Last updated: 2026-09-04
Decagon runs autonomous AI agents that resolve customer support issues end to end across chat, voice, email, and SMS, with 80% of production traffic on its own fine-tuned models as of March 2026. It targets airlines, banks, telecoms, and retailers that need high-volume enterprise support automation.
About Decagon
Decagon is an enterprise AI customer support platform founded in 2023 by Jesse Zhang and Ashwin Sreenivas, valued at $4.5 billion after a $250 million Series D led by Coatue Management and Index Ventures in January 2026. The company calls its product an "AI concierge": autonomous agents that resolve customer issues end to end across chat, email, voice, and SMS rather than just pointing users to a help article. Decagon runs on a multi-model architecture, combining foundation models from OpenAI, Anthropic, and Cohere with its own fine-tuned models trained specifically on support conversations. A group of specialized agents cross-checks responses, with a supervisor model flagging hallucinations before a reply reaches the customer, and a persistent "user memory" layer carries context across channels so a conversation can continue from chat straight into a phone call without the customer repeating themselves. Decagon Voice adds phone support for renewal calls, appointment reminders, and proactive outreach, alongside direct integrations with Zendesk, Salesforce, Intercom, and Kustomer that let agents create cases, update records, and tag tickets without leaving the conversation. Decagon serves 100+ enterprise customers in airlines, banking, telecom, and retail, and is best suited to organizations processing 10,000+ support tickets per month with 50+ agents. Pricing is fully custom with no free tier or self-serve signup, and deployments go live through dedicated enterprise sales and implementation support. G2 reviewers cite fast implementation and strong response quality as the most common reasons enterprise buyers renew their contracts.
Screenshots
Pricing
Custom enterprise pricing only, no public rates and no free tier. A $50,000/year platform fee applies regardless of plan, covering all channels, integrations, Agent Operating Procedures, Watchtower QA monitoring, testing tools, and analytics — on top of that, usage is billed either per-conversation (~$0.99 default) or per-resolution, buyer's choice. Typical annual contracts range from $95,000 to $590,000+ depending on ticket volume, with median deals around $400,000/year. Implementation takes roughly 6 weeks and includes a dedicated Agent Product Manager and Forward-Deployed Engineer.
Feature Comparison by Tier
| Feature | Enterprise (Custom) |
|---|---|
| Price | $50,000/yr platform fee + usage; contracts typically $95,000-$590,000/yr, median ~$400,000 |
| Billing model | Per-conversation (~$0.99 default) or per-resolution — buyer's choice |
| What the platform fee covers | All channels, integrations, Agent Operating Procedures, Watchtower QA monitoring, testing tools, analytics |
| Channels included | Chat, email, voice, SMS — one contract, no per-channel add-on pricing |
| Implementation | About 6 weeks from signature to go-live, with legal/procurement review in parallel |
| Onboarding support | Dedicated Agent Product Manager and Forward-Deployed Engineer, white-glove |
| Self-serve or monthly option | None published at any contract size |
Key Features
- Cross-channel AI concierge: A single agent handles chat, email, voice, and SMS with a persistent 'user memory' that carries conversation context across channels so customers never repeat themselves.
- Decagon Voice 2.0: Inbound and outbound phone agents run with sub-second latency, interruption handling, branded caller IDs, and customizable tone for calls like renewals and appointment reminders.
- Duet AOP builder: Launched in March 2026, Duet uses AI to help CX teams draft and refine the Agent Operating Procedures that govern how Decagon's agents behave.
- Deep helpdesk and CRM integrations: Direct API connections to Zendesk, Salesforce, Intercom, and Kustomer give agents full Customer 360 context to create cases, update opportunity stages, and auto-tag tickets.
- In-house fine-tuned models: As of March 2026, 80% of Decagon's production traffic runs on models it trained in-house on support conversations rather than general-purpose models from OpenAI or Anthropic.
- 70% combined resolution rate: A published case study reports Decagon agents resolving 70% of chat and voice interactions without human escalation.
Pros
- Independently reported resolution rates translate into meaningfully fewer live-agent escalations, a differentiator G2 reviewers frequently cite over rivals with less mature automation.
