Decagon
Enterprise AI customer-service agents for chat, email and voice, built from written operating procedures and sold by demo
Enterprise AI customer-service agents for chat, email and voice, built from written operating procedures and sold by demo
Fast-read signals for fit, pricing, and trust.
Decagon is aimed at a large consumer or fintech brand with high support volume and an engineering-minded support operation that wants to write agent behaviour as procedures and test it before it goes live. There is no self-serve path: no price, plan or trial is published, the product documentation sits behind a sign-in, and every commercial route on the site is a demo request.
Onboarding is sales-led and begins with a demo request. The team connects its help desk, CRM, knowledge sources and telephony, writes the Agent Operating Procedures, runs simulations against them, deploys the agent on chat, email and voice, and keeps watching it through Watchtower QA and the insights views. Decagon's public case studies are large consumer and fintech brands, which is the scale the product is described for.
Decagon publishes no price, plan ladder or self-serve purchase path: its pricing address serves an empty page, every commercial call to action leads to 'Get a demo', and the product documentation requires a sign-in. Its public pages offer neither a free plan nor a trial nor a way to sign up yourself; access starts from the demo.
On data, Decagon states zero-data retention with its AI model providers, meaning no conversation data is stored by them or used for training, AES-256 encryption at rest and TLS 1.2 or later in transit, SSO through identity providers such as Okta and Microsoft Entra, and automatic PII redaction using Google's DLP service. Decagon states these itself; certifications are documented on its separate Trust Center.
Quick fit check against how you actually work.
What you can actually do with this tool.
Behaviour is defined in natural-language Agent Operating Procedures, which the team can read and change, rather than in flow diagrams. That is the design choice that distinguishes how Decagon is configured.
Simulation testing and A/B experiments run before rollout, and the Watchtower QA layer and insights monitor conversations afterwards.
The voice product adds automatic language detection and switching across 70+ languages and outbound calling. Which channels a contract includes is not published.
Salesforce, Intercom and Zendesk for tickets and CRM; Confluence, Contentful and Kustomer for knowledge; Amazon Connect and RingCentral for telephony; APIs, MCP and SIP trunking for the rest. The vendor calls the list illustrative.
Tools and platforms this connects with.
Common questions about this tool, answered.
No price, plan ladder or self-serve purchase path is published; the only commercial route on Decagon's site is a demo request, and pricing comes out of that conversation. Decagon's public pages offer no free plan or trial to fall back on.
Chat, email and voice are the channels the agent serves, with 70+ languages on voice; which channels come with a particular contract is not something Decagon publishes. The named integrations are Salesforce, Zendesk, Intercom, Confluence, Contentful, Kustomer, Amazon Connect and RingCentral, with APIs, MCP and SIP trunking covering anything else, and Decagon calls that list illustrative.
Not according to Decagon: it states zero-data retention with its AI model providers, which it describes as no storage of conversation data by them and no use of it for training. Its Trust Center holds the certifications behind that statement.