Decagon
by Decagon AI, Inc.
AI concierge agents that resolve customer support across chat, email, voice and SMS
Decagon builds enterprise AI customer-service agents that resolve tickets end to end across chat, email, voice and SMS rather than deflecting them to a help centre. Support and CX leaders define behaviour in natural-language workflows instead of code, then test, A/B and monitor those agents with a QA stack aimed at making autonomous resolution auditable enough to trust in production.
Decagon develops enterprise AI agents that handle customer-service interactions end to end across chat, email, voice and SMS on one shared intelligence layer. It was founded in August 2023 by Jesse Zhang and Ashwin Sreenivas — who previously sold Lowkey to Niantic and Helia to Scale AI respectively — and is headquartered in San Francisco with additional offices in New York City and London. Its central abstraction is the Agent Operating Procedure, a natural-language workflow definition that lets support operations staff specify agent behaviour without writing code, which is what lets a CX team iterate without a developer in the loop for routine changes. Around that sits an evaluation stack unusual in this category: simulation-based testing at scale, A/B experimentation between agent versions, a Watchtower continuous quality monitor, Voice-of-the-Customer analytics, and AI-generated knowledge suggestions that surface gaps in the underlying documentation. Integrations are native to Salesforce, Zendesk including Sunshine, Intercom and Kustomer for tickets and customer context, Confluence and Contentful for knowledge, and Amazon Connect, RingCentral and SIP trunking for voice, with self-serve APIs, custom tool connectors and MCP for anything not pre-built. Compliance coverage spans SOC 2 Type II, ISO 27001:2022, HIPAA, PCI DSS, GDPR and CCPA, with zero-day retention at the LLM providers. Named customers include Chime, Duolingo, Block, Affirm, Avis Budget Group, Deutsche Telekom, Oura, American Airlines, Ticketmaster, Snap and Fanatics; Chime reports 70% chat and voice resolution and Duolingo an 80% deflection rate. In January 2026 Decagon raised a $250M Series D led by Coatue and Index Ventures at a $4.5B valuation, roughly triple its mark six months earlier.
VPs of Customer Experience or Support at high-volume consumer businesses — fintech, travel, marketplaces, subscription apps — who are trying to move from deflection to genuine autonomous resolution.
Autonomous resolution of a large share of contacts across every channel, with simulation testing and continuous QA so the behaviour is auditable before it reaches customers.
At a Glance
- Category
- AI Agents & Orchestration
- Pricing
- Contact for pricing, Usage-based
- Target Market
- CIOs, CTOs, VP Customer Experience, Support Operations Leaders
- Deployment
- Cloud-only, API-based
- Founded
- 2023
- Headquarters
- San Francisco, United States
- Team Size
- 201-500
- Customers
- 100+ new global enterprise customers added in the last fiscal year (company-reported, January 2026)
Key Features
- ✓Agent Operating Procedures
Natural-language workflow definitions let support teams specify and change agent behaviour without writing or deploying code.
- ✓Omnichannel agents on one layer
Chat, email, voice and SMS run off a single intelligence layer, so context and policy do not fork per channel.
- ✓Simulation testing and QA
Agents are exercised against simulated conversations at scale before release, which is what makes autonomous resolution reviewable.
- ✓A/B experimentation
Competing agent versions can be tested against live traffic to measure resolution and containment before a full rollout.
- ✓Watchtower continuous monitoring
Ongoing quality assurance flags degrading conversations in production rather than waiting for a customer complaint.
- ✓Voice of the Customer analytics
Aggregates conversation data into themes so product and support leaders see what is actually driving contact volume.
- ✓Native ticketing and voice integrations
Salesforce, Zendesk, Intercom, Kustomer, Confluence, Contentful, Amazon Connect, RingCentral and SIP trunking connect without custom code.
Capabilities
Use Cases
- •Tier-1 ticket deflection at scale
Resolve high-volume repetitive contacts autonomously so human agents only receive escalations that need judgement.
- •Voice IVR replacement
Replace menu-tree phone systems with a natural-dialogue agent over Amazon Connect, RingCentral or SIP trunking.
