Decagon

TL;DR. Decagon is an enterprise AI customer support platform that deploys autonomous "AI concierge" agents across chat, email, voice, and SMS. Its Agent Operating Procedures let teams define workflows in natural language. Founded in 2023, it serves 100+ enterprises including Notion, Duolingo, and Chime, reaching a $4.5 billion valuation.

Overview

Decagon is an enterprise AI company that builds, deploys, and operates autonomous AI agents for customer support across chat, email, voice, and SMS channels. The company markets its product as an "AI concierge," positioning its agents not as chatbots that deflect tickets but as systems that resolve customer issues end to end, from answering product questions to processing refunds, canceling subscriptions, and replacing credit cards.

What the product does. Decagon's platform unifies chat, voice, and email within a single intelligence layer, enabling customer experiences to stay consistent across every channel. Its core differentiator is Agent Operating Procedures (AOPs), which let teams define agent workflows in natural language rather than rigid configuration languages or decision trees. The platform includes testing, observability, experimentation, and analytics tools so teams can iterate on agent logic without engineering sprints or vendor tickets. AI agents are built on a mix of third-party large language models from OpenAI, Anthropic, and Cohere, supplemented by proprietary fine-tuned models trained on enterprise data such as knowledge bases, manuals, and historical support conversations. The company reports serving over 10 million customers with an 80% deflection rate, 65% reduction in support operations costs, and a 93% agent quality score.

Founding story. Decagon was founded in August 2023 by Jesse Zhang and Ashwin Sreenivas, who met at a founders' retreat in Utah. Before building Decagon, the two surveyed dozens of startups to identify a product companies would pay for, arriving at AI-powered customer service. Zhang previously founded Lowkey, a social gaming clip-sharing app acquired by Niantic (maker of Pokémon Go) in 2021. Sreenivas founded Helia, an AI startup focused on real-time video, which was acquired by Scale AI in 2020. Both founders also held technical roles at companies including Google, Citadel Securities, and Palantir. Zhang serves as CEO; Sreenivas serves as President (and is listed as CTO in some sources).

Industry and market positioning. Decagon operates in the AI-powered customer service automation market, competing with legacy customer support tools (Salesforce, Zendesk) and a new wave of agentic AI platforms. The company's pitch is that its agents can resolve issues autonomously rather than simply routing or deflecting them. Decagon won many of its initial enterprise contracts through competitive "bake-offs" in which its AI was pitted against incumbent chatbots, with Zhang claiming Decagon won each evaluation. Customers include F100 enterprises across airlines, banks, telecom, and retail (Avis Budget Group, Block, Deutsche Telekom, 1-800-FLOWERS.COM, Hunter Douglas) as well as high-growth technology companies (Notion, Duolingo, Rippling, Bilt, Substack, Oura Health, Affirm, Chime, Eventbrite).

Funding history. Decagon has raised approximately $481 million in total funding across six rounds, reaching a $4.5 billion valuation as of January 2026:

Key investors across rounds include a16z, Accel, Bain Capital Ventures, BOND, Coatue, Index Ventures, Elad Gil, Forerunner, Ribbit Capital, and T.Capital. In March 2026, Decagon completed its first employee tender offer at the same $4.5 billion valuation, giving more than 300 employees the opportunity to sell vested shares.

Company facts. Decagon is headquartered in San Francisco, California, with additional offices in New York City (opened July 2025) and London (opened November 2025). Employee estimates range from approximately 300 to 450, with Tracxn reporting 434 employees as of May 2026 and PitchBook reporting 446. Sacra estimates Decagon hit approximately $35 million in annualized revenue by October 2025, up from $10 million at the end of 2024, with Q3 2025 GAAP revenue and ARR each growing more than 3x year over year. The company added over 100 new global enterprise customers in 2025.

Key links.

• Website: decagon.ai

• About page: decagon.ai/about

• Blog: decagon.ai/blog

• LinkedIn: linkedin.com/company/decagon-ai

Products & Features

Decagon is an enterprise AI concierge platform that automates customer support across chat, email, and voice channels. The core differentiator is its Agent Operating Procedures (AOPs) system, which lets non-technical teams define agent workflows in natural language while engineers retain control over integrations and guardrails, eliminating the need for decision trees or custom code for routine logic changes.

Core Product

Decagon positions itself as a unified conversational AI platform, not a chatbot layer. Its agents resolve customer issues end-to-end by retrieving data from connected systems, taking real actions (refunds, account changes, order updates), and escalating to human agents when needed. The platform is built on a proprietary AI Agent Engine that maintains customer context across sessions and channels, and is described as self-improving through a data flywheel that learns from every interaction.

The product lifecycle follows three phases: Build (define workflows via AOPs with Duet as an AI assistant), Optimize (validate with simulated conversations, unit testing, versioning, and A/B testing), and Scale (leverage analytics, Voice of Customer insights, and Watchtower monitoring). As of March 2026, approximately 80% of platform traffic runs on models Decagon trained in-house specifically on customer support conversations, rather than on third-party providers like OpenAI or Anthropic.

Key Features and Capabilities

Agent Operating Procedures (AOPs): The flagship feature. AOPs let CX teams write business logic in plain English (e.g., when to issue a refund, when to escalate, how to handle specific complaint types), which Decagon compiles into executable agent logic. AOPs can trigger real backend operations including refunds, account changes, and queries to third-party systems. Engineers retain control over core code, guardrails, and integrations. Hundreds of AOPs can be deployed at scale.

Duet: An AI assistant for building and refining agents. Released in March 2026, Duet analyzes customer conversations to identify workflow gaps and automatically generates, tests, and refines AOPs. It compresses days of iteration into minutes and acts as a thought partner with best practices.

