Decagon Review Australia 2026: The Enterprise AI Concierge with the Metrics to Back It Up

Verdict (TL;DR): Decagon is a US-based enterprise AI customer-service platform that deploys AI agents across voice, chat, and email — resolving customer issues end to end without a human in the loop. The customer results are striking: 80% deflection at Duolingo, 95% cost reduction at ClassPass, 70% chat and voice resolution at Chime. Backed by a16z, Accel, and Bain Capital, this is serious infrastructure for serious contact-centre operators. There is no public pricing and no self-serve entry point. It is sold through a sales process to large enterprises. For Australian SMBs: this is not your tool. For large Australian enterprises running high-volume contact centres — financial services, retail, telecommunications, health — it is worth a close evaluation. Score: 14/25, weighted down by inaccessible pricing, no AUD native model, and no Australian data residency option confirmed publicly.

Score: 14/25 — ⭐⭐⭐ Good

Last reviewed: 26 June 2026 by Adam.

Not for small business. Decagon is enterprise software: sales-led, no public pricing, US-based. If you are a small or medium Australian business looking for an AI customer-service chatbot you can try today, go to our best AI customer-service chatbots for Australia 2026 roundup. We review Intercom Fin, Freshdesk Freddy AI, HubSpot Breeze, and others there with actual AUD prices and free tiers you can test.

What Decagon is

Decagon is a San Francisco–based company founded by Jesse Zhang (CEO) and Ashwin Sreenivas (President). It launched in 2023 and has built a platform specifically for deploying AI agents that handle customer service conversations — not just answering FAQs, but taking actions: processing refunds, rescheduling bookings, updating account details, resolving billing disputes.

The company is backed by a16z, Accel, Bain Capital Ventures, Coatue, and Index Ventures, among others. (exact total funding amount — Decagon has not publicly disclosed figures as of June 2026.)

The core idea is that most customer service software was built for human agents and then bolted with AI. Decagon was built the other way around: the AI agent is the primary operator, with humans available for escalation. The platform uses what it calls Agent Operating Procedures (AOPs) — natural-language workflow definitions that let CX teams configure agent behaviour without writing code or filing engineering tickets.

Customers on the public record include Duolingo, Chime, ClassPass, Rippling, Oura Health, Hertz, Noom, and Affirm. These are not small companies.

In June 2026, Decagon launched Duet Autopilot, described as a self-improving agent that can autonomously identify gaps in its own knowledge and performance and propose corrections without human intervention.

What it costs in Australia

Decagon publishes no pricing. None. The website has a "Get a demo" button and nothing else on cost. Based on the company's positioning, customer profile (Duolingo, Hertz, Rippling), and the US enterprise AI market, pricing is almost certainly structured as:

• Annual contract, billed in USD
• Likely priced per resolved conversation or per seat — (USD $200,000–$500,000+ per year for mid-tier enterprise deployments, based on comparable enterprise AI customer-service platforms; this is an estimate only, not a Decagon-confirmed figure)
• Converted to AUD at 1.4045 (ECB reference rate, 18 September 2026): (A$281,000–A$702,000+ per year before GST)

Australian companies would also need to account for GST implications on imported software services. There is no free trial, no self-serve onboarding, and no monthly plan.

How it works in practice

Decagon deploys across three channels: voice (phone-based AI agents), chat (website and messaging), and email (automated resolution of inbound support emails). All three share a single intelligence layer, so a customer who starts on chat and calls back gets a consistent experience.

Configuration uses AOPs — written in plain English, not code. A CX manager can write something like "If a customer requests a refund and the order is within 30 days, process it; if outside 30 days, escalate to a human with a reason summary" and the agent will follow it. Decagon claims this lets teams iterate on agent behaviour in hours, not weeks.

The platform connects to existing tech stacks through integrations: Salesforce, Zendesk, Shopify, Stripe, Intercom, and others. It supports SSO, RBAC, and integrates with Okta and Microsoft Entra for identity management.

On the AI safety side, Decagon enforces zero-day retention with AI providers (OpenAI, Anthropic) — meaning conversation data is not stored or used for model training by those providers. A built-in supervisor model checks responses before they are sent and flags hallucinations. PII is automatically redacted using Google's DLP service at the end of each conversation. The platform carries SOC 2 Type 2 certification and maintains uptime SLAs, though specific figures are not publicly disclosed. (exact SLA uptime percentage — not published on Decagon's public website as of June 2026.)

What it does well

Verified results from large deployments. Unlike many AI customer-service vendors making claims without evidence, Decagon publishes detailed case studies with named clients and specific metrics. Duolingo achieved 80% deflection. ClassPass reduced costs by 95%. Chime resolved 70% of chat and voice contacts without a human. These are primary-source claims from the vendor's website; we have not independently verified them, but they are named, attributed, and not anonymous.

The AOP approach is genuinely different. Most contact-centre AI still relies on decision trees or intent classification — fragile systems that break when customers phrase things unexpectedly. AOPs written in natural language give CX teams more control and faster iteration than traditional bot-building tools. The Rippling case study notes that Rippling can tailor agent behaviour to each of its product lines without engineering involvement.