- The cross-channel memory layer is the most-cited reason enterprise buyers pick Decagon over single-channel competitors, since agents don't lose context switching between chat and voice.
- Buyers already running Salesforce or Zendesk report the fastest time-to-value of any AI concierge they evaluated, since existing case and ticket data is usable on day one with no separate data migration.
Cons
- Enterprise-only custom pricing with a mandatory annual platform fee on top of usage charges, pricing out SMBs and startups regardless of ticket volume.
- G2 reviewers report difficulty tracing why the AI agent took a specific action and describe permissioning as rudimentary, complicating QA, audit, and compliance review for regulated industries.
- Implementation typically takes several weeks to months and requires a dedicated in-house Agent Engineer to build and maintain Agent Operating Procedures.
- Usage-based pricing has no baseline rate or cap, making month-to-month cost forecasting difficult during ticket volume spikes.
Data Handling
- Training-data policy
- Zero-day data retention is enforced with all third-party LLM providers. Decagon uses customer conversation data to fine-tune company-specific in-house models under enterprise agreements.
- Compliance
- SOC 2 Type II · GDPR · HIPAA (BAA available)
Frequently Asked Questions
What are Decagon's pricing plans in 2026?
Decagon sells custom enterprise contracts only: a flat $50,000 annual platform fee, then usage-based charges per conversation or per resolution. Most deals land between $95,000 and $590,000 a year, with the median near $400,000. Monthly and self-serve options do not exist, and implementation runs about six weeks before go-live.
Does Decagon have a free plan?
No. Every Decagon engagement is a custom enterprise contract with a sales demo in front of it and a multi-week implementation behind it, so there is nothing to test without committing. Teams that need to try before buying usually evaluate self-serve chatbot tools with a real free plan first.
What should you use instead of Decagon?
Sierra is the closest comparison, since it is also an LLM-native enterprise CX platform with similarly custom six-figure pricing. Intercom Fin costs far less at roughly $29 per seat per month, making it a better fit for smaller support teams. Zendesk AI suits teams that want AI bundled inside an existing Zendesk subscription rather than a standalone platform.
What separates Decagon from Sierra?
Both are LLM-native enterprise CX platforms priced in six figures, which is why they get shortlisted together. Decagon's edge is cross-channel user memory plus native two-way write access into Salesforce and Zendesk; Sierra is usually judged on its own agent-building workflow instead. Since neither publishes rates, bake-offs come down to voice latency, integration depth with your helpdesk, and total cost at your ticket volume.
How long does it take to get going with Decagon?
Roughly six weeks from signature, and longer once procurement is counted. The path opens with a sales demo, since nothing about Decagon is self-serve, then its team scopes an Agent Operating Procedure for your highest-volume ticket categories. Connecting Zendesk or Salesforce and running implementation fills the weeks before go-live, with legal review happening in parallel because every contract is bespoke.
Top Alternatives
- Sierra: Sierra offers a comparable LLM-native layer built around a different agent-building workflow. Decagon is the one to back if cross-channel voice and persistent user memory are non-negotiable.
- Intercom Fin: Intercom Fin gets you in the door at $29 a seat. Decagon is a bespoke enterprise build with deep Salesforce and voice work behind it.
- Zendesk AI: Zendesk AI arrives inside a subscription you already hold. Decagon is bought separately and can sit on top of that same Zendesk.
- Ada: Ada is the faster, lighter rollout when chat is the whole scope. Decagon is what you buy when voice, SMS, and CRM write-access come with it.
HokAI guides covering Decagon
- Best AI Customer Support Software in 2026: What Per-Resolution Pricing Actually Costs: AI customer support pricing shifted from per-seat to per-resolution in 2026. See what Decagon, Intercom, Zendesk, Ada and Sierra charge, and model the cost.
- Is Notion AI Worth It? The $10 Add-On Is Gone: Notion dropped its $10 AI add-on in 2025. Full AI now needs the $20-a-seat Business plan, built for agents and custom code, billed separately in credits.
- What Meta AI Is Actually For, Now Its Model Keeps Changing Its License: Meta AI's model went closed in April 2026, then reopened in August. In between, Meta added $7.99-$19.99 subscriptions and $2-per-million-token business billing.