Transparent Observability (Trace View): Surfaces step-by-step traceability for every agent decision, showing model calls, workflow triggers, and knowledge articles referenced. Introduced in 2024 to address transparency concerns.

Watchtower: Provides 24/7 always-on monitoring and quality assurance against custom criteria for all customer conversations.

Agent Workbench: Launched Spring 2026, adds autonomous debugging and root-cause analysis for agents.

Versioning and Experiments: Teams can collaborate, safely roll out new versions, and A/B test workflow updates against real conversations with full impact measurement.

Testing and Simulations: Evaluate agent updates with simulated conversations and unit testing before deployment.

Voice of Customer: Uncovers what customers care about most and where they experience friction, turning insights into actionable product roadmap decisions.

Omnichannel Deployment

Decagon unifies three channels within a single intelligence layer:

• Chat: Flexible, personalized chat that executes complex workflows with 24/7 availability in any language. Supports audit logging, proactive gap identification, and conversation insights.

• Email: Always-on email resolution that handles inquiries through Zendesk and Intercom with smooth handoffs. Includes automated email routing.

• Voice: Launched in 2025 in partnership with ElevenLabs for voice synthesis. Handles inbound calls with real-time responsiveness, interruption management, customizable voice profiles (language, style, speed, tone, pronunciation), and seamless human escalations with summaries. The Spring 2026 update added outbound campaign capabilities (renewals, follow-ups, outreach) and persistent user memory across interactions.

All three channels share the same AOP logic and agent "brain," ensuring consistent experiences. Cross-channel memory means customers do not need to repeat themselves when switching channels.

Integrations

Decagon offers pre-built integrations, APIs, and MCP (Model Context Protocol) support across four categories:

• CRM, helpdesk, and ticketing: Salesforce, Zendesk, Intercom, Zendesk Sunshine, and more. Supports two-way sync for ticketing, knowledge base content, and escalation handling.

• Knowledge base syncs: Confluence, Contentful, Kustomer, and similar systems to populate the agent's knowledge base.

• CPaaS and telephony: Amazon Connect, RingCentral, and SIP trunking for voice calls with seamless handoffs.

• Custom and open connectivity: MCP-based open connectivity to any data system, self-serve API integrations, and custom tool integrations for retrieving data and triggering actions.

Integrations enable the agent to take real actions such as processing refunds, updating orders, verifying identity, and creating tickets without escalating to a human agent. The company states many integrations require "no custom code," though non-standard setups require engineering involvement.

Platform Availability

Decagon is primarily a web-based platform accessed through a browser-based console for building, optimizing, and monitoring agents. Customer-facing agents deploy across web chat, email, and voice/telephony channels. There is no standalone mobile or desktop application; the platform operates as a cloud service with API and MCP connectivity for system integration.

Pricing

Decagon does not publish pricing on its website. The platform is enterprise-focused and sold via a sales-led process (demo required). Implementation timelines range from four to twelve weeks depending on integration complexity, and Decagon assigns implementation engineers to guide deployment. No self-serve pricing tiers or free trial are publicly listed, though the company's site mentions a "set up in minutes, no credit card required" prompt on some pages, suggesting limited self-serve exploration may be available.

Recent Product Updates

• March 2026 (Spring Launch): Introduced Duet (AI-powered AOP generation and refinement), Agent Workbench (autonomous debugging), outbound voice campaigns, and persistent user memory across interactions.

• 2025: Launched Decagon Voice with ElevenLabs partnership for AI-powered phone support. Shifted approximately 80% of traffic to in-house trained models.

• 2024: Introduced Trace View for agent decision traceability.

Target Use Cases

Published use cases span inbound support (account access, returns, charge disputes, technical troubleshooting), appointment reminders, reservation booking, inbound lead qualification, trip disruption support, and outbound campaigns. Customers include Chime, Duolingo, Rippling, Figma, Notion, Dropbox, ClassPass, Hertz, Bilt, and Substack across fintech, retail, travel, and software verticals.

Security & Compliance

Decagon maintains an enterprise-grade security posture built around SOC 2 Type II compliance, zero-day LLM data retention, AES-256 encryption, and layered AI guardrails. The platform is designed for regulated industries, with HIPAA options available for healthcare customers and GDPR compliance for European operations.

Certifications and Compliance

Decagon holds SOC 2 Type II certification, confirmed on both its official security page and corroborated by multiple independent sources. The company offers HIPAA compliance options, including Business Associate Agreements (BAAs) for healthcare customers on enterprise contracts. The platform is also GDPR compliant, serving European enterprise customers. Decagon does not publicly list ISO 27001 or PCI-DSS certifications as of the latest available information. The company maintains a public Trust Center at trust.decagon.ai for security documentation access.

Data Encryption and Protection

All data is encrypted at every stage. AES-256 encryption is used for data at rest and backups, and TLS 1.2+ is enforced for all network transmission. Encryption keys are centrally managed with strict access controls, expiration policies, and rotation schedules that maintain continuous protection. Decagon uses Google's Data Loss Prevention (DLP) service to automatically detect and redact personally identifiable information (PII), sanitizing logs and transcripts immediately after conversations end.

Data Retention and LLM Provider Practices

Decagon enforces zero-day retention with all AI model providers, including OpenAI and Anthropic. This means no conversation data is stored or used for model training by these third parties. The company's privacy policy, effective May 6, 2025, clarifies that data processed on behalf of business customers is governed by individual customer agreements, not the general privacy policy.

Enterprise Access Controls

Decagon provides a comprehensive suite of enterprise access controls:

• Single Sign-On (SSO): Integrates with identity providers including Okta and Microsoft Entra, with SAML and SCIM support for automated user provisioning.

