluation ²¹.