- •Regulated-industry support automation
Automate fintech, healthcare or airline support where SOC 2, HIPAA and PCI coverage plus PII redaction are contractual prerequisites.
- •Knowledge-gap discovery
AI-generated suggestions surface missing or stale help-centre articles that the agent could not answer from.
- •Pre-release behaviour testing
Simulate thousands of conversations against a new Agent Operating Procedure before it touches real customers.
Ideal For
Best For
- ✓High-volume consumer support organisations where ticket growth outpaces headcount budget
- ✓Regulated industries needing SOC 2 Type II, HIPAA, PCI DSS and ISO 27001 coverage on an autonomous support agent
- ✓Teams consolidating chat, email, voice and SMS onto one agent instead of separate per-channel bots
- ✓CX operations groups that want to author and A/B test agent behaviour in natural language rather than code
- ✓Enterprises already standardised on Salesforce, Zendesk, Intercom or Kustomer who need deep native ticket and identity integration
Not Ideal For
- ✗Startups and SMBs — there is no self-serve signup, no trial and no public pricing, and independent coverage reports contracts typically starting in six figures annually
- ✗Teams that need to evaluate before talking to sales; there is no public documentation site and no free tier to prototype against
- ✗Non-technical teams attempting advanced custom workflows or new API integrations, which reviewers say still require developers on standby
- ✗Organisations that require full explainability of every agent decision, since a recurring complaint is that it is hard to see why an agent chose a given path
Deployment
Market & Ratings
100+ new global enterprise customers added in the last fiscal year (company-reported, January 2026)
Market Analysis
Pros
- ✓Named production deployments at recognisable scale — Chime, Duolingo, Block, Affirm, Deutsche Telekom, Avis Budget Group, American Airlines — with Chime reporting 70% chat and voice resolution
- ✓Compliance coverage (SOC 2 Type II, ISO 27001:2022, HIPAA, PCI DSS 4.0.1, GDPR, CCPA) clears procurement in regulated industries that usually block AI support pilots
- ✓Natural-language Agent Operating Procedures put day-to-day behaviour changes in the hands of support operations rather than engineering
- ✓The testing and QA layer — simulations, A/B experiments, Watchtower — is the part most competitors treat as an afterthought
Cons
- ✗Reviewers consistently describe the agent as a black box: it is hard to see why it made a particular decision, which complicates review and debugging when it gets something wrong
- ✗Advanced work still needs engineers — custom AOP workflows and new API integrations go beyond what non-technical users can do alone
- ✗No public pricing, no self-serve signup, no trial and no public documentation site, so the product cannot be evaluated without engaging sales
- ✗Usage-based billing on a six-figure floor makes budgets hard to forecast when contact volume spikes, and the contractual definition of 'resolution' drives real billing disputes
- ✗Requires close monitoring in the early weeks to tune behaviour, which pushes out time to full ROI
Pricing
Enterprise (custom quote)
Contact for pricing
- ✓Per-conversation or per-resolution usage billing
- ✓Chat, email, voice and SMS agents
- ✓AOP authoring, simulation testing and A/B experimentation
- ✓Native helpdesk, CRM, knowledge and telephony integrations
- ✓SOC 2 Type II, ISO 27001, HIPAA, PCI DSS and GDPR coverage
Decagon publishes no rate card, has no self-serve signup and offers no trial, so every number comes out of a sales call; independent coverage reports contracts typically starting in six figures annually, which puts it out of reach for most startups and SMBs. Billing runs on one of two usage models: per conversation, where you pay for every interaction the agent touches, or per resolution, an outcome model that costs more per unit but only charges for tickets closed without a human. How 'resolution' is defined in the contract is the term that actually determines the bill, and quotes also move with ticket volume, workflow complexity and unpublished onboarding fees.
Security & Compliance
Connect
Sources
This page was written from 6 sources, 3 on domains other than decagon.ai.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
xpander.ai
Vendor-neutral enterprise AI agent platform with a universal agent harness and Omni, an AI forward-deployed engineer
WSO2 Agent Manager
Open-source control plane for governing AI agents across any framework, model or cloud
Moveworks
Agentic AI assistant and enterprise search for employee IT, HR and finance requests
Convey
Digital teammates that own back-office operations, trained by screen-share not code