• Role-Based Access Control (RBAC): Platform access is defined and enforced by role, ensuring least-privilege permissions for every user.

• Two-Factor Authentication (2FA): Required for application access control.

• Just-in-time API tokens: Short-lived JWT tokens give AI agents real-time access to customer systems, scoped for minimal privilege and discarded after each session.

• Voice authentication: Flexible authentication methods to verify end-users over voice channels for sensitive workflows.

• Audit logs: Comprehensive, tamper-protected logs capture key events from logins to unusual activity, accessible only to senior engineering leadership.

Network and Infrastructure Security

Decagon's production services are hosted on leading cloud infrastructure providers, primarily Google Cloud. The company uses Cloudflare and Google's Virtual Private Cloud (VPC) for network perimeter protection, supplemented by web application firewalls and regular vulnerability scanning. The platform features multi-region infrastructure with model redundancy, autoscaling, auto-failover, and ongoing health checks to maintain platform uptime SLAs.

AI Model Safety Measures

Decagon implements multiple layers of AI-specific safety controls:

• Bad actor detection: A specialized system identifies adversarial prompts, manipulation attempts, and policy violations, deflecting unsafe inputs or escalating to human agents.

• Supervisor model: A built-in supervisor model detects hallucinations before responses are sent to users, automatically revising responses that stray from factual grounding.

• Watchtower: An always-on quality assurance system that reviews every conversation (both AI and human) against custom criteria, flagging compliance risks, PII exposure, and outlier behavior for rapid triage. Watchtower monitors for policy compliance, PII protection, and payment handling risks.

• Guardrails for sensitive actions: Strict guardrails are applied to sensitive workflows like refunds, with layered protections designed to ensure AI agents execute actions safely and within authorized parameters.

Responsible Disclosure and Incident History

Decagon operates a responsible disclosure program, encouraging the community to report security vulnerabilities to security@decagon.ai. The company maintains a public status page at status.decagon.ai for tracking platform incidents. No known data breaches or security incidents have been publicly reported as of the latest available information. The Nudge Security profile for decagon.ai provides additional third-party security assessment data including supply chain details and breach history.

User Feedback & Adoption

Decagon has built strong user sentiment in a short time, with a 4.9/5 rating on G2 (from approximately 18 reviews) and a 4.8/5 score on FeaturedCustomers (based on 584 reference ratings, 23 testimonials, and 17 case studies). The company reports 100+ enterprise customers, including Notion, Duolingo, Chime, Hertz, ClassPass, Rippling, Substack, Eventbrite, Bilt, Block, and Mercado Libre. Common praise centers on fast implementation, high deflection rates, and responsive partnership, while recurring complaints focus on opaque pricing, heavy engineering requirements for setup, limited visibility into agent decisions, and performance degradation under high ticket volumes.

Review Ratings and Volume

Decagon earns high marks across review platforms, though review volume remains modest given the company's enterprise-only sales model.

• G2: 4.9/5 from approximately 18 reviews. Reviewers repeatedly highlight fast implementation, responsive support, and best-in-class AI quality. One reviewer noted "Implementation was very quick (<1 week) and the team is constantly supportive." Another praised the depth of analytics: "the AI solution Decagon provides is best-in-class" for boiling down qualitative data.

• FeaturedCustomers: 4.8/5 based on 584 reference ratings, aggregating 23 testimonials, 17 case studies, and 4 customer videos.

• G2 is not directly accessible for full review scraping, but third-party teardowns consistently cite the 4.9/5 figure and the review themes above.

Common Praise Themes

Fast implementation and time-to-value. Duolingo's DET team went live in one month after a year of failed attempts with a previous vendor, achieving 80% chat deflection immediately. G2 reviewers corroborate quick onboarding, with one reporting under one week from kickoff.

High deflection and resolution rates. Decagon claims an 80% average deflection rate across its customer base. Published case study results include:

• Duolingo: 80% chat deflection at launch

• Flashfood: 90%+ automatic resolution

• ClassPass: 10x increase in chat deflection rate

• Notion: up to 34% improvement in ticket resolution time, with a 3.4% average ask-for-human rate

• 1-800-Flowers: 93% CSAT

• Chime: 70% combined chat and voice resolution

These are vendor-reported figures from self-selected best cases, not independently audited benchmarks.

Ease of use and reduced maintenance burden. Duolingo's Ian Riggins (Senior Operations Manager) stated: "With the previous vendor, at least half my week was dedicated to maintaining their system. With Decagon, it's been a night-and-day difference." Decagon's automatic FAQ syncing and intuitive backend tools were frequently cited as improvements over prior solutions.

Strong vendor partnership. Multiple testimonials reference Decagon's team responsiveness and willingness to incorporate customer feedback. Notion selected Decagon after a thorough RFP process specifically citing the desire for "a collaborative team that would value their feedback and build alongside them."

Common Complaint Themes

Opaque, enterprise-only pricing. Decagon does not publish pricing and offers no self-serve option. Third-party estimates place annual costs starting around $50,000 to $95,000 depending on usage model. Reviewers and analysts consistently flag this as a barrier for mid-market or smaller teams.

Engineering resources required for setup. While writing Agent Operating Procedures (AOPs) is accessible to non-technical staff, connecting those AOPs to backend systems (refund processors, account management APIs, CRM records) requires developers. Implementation timelines range from 4 to 12 weeks. Teams without dedicated engineering bandwidth cite this as a significant barrier.

Limited visibility into agent decisions. A recurring concern among customers is difficulty auditing why the AI agent made a particular decision. Decagon introduced Trace View and Watchtower to address this, but community feedback indicates the auditing experience remains inconsistent in practice, particularly for teams requiring compliance-grade audit trails.

Performance degradation under load. G2 reviewers and independent user reports describe slower response times, higher escalation rates, and occasional agent errors during spikes in ticket volume.

