Decagon Review 2026: Pricing, Features and Verdict

What it is

Decagon makes AI agents that handle customer service conversations across chat, email, voice and SMS. It was founded in August 2023 by Jesse Zhang and Ashwin Sreenivas, is based in San Francisco with offices in New York and London, and in January 2026 announced a $250 million Series D led by Coatue and Index Ventures at a $4.5 billion valuation, according to Wikipedia.

Decagon pitches itself as an "AI concierge": one agent that knows the customer across channels, takes actions in your systems and escalates to humans with context. The platform has four parts on its site: Build (AOPs, workflows, integrations), Optimize (experiments and simulations), Scale (insights, Watchtower QA, knowledge suggestions) and Duet, an AI partner that helps teams build and improve agents.

It is not a helpdesk. Decagon sits on top of Zendesk, Salesforce, Intercom, Kustomer or a contact center platform and hands off to your human team there.

Pricing explained

Decagon publishes no prices. Its own post on pricing, written by its director of product, describes two contract models:

Model	How it bills	What Decagon says
Per conversation	A fixed rate for every incoming conversation, lower at higher volumes	Most customers choose this; costs are predictable
Per resolution	A higher fixed rate for each fully resolved conversation, no charge for escalations, lower with larger commitments	Available, but less common

Decagon argues for per conversation billing because it avoids disputes over what counts as a resolution, such as a frustrated customer who stops replying. That is a fair point, and it is the opposite of how Intercom Fin bills assumed resolutions. The tradeoff is simple: on per conversation pricing you pay for every conversation the AI touches, including the ones it escalates.

To compare a Decagon quote with published prices, convert it to cost per resolved conversation. If Decagon quotes a per conversation rate and resolves 70% of conversations, divide the rate by 0.7. For context, Fin is $0.99 per outcome, Zendesk is $1.50 to $2.00 per verified resolution and HubSpot is $0.50; see the AI Price Index.

Key features

Agent Operating Procedures

AOPs are the core idea. You write workflows in natural language, inbound or outbound, the way you would train a person ("check eligibility, offer the upgrade, if declined, apply the credit"). Decagon says CX teams can build and change them without engineering or vendor professional services, while engineers keep control through Git based version tracking. Guardrails for brand voice, escalation rules and hallucination checks apply across chat, voice and email. You can trace why the agent did what it did at any step.

Voice

Decagon Voice runs on Chord, Decagon's own speech model. Published features include interruption handling, talking while the agent works in the background, hundreds of voices including custom tuned ones, 70+ languages with automatic detection and switching, outbound calling, accuracy across accents and noise, and escalations that pass a summary to the human. Chime's COO is quoted on Decagon's site about combining voice with cross channel memory, and Chime reports 70% chat and voice resolution.

Chat and email

On chat, Decagon emphasizes safe, on brand replies; on email, contextual resolutions with routing through Zendesk and Intercom. Interactive actions run through integrations, so the agent can look up orders, change plans or issue credits where your systems allow.

Testing, experiments and QA

Decagon includes simulations at scale to test agents before launch, live A/B experiments to compare versions, insights and voice of the customer reporting, Watchtower, an always on QA system that reviews every conversation against your criteria, and suggestions that propose knowledge updates.

Integrations

Listed connectors cover CRMs and helpdesks (Salesforce, Zendesk, Intercom, Kustomer), knowledge bases (Confluence, Contentful), CPaaS and contact center platforms (Amazon Connect, RingCentral) and SIP trunking. MCP and API integrations connect anything else. Decagon also offers Decagon Assist, which drafts summaries, replies and actions for human agents.

Personal agents

Decagon now detects AI agents acting for consumers (for example, a shopper's assistant) and serves them through a dedicated channel with permissions you control. Few vendors in this category address that yet.

What users say

Decagon does not have a public review base large enough to rate on the sites we checked, which is common for enterprise only vendors. The evidence available is customer case studies on its own site: Chime (70% chat and voice resolution), Rippling (32% increase in deflection), and other named brands reporting 80% deflection, 95% cost reduction and $1 million in revenue from fully AI handled conversations. One operations manager quoted says maintaining the previous vendor took half his week. Treat vendor case studies as best case results and ask for references in your industry.

Privacy and security

Decagon's security page is detailed:

• Role based access control and SSO through Okta and Microsoft Entra
• Short lived, scoped JWT tokens for agent access to your systems, discarded after each session
• Voice authentication options for sensitive workflows
• AES-256 encryption at rest and TLS 1.2+ in transit, with managed key rotation
• Zero data retention with AI providers such as OpenAI and Anthropic, so conversation data is not stored or used for training
• Automatic PII redaction in logs and transcripts using Google's DLP service
• Bad actor detection for adversarial prompts and a supervisor model that checks responses for hallucinations
• Tamper protected audit logs, multi region infrastructure, model redundancy and uptime SLAs

The page displays SOC 2, HIPAA, PCI, GDPR, CCPA, ISO and EU AI Act badges, with full reports in its trust center on request.

Who it is for

• Enterprise and fast growing companies with large support volumes across chat, email and phone
• Fintech, travel, health, telecom, media and subscription businesses, which Decagon targets by industry
• CX teams that want to own agent logic in plain language without waiting on engineering
• Buyers who prefer predictable per conversation billing over resolution disputes

Who should skip it

• Small and mid size teams; there is no self serve plan or published price
• Teams that want to pay only when the AI succeeds and do not want to negotiate it; Fin and Zendesk publish per resolution rates
• Buyers who need a helpdesk; Decagon sits on top of one
• Anyone who needs a long public review record before buying

Alternatives

• Sierra: the closest rival, with outcome based pricing and a longer compliance list; see Decagon vs Sierra
• Intercom Fin: published $0.99 per outcome and a full helpdesk; see Intercom Fin vs Decagon
• Ada: enterprise AI agent across voice and digital channels, also quote only
• Zendesk AI agents: if you want AI inside your existing Zendesk

Verdict

Decagon is a strong buy for enterprise support teams that want the AI to be owned by the CX team, not by engineers or the vendor's professional services. AOPs, a proprietary voice model, built in experiments and QA, and a clear zero retention stance with LLM providers put it near the top of the category on features. The weaknesses are commercial: no published prices, a default per conversation model that bills escalations, and a young company with no public review base. Get quotes from Decagon and Sierra, convert both to cost per resolved conversation, and compare against Fin's published $0.99.