No self-serve or trial access. There is no public documentation site, no free trial, and no way to test the product without engaging sales. This contrasts with competitors that offer self-serve onboarding.

Adoption Metrics

• 100+ enterprise customers reported by the company, spanning fintech, retail, travel, telecom, and software.

• Named customers include Notion, Duolingo, Rippling, Chime, Hertz, Eventbrite, Substack, ClassPass, Bilt, Block, Oura Health, Affirm, Mercado Libre, 1-800-Flowers, and Hunter Douglas.

• The company has 300+ employees across San Francisco, New York, London, and Australia, and was named to the 2025 Forbes AI 50 list.

• Decagon raised a $250M Series D in January 2026 at a $4.5B valuation, reflecting investor confidence in adoption trajectory.

Notable Customer Case Studies

Duolingo (DET). Replaced a vendor that deflected ~30% of email tickets and could not launch chat after a year. Decagon went live in one month, achieving 80% chat deflection. Knowledge management improved with automatic hourly FAQ syncs.

Notion. After a formal RFP evaluating build-vs-buy and multiple agentic solutions, Notion chose Decagon. Results included up to 34% improvement in ticket resolution time, a 3.4% average ask-for-human rate, and a strategic shift elevating CX from a transactional function to a growth driver.

ClassPass. Achieved a 10x increase in chat deflection and expanded customer chat to 24/7 coverage.

Hunter Douglas Group. Deployed localized AI agents with native accents and culturally appropriate tone per market (e.g., "Archie" and "Roman" in the UK, "Buddy" in Australia). The company reports $1M in revenue from fully AI-handled conversations.

Podium. Previously capped at a 25-30% deflection rate with a prior vendor. Partnered with Decagon to push deflection higher and free up human agents for more complex interactions.

Barriers to Adoption

The most frequently cited barriers are cost (six-figure contracts with no transparent pricing), engineering requirements for integration, lack of self-serve access, and the enterprise-only go-to-market model. Decagon's platform is optimized for high-volume support operations with dedicated technical staff. Smaller teams or those on helpdesks like Freshdesk (not currently supported) may face additional friction. Despite these barriers, sentiment among enterprise buyers is strongly positive, with the G2 score and customer testimonials reflecting high satisfaction for the target customer profile.

Monetization & Business Model

Decagon operates a usage-based enterprise SaaS model with two pricing options: per-conversation (a fixed rate for every incoming conversation, with volume discounts) and per-resolution (a higher fixed rate charged only when the AI fully resolves an issue without human escalation). The company does not publish list prices and sells exclusively through a sales-led, white-glove engagement with no self-serve option or free tier.

Revenue Model

Decagon's pricing departs from traditional per-seat SaaS. Because AI agents perform entire workflows autonomously rather than serving as tools for human users, Decagon benchmarks value against human labor rather than seat counts. The company offers two pricing structures, both usage-based:

• Per-conversation pricing: A fixed rate for every incoming conversation the AI agent handles, with volume discounts at higher volumes. The majority of Decagon's customers choose this model because it is predictable and avoids disputes over what constitutes a "resolution."

• Per-resolution pricing: A higher fixed rate charged only for conversations the AI fully resolves without escalation to a human. No charge applies to escalated cases. Larger resolution commitments lower the per-resolution rate.

Decagon does not charge per-seat fees. There is no public pricing page, no self-serve signup, and no free tier or trial. All contracts are negotiated through enterprise sales, with deployment supported by dedicated Agent Product Managers and Forward-Deployed Engineers over an approximately six-week onboarding period.

Pricing Estimates

Because Decagon does not disclose pricing publicly, contract values must be inferred from third-party analyses and customer reports. A 2026 guide by MyAskAI estimated a median annual contract value of approximately $386,000, with a range from $95,000 to over $590,000. The same source estimated a per-resolution rate of roughly $1.50. Fin AI reported a per-resolution rate of $0.50, though noted this appeared to be a negotiated enterprise rate. These figures are unofficial and should be treated as estimates.

Revenue and Growth Metrics

Decagon disclosed reaching 8-figure ARR (i.e., at least $10 million) by mid-2025, one year after emerging from stealth. Sacra estimated Decagon's annualized revenue at approximately $35 million as of October 2025, up from $10 million at the end of 2024, with Q3 2025 GAAP revenue and ARR each growing more than 3x year over year. Decagon has not publicly confirmed these revenue figures.

By January 2026, Decagon reported adding more than 100 new enterprise customers in the prior year, including Avis Budget Group, Block, Deutsche Telekom, Mercado Libre, Oura Health, Affirm, and Chime. The company employs over 300 people across offices in San Francisco, New York, and London.

Funding History

Decagon has raised approximately $481 million to $536 million in total funding across multiple rounds, depending on the source and whether debt or secondary transactions are included. The company's trajectory from stealth to a $4.5 billion valuation took roughly 19 months:

The Series C brought total funding to $231 million at the time, drawing 5x more investor interest than the round size. The Series D tripled the valuation from $1.5 billion to $4.5 billion in under six months. In March 2026, Decagon completed its first employee tender offer at the $4.5 billion valuation, allowing over 300 employees to sell a portion of vested shares.

Key investors across rounds include Andreessen Horowitz, Accel, Bain Capital Ventures, BOND, Coatue Management, Index Ventures, Elad Gil, Forerunner, Ribbit Capital, ChemistryVC, Definition Capital, Starwood Capital, and T.Capital (the investment arm of Deutsche Telekom).

Market Opportunity

Decagon targets the AI-powered customer service automation market, serving F100 enterprises across airlines, banks, telecom, and retail as well as high-growth technology companies. The company cites Qualtrics XM Institute data quantifying $3.7 trillion in annual revenue at risk from bad customer experiences. Decagon positions itself against legacy configuration-driven chatbot and CRM systems, differentiating through its Agent Operating Procedures (AOPs) that enable non-technical teams to define and iterate on AI agent behavior using natural language.

Growth Stage

Decagon is a late-stage, pre-IPO startup. Founded in August 2023 by Jesse Zhang and Ashwin Sreenivas, the company emerged from stealth in June 2024 and reached unicorn status within a year. Its rapid valuation ascent from approximately $650 million (October 2024) to $4.5 billion (January 2026) reflects sustained investor demand for enterprise AI customer support platforms. Competitors include Sierra Technologies, Forethought, and established enterprise software providers such as Salesforce.

Leadership & Team

Decagon was co-founded in 2023 by Jesse Zhang (CEO) and Ashwin Sreenivas (President), both repeat founders with prior acquisitions by major tech companies. The San Francisco-based company has grown to roughly 200+ employees and is backed by leading venture firms including Andreessen Horowitz, Accel, Bain Capital Ventures, Coatue, and Index Ventures.

Founders & Executive Leadership

Jesse Zhang and Ashwin Sreenivas met at a Utah retreat hosted by Andreessen Horowitz (a16z) before founding Decagon. At the time of Decagon's Series A, Sreenivas held the title of CTO; he is currently listed as Co-founder & President on Decagon's official website. Each co-founder holds roughly a 25% stake in the company, per Forbes estimates.

Zhang graduated early from Harvard with a degree in computer science and went through Y Combinator with his first startup, Lowkey. He credits the Niantic acquisition to timing and market fit, and has described building Decagon with a radically customer-driven approach from day one, lining up conversations with large enterprises before committing to a specific product direction.

Sreenivas earned both his bachelor's (2017) and master's (2019) degrees in computer science from Stanford University. Before co-founding Decagon, he built Helia, which applied AI to real-time video and was acquired by Scale AI approximately one year after founding. He also spent about a year as a Deployment Strategist at Palantir Technologies in New York.

Board Members

Ivan Zhou joined Decagon's board at the Series A stage, representing Accel. He has a personal connection to the founding story, having been a colleague of Zhang's at Niantic. Zhou famously promised to shave his head if Decagon increased its revenue tenfold, a milestone the company hit.

Notable Investors & Angels

Decagon's investor base includes prominent venture firms and individual angels from across the enterprise software world:

• Venture firms: Andreessen Horowitz (a16z), Accel, Bain Capital Ventures, Coatue Management, Index Ventures, A*, Avra, Forerunner, Ribbit Capital, T.Capital, ChemistryVC, Definition Capital, Starwood Capital, GIC

• Angel investors (Series A): Aaron Levie (CEO, Box), Howie Liu (CEO, Airtable), Matt MacInnis (COO, Rippling), Mike Vernal (former Sequoia), Frederic Kerrest (Cofounder, Okta), Jack Altman (CEO, Lattice), Ed Hallen (Cofounder, Klaviyo), Elad Gil

Team Size & Growth

Decagon has experienced rapid headcount growth since its 2023 founding. The company grew from a two-person founding team to approximately 200 employees by late 2025, as reported by Forbes. Tracxn lists the employee count at 434 as of May 2026, though third-party headcount trackers vary significantly. The company experienced over 500% headcount growth within a 12-month period during its early scaling phase.

Decagon is headquartered in San Francisco and has been operating with a primarily in-person work culture, with a growing New York office for senior sales personnel.

Company Culture & Values

Decagon's culture is described as intensely customer-centric and fast-moving. Zhang has emphasized several core cultural pillars:

• Customer obsession: The company was built by interviewing large enterprises first and letting use cases emerge from those conversations, rather than starting with a fixed product idea.

• Intensity and work ethic: Zhang has described instituting a culture of intensity, where employees take pride in outworking competitors. He emphasizes "clock speed," the pace at which people can think and adapt, as a key hiring criterion across all functions.

• In-person collaboration: Decagon operates largely in-person at its San Francisco office, with Zhang noting this approach helps culture assimilate during rapid growth.

• AI-native team: The company positions itself as an "AI-native team building AI-native technology," with smart people across engineering, sales, and marketing rather than concentrating talent solely in engineering.

Zhang has stated that Decagon's AI agents are about "enhancing jobs, not replacing them," freeing human workers from mundane, repetitive tasks to focus on higher-value work.

Target Audience & Use Cases

Decagon targets large enterprises and high-growth technology companies that need to automate high-volume customer support across chat, email, voice, and SMS. Its AI agents are designed for organizations with complex support operations that want to resolve customer inquiries autonomously while freeing human agents for higher-value interactions.

Primary Market Segments

Decagon operates exclusively in the enterprise segment. The company does not target SMBs or individual users. Multiple third-party analyses confirm that Decagon's setup is a structured enterprise engagement with dedicated support staff, and its customer base skews heavily toward well-funded technology companies and large enterprises. The platform is not positioned for small teams or self-serve adoption.

Named Customers and Scale

As of 2025-2026, Decagon's AI agents are used by more than 100 companies. Named customers include:

• Consumer technology / SaaS: Notion, Duolingo, Eventbrite, Substack, ClassPass, Curology, Oura Health, Figma, Dropbox, SimplePractice, Whop, Fourthwall

• Financial services / fintech: Bilt, Chime, Affirm, Block, Valon, NG.CASH, Rippling

• Travel & hospitality: Hertz, Away, 1-800-Flowers.com

• Retail & ecommerce: Mercado Libre, Faire, Hunter Douglas Group, Flashfood, Rituals

• Health & wellness: Noom

Forbes lists clients including public companies like Duolingo and Hertz alongside high-growth startups such as Notion, Bilt, and Rippling. The company has also noted strategic partnerships with TaskUs and Deutsche Telekom.

Target Personas

Decagon's primary buyers and users are:

1. Customer Experience (CX) leaders and executives who own support strategy and are accountable for metrics like CSAT, deflection rate, and cost per contact.
2. Support operations managers responsible for day-to-day ticket volume, staffing, and workflow management.
3. Product teams who use conversation analytics and customer insight dashboards to identify friction points and prioritize product improvements.
4. Technical teams that manage integrations with existing support stacks (Salesforce, Zendesk, internal tools) and oversee guardrails, though Decagon's managed-service model means the company's own team handles much of the retrieval logic and ongoing tuning.

Top Use Cases

1. Autonomous ticket resolution. Decagon's AI agents handle inbound customer inquiries across chat, email, and voice, resolving issues end-to-end without human intervention. Customers report deflection rates of 70-93% and cost reductions of 65-95%.

2. Transactional workflow execution. Unlike basic chatbots that only surface information, Decagon's agents execute multi-step workflows such as processing refunds, canceling subscriptions, ordering replacement credit cards, handling dispute workflows, and managing returns. The agents check eligibility, interact with payment and backend systems, and complete the same steps a human agent would.

3. Proactive outbound engagement. Decagon's Proactive Agents can anticipate customer needs and resolve issues before they arise. Hertz, for example, uses this capability to address potential problems before customers need to contact support.

4. Voice-based customer support. Decagon offers AI voice agents for phone-based support, designed for natural dialog with brand customization. Customers like Chime and Valon use Decagon Voice, with Chime reporting 70% chat and voice resolution.

5. Customer insight and product intelligence. Decagon's analytics suite turns every conversation into structured insights, helping CX teams monitor emerging issues, track support volume, and feed learnings back to product and engineering teams. SimplePractice uses Decagon Duet to transform customer support into product intelligence.

Industry Verticals

Decagon has dedicated industry solutions and published content for several verticals:

• Financial services / fintech: Handling secure account issues including password resets, balance inquiries, fraud alerts, and dispute workflows with compliance, validation, and auditability requirements. Named fintech customers include Chime, Bilt, Affirm, Valon, and NG.CASH.

• Ecommerce and retail: Processing returns, managing order modifications, and scaling support during peak periods. Rituals reported zero ticket backlog during peak Black Friday season.

• Travel and hospitality: Hertz uses proactive outbound agents; Away uses Decagon for scaled support.

• SaaS and consumer technology: Notion, Duolingo, Substack, Rippling, and ClassPass use Decagon to handle product questions, subscription management, and high-volume user inquiries.

• Health and wellness: Noom and Curology deploy Decagon for member and patient-facing support.

Company Size Sweet Spot

Decagon's customers are mid-to-large enterprises and well-funded growth-stage companies. The platform is built for organizations with sufficient support volume to justify an enterprise deployment. The Assembled analysis notes that Decagon serves major brands with significant annual recurring revenue, and multiple sources characterize the customer base as large enterprises and high-growth tech companies rather than SMBs. The company's structured onboarding model and managed-service approach reflect an enterprise-class engagement model.

Geographic Focus

Decagon is headquartered in San Francisco with additional offices in New York City and London. While the company is U.S.-based, its customer base is global, including Mercado Libre (Latin America), Rituals (Netherlands/Europe), Hunter Douglas Group (multinational), and NG.CASH (Brazil). The platform supports multilingual customer engagement, enabling deployment across international markets.

Tags & Categories

Decagon is an enterprise-grade AI customer service platform that builds, deploys, and manages autonomous AI agents across chat, voice, and email channels. The company classifies its product category as "AI concierge," positioning itself beyond traditional chatbots toward agents that handle complete customer interactions end to end.

Primary Category: AI Customer Service

Categories: AI Customer Service, AI Infrastructure, AI Other

Tags: conversational ai, customer support automation, ai agents, omnichannel support, voice ai, ticket resolution, enterprise, llm-powered, zendesk integration, salesforce integration, saas, fintech, soc 2 type ii, hipaa, agentic ai, knowledge base sync

Decagon's platform is built on foundation models from OpenAI, Anthropic, and Cohere, layered with company-specific data. Voice capabilities are powered through a partnership with ElevenLabs. The platform's core innovation is Agent Operating Procedures (AOPs), which allow non-technical teams to define agent workflows in natural language while technical teams retain control over guardrails, integrations, and versioning.

The platform integrates with major support tools including Zendesk, Salesforce, Intercom, Kustomer, Confluence, Contentful, Amazon Connect, RingCentral, Stripe, Shopify, and ServiceNow. It also supports MCP (Model Context Protocol) connectivity and custom API integrations for connecting to any system or endpoint.

Decagon targets large enterprises and high-growth companies across multiple verticals. Named customers span financial services (Chime, Affirm, Block, Bilt), technology (Notion, Rippling, Eventbrite, Figma, Substack), travel (Hertz, Avis Budget Group), health and wellness (Oura, Noom, ClassPass), retail (Mercado Libre, Quince), and media (Riot Games, Duolingo). The company also announced a commercial pilot with Deutsche Telekom, expanding into telecom.

Security certifications include SOC 2 Type II, HIPAA eligibility, GDPR compliance, and PCI compliance, with AES-256 encryption at rest and TLS 1.2+ in transit.

Impact & Recommendations

Decagon has emerged as one of the fastest-scaling startups in AI customer support, raising over $500 million in total funding and reaching a $4.5 billion valuation by January 2026, roughly two and a half years after its 2023 founding. Its trajectory places it among the top three enterprise AI agent vendors alongside Sierra and Intercom Fin, with a differentiated position built on action-oriented AI agents, proprietary fine-tuned models, and a roster of F100 enterprise customers.

Market Position and Competitive Landscape

Decagon operates in the rapidly growing AI-powered customer service automation market. Sacra estimates the company hit $35 million in annualized revenue by October 2025, up from $10 million at the end of 2024, with Q3 2025 ARR growing more than 3x year over year. Forbes reported an estimated $12 million in 2025 revenue, illustrating some variance between sources, but both agree the company is scaling rapidly from a small base relative to its $4.5B valuation.

The competitive landscape divides into three tiers:

1. AI-native resolution agents: Sierra, Decagon, Ada, Forethought, and Intercom Fin. These platforms go beyond deflection to take end-to-end actions (issuing refunds, updating accounts, modifying subscriptions). Sierra is the most direct competitor, reportedly valued at $10B with an estimated $150M ARR, positioning it ahead of Decagon on revenue scale. Intercom Fin is strongest for SMB-to-mid-market teams already in the Intercom ecosystem and offers faster deployment at approximately $0.99 per resolution.

2. Legacy incumbent AI layers: Salesforce, Zendesk, and ServiceNow have embedded AI assistants into their existing helpdesk platforms. These incumbents exceed Decagon on installed base and revenue but lack the AI-native architecture and action-taking depth that Decagon and Sierra provide.

3. Specialized and vertical tools: Gorgias (e-commerce/Shopify), Cognigy (hybrid rule-based plus AI with on-premise options), and voice-first platforms like Ringg AI compete in specific segments.

In head-to-head resolution testing cited by Fin (Intercom's product), Fin reportedly achieved 73% resolution versus Decagon at 49%, though these results are published by a competitor and should be treated cautiously. Decagon's own customer data shows 80%+ deflection rates across deployments, with specific customers like Chime reaching 70% voice resolution and ClassPass achieving a 10x increase in deflection rate.

Key Growth Signals

• Funding velocity: Decagon raised four rounds in under two years: a $35M Series A (July 2024), a $65M Series B at a $650M valuation (October 2024), a $131M Series C at a $1.5B valuation (June 2025), and a $250M Series D at a $4.5B valuation (January 2026). The valuation tripled in six months between Series C and Series D.

• Enterprise customer acquisition: More than 100 new global enterprise customers signed in 2025, including Avis Budget Group, Block, Deutsche Telekom, and Mercado Libre. The customer roster also includes Notion, Duolingo, Rippling, Chime, Oura Health, Affirm, Hertz, Bilt, Eventbrite, Substack, Wealthsimple, and Hunter Douglas.

• Investor quality: The cap table includes Andreessen Horowitz, Accel, Bain Capital Ventures, Coatue Management, Index Ventures, Ribbit Capital, Forerunner, and Elad Gil. Andreessen Horowitz published a lengthy thesis piece in March 2026 naming Decagon as the portfolio company at the heart of the AI customer service shift.

• Product expansion: Decagon extended from chat and email into enterprise voice (Decagon Voice, built with ElevenLabs; Voice 2.0 launched September 2025 with 65% latency reduction), proactive outbound agents, and an AOP Copilot for agent management. Over 80% of model traffic runs on Decagon's self-trained models built for CX accuracy.

• Geographic expansion: Offices opened in New York (July 2025) and London (November 2025), extending beyond San Francisco to cover the U.S. East Coast and Europe. A strategic partnership with TaskUs extends go-to-market reach across digital and voice channels.

Risk Factors

• Valuation vs. revenue gap: At a $4.5B valuation against estimated revenue of $12M to $35M, Decagon trades at a revenue multiple well above 100x. Any slowdown in enterprise AI spending or failure to scale ARR proportionally could pressure the valuation significantly.

• Intense competition: Sierra is reportedly larger by revenue and valuation. Legacy incumbents (Salesforce, Zendesk) are rapidly improving their embedded AI capabilities. Intercom Fin competes aggressively on price and deployment speed for mid-market buyers. The category is attracting continued venture funding, which will produce more entrants.

• Enterprise sales cycle dependency: Decagon has no self-serve product for SMB buyers. Revenue depends on high-touch enterprise sales cycles, which are longer and more concentrated. The loss of a few large customers could materially impact growth.

• Model dependency and cost: While Decagon uses proprietary fine-tuned models for over 80% of traffic, it still relies on third-party models from OpenAI, Anthropic, and Cohere for the remainder. Pricing models (per-conversation and per-resolution) must manage inference costs as volume scales.

• Young company, young founders: Co-founders Jesse Zhang (28) and Ashwin Sreenivas were honored on the 2026 Forbes Under 30 list. While their prior exits (Zhang's gaming startup Lowkey was acquired by Niantic) demonstrate execution ability, scaling a multi-billion-dollar enterprise software company presents a different order of operational complexity.

Competitive Advantages (Moat)

• Agent Operating Procedures (AOPs): Decagon's signature feature lets CX teams write workflows in natural language that compile directly into code, enabling non-engineers closest to the customer to shape agent behavior while engineers retain control over integrations, guardrails, and versioning. This is a meaningful workflow moat that competitors' configuration-driven approaches do not match.

• Proprietary models: Over 80% of model traffic runs on self-trained models purpose-built for CX accuracy rather than adapted from general-purpose LLMs. This vertical model specialization compounds in edge cases and is difficult for competitors using off-the-shelf foundation models to replicate.

• Action-taking depth: Unlike chatbots that deflect or agent-assist tools that draft replies, Decagon agents execute end-to-end actions through APIs: processing refunds, updating accounts, canceling subscriptions, disputing transactions, and replacing credit cards. This action capability is the core differentiator versus first-wave AI assistants.

• Enterprise customer lock-in: Once Decagon's agents are integrated into a company's backend systems, knowledge bases, and workflows, switching costs are substantial. The AOP framework deepens this lock-in by embedding Decagon into the customer's operational processes.

• Talent and capital advantage: With over 300 employees and $500M+ in funding, Decagon has the resources to invest in proprietary model training, enterprise-grade infrastructure, and geographic expansion at a pace smaller competitors cannot match.

Comparison to Closest Competitors

ICP Fit for Chiri Atlas Audience

Decagon is an excellent fit for Chiri Atlas readers evaluating enterprise AI tools for customer support transformation. The ideal customer profile is:

• Company size: Mid-market to enterprise (500+ employees) with high-volume support operations

• Industries: Fintech, telecom, travel, e-commerce, SaaS, and healthcare (regulated environments requiring SOC 2, GDPR, CCPA, HIPAA compliance)

• Support volume: Organizations where support is operationally painful at scale and AI resolution can deliver measurable ROI

• Technical readiness: Teams with engineering resources to support integration, though AOPs reduce the engineering burden for ongoing agent management

• Not a fit for: SMBs or teams wanting a self-serve, quick-deploy chatbot widget. Decagon requires enterprise-level commitment and engagement.

Overall Recommendation and Outlook

Decagon is a category leader in the AI-native customer support agent market with genuine product differentiation (AOPs, proprietary models, end-to-end action-taking), a blue-chip enterprise customer base, and top-tier investor backing. Its growth signals (100+ enterprise customers in 2025, 3x ARR growth, tripled valuation) are among the strongest in the AI agent category.

The primary risk is the valuation-to-revenue gap, which demands sustained hypergrowth to justify. Competition from Sierra (larger by revenue) and improving legacy incumbents will intensify. The company's ability to expand from reactive support into proactive concierge experiences and revenue-generating interactions (as demonstrated by Hunter Douglas generating $1M in AI-attributed revenue) will determine whether it grows into its valuation.

For enterprises evaluating AI customer support platforms, Decagon warrants a shortlist position alongside Sierra and Intercom Fin. The decision between them should come down to configurability needs (Decagon), managed-service preference (Sierra), or ecosystem and speed-to-deploy (Intercom Fin). Decagon is best suited for organizations that need deep workflow customization, action-taking agents, and have the scale and technical readiness to justify an enterprise investment.

Chiri Analysis

Chiri Score: 86/100

Verdict

Best for:

• High-volume enterprise support operations with dedicated engineering resources for backend integrations

• Fintech, retail, travel, telecom, and software companies needing autonomous resolution across chat, email, and voice

• Teams replacing underperforming chatbot vendors capped at 25-40% deflection

• Regulated industries requiring SOC 2 Type II, HIPAA/BAA, and GDPR compliance

• CX organizations wanting non-technical staff to define agent logic in natural language via AOPs

Not for:

• Mid-market or small teams sensitive to six-figure, sales-only pricing

• Buyers wanting a free trial or self-serve onboarding without engaging sales

• Teams on unsupported helpdesks like Freshdesk

• Organizations without engineering bandwidth for 4-12 week integrations

• Compliance teams requiring published ISO 27001 or PCI-DSS certifications

Head-to-head

Frequently Asked Questions

Is Decagon SOC 2 compliant?

Yes. Decagon holds SOC 2 Type II certification, confirmed on its security page and by independent sources. It also offers HIPAA compliance with Business Associate Agreements for healthcare enterprise customers and is GDPR compliant. It does not publicly list ISO 27001 or PCI-DSS.

How much does Decagon cost?

Decagon does not publish pricing and sells through a sales-led enterprise process requiring a demo. It uses usage-based models: per-conversation or per-resolution. Third-party estimates place annual contracts starting around $50,000 to $95,000 depending on volume. There is no free trial or self-serve tier.

Who are Decagon's competitors?

Decagon competes with legacy support platforms Zendesk and Salesforce (Agentforce), agentic AI vendor Intercom (Fin), and direct startup peer Sierra. Its differentiator is autonomous end-to-end resolution across chat, email, and voice via natural-language Agent Operating Procedures rather than ticket deflection or routing.

Is Decagon good for enterprise?

Yes. Decagon is enterprise-only, serving 100+ customers including Notion, Duolingo, Chime, Deutsche Telekom, and Avis Budget Group. It provides SSO, SCIM, RBAC, audit logs, multi-region infrastructure, and dedicated implementation engineers. It is optimized for high-volume support teams with engineering resources.

What is Decagon's Agent Operating Procedures feature?

Agent Operating Procedures (AOPs) are Decagon's core differentiator. They let non-technical CX teams define agent workflows in plain English while engineers control integrations and guardrails. AOPs trigger real backend operations including refunds, account changes, and third-party queries, eliminating decision trees and custom code.

How much funding has Decagon raised?

Decagon has raised approximately $481 million across six rounds. It closed a $250 million Series D in January 2026 led by Coatue and Index Ventures at a $4.5 billion valuation, tripling from its $1.5 billion Series C valuation in June 2025.

Does Decagon support voice and phone support?

Yes. Decagon Voice launched in 2025 in partnership with ElevenLabs for voice synthesis, handling inbound calls with interruption management and human escalation. The Spring 2026 update added outbound voice campaigns and persistent user memory across interactions.

How long does Decagon take to implement?

Implementation typically ranges from four to twelve weeks depending on integration complexity, with Decagon assigning dedicated implementation engineers. Some customers report faster launches; Duolingo went live in one month and one G2 reviewer reported under one week. Writing AOPs is non-technical, but backend integration requires developers.

---

Reviewed by Chiri Atlas Research Desk (AI Tooling Analyst) on 2026-07-05.