Omnichannel with memory. Voice, chat, and email share a context layer. Chime's COO specifically called out "cross-channel memory" as a key differentiator — the AI agent knows what happened in a previous chat when a member calls.

Watchtower — always-on QA. Every conversation is reviewed against custom criteria automatically. This matters for regulated industries: you can set compliance rules and have the system flag deviations in real time.

Duet Autopilot. The June 2026 launch of Autopilot, a capability that identifies gaps and proposes knowledge updates autonomously, is a meaningful step. For enterprise contact centres handling thousands of SKUs or policy changes, continuous self-improvement without manual maintenance is a real operational advantage.

What it doesn't do well

No pricing transparency. For any business trying to plan a technology budget, the complete absence of public pricing is a friction point. You must engage a sales rep before you know whether you are in the right range. Some large enterprises accept this; many prefer to do pre-qualification with numbers.

Not designed for small business — at all. There is no pathway for a 10-person team to use Decagon. No starter plan, no self-serve trial, no month-to-month contract. The platform assumes you have a CX operations team, a tech stack to integrate with, and a significant support volume to justify the investment.

US-based, US-data infrastructure. Decagon's security page references multi-region infrastructure but does not name specific regions. It does not mention Australian or Asia-Pacific data residency on its public website. For Australian enterprises in regulated sectors — financial services, healthcare, government — this is a real issue. Data sovereignty requirements under Australian law may prevent use of a platform that processes customer data in the US. (specific AWS/Azure region configurations for Australian customers — requires direct vendor confirmation.)

No Australian presence confirmed. There is no Australian office, no Australian sales team listed, and no AU-specific documentation on the website. Enterprise procurement typically requires a local entity or at least a formal APAC presence for contract, support, and compliance purposes.

High implementation complexity. Integrating Decagon into an existing contact-centre tech stack — CRM, CCaaS, ticketing — is a project, not a product activation. The case studies imply strong vendor support during onboarding, but you are still signing up for a significant implementation effort.

Australian fit

Decagon is a strong platform for the right buyer — but it is a poor fit for most Australian organisations reading this publication.

Positive signals: The platform's zero-day data retention with LLM providers and PII redaction are genuine enterprise-grade controls. The integration breadth (Salesforce, Zendesk, Shopify, Stripe, Intercom, and more) covers common Australian tech stacks. The voice channel is particularly relevant for Australian contact centres, where phone remains a dominant support channel.

Problems for Australian buyers:

• No published Australian data residency — likely US-based, which creates issues for regulated sectors
• No Australian entity or APAC office confirmed (as of June 2026)
• Pricing in USD, no AUD support, no GST documentation public
• No AU-specific case studies or references

For an ASX-listed company, a major Australian bank, or a large telco, these are not necessarily deal-breakers — they are negotiation items in a sales process. For a mid-tier Australian business, they are reasons to look elsewhere first.

Australian Fit score: 1/5. No confirmed data residency, no local presence, no AUD pricing, no AU case studies.

Alternatives

For Australian SMBs: Do not spend another minute on Decagon. Our best AI customer-service chatbots for Australia 2026 roundup covers tools with AUD pricing, free trials, and self-serve onboarding. Intercom Fin AI and Freshdesk Freddy AI are reviewed in detail there.

For Australian enterprise: If you are evaluating Decagon, you should also talk to Sierra (outcome-based pricing, similar enterprise positioning), Cresta (stronger contact-centre integration with APAC expansion in 2025, see our review below), and Parloa (strong compliance certifications including DORA, ISO 27001, and PCI DSS; European data-sovereignty model may be negotiable for APAC).

Bottom line

Decagon is building something genuinely impressive. The customer results are the best-documented in the enterprise AI customer-service space. The AOP approach is practical and the omnichannel memory is differentiated. If you are running a large Australian contact centre — retail, fintech, health, telco — and you have the budget, Decagon belongs on your shortlist alongside Sierra and Cresta.

But if you are a small or medium Australian business: this is not for you. Go to our customer-service chatbot roundup. Come back to Decagon when you have a contact centre team and a procurement budget to match.

Score

Frequently asked questions

Is Decagon available in Australia?

Decagon operates globally and Australian enterprises can engage via its sales team, but Decagon does not have an Australian office, does not publish Australian data residency options, and does not list AUD pricing. Large Australian enterprises in financial services or healthcare should verify data sovereignty requirements before proceeding.

How much does Decagon cost?

Decagon does not publish pricing. It is sold through a sales team on annual enterprise contracts, priced in US dollars. Based on the profile of its customers and the enterprise AI market, costs are likely in the six-figure USD range annually. There is no free trial or self-serve plan.

What is the difference between Decagon and a standard AI chatbot?

Standard AI chatbots typically answer questions. Decagon deploys AI agents that take actions — processing refunds, updating bookings, resolving billing issues — across voice, chat, and email, without a human in the loop for the majority of contacts. It uses natural-language Agent Operating Procedures (AOPs) to define agent behaviour rather than decision trees or intent classification.

Sources

• Decagon website — product overview, channels, security: https://decagon.ai (checked 26 June 2026) — Wayback 29 June 2026
• Decagon About page — founders, investors, customer metrics: https://decagon.ai/about (checked 26 June 2026 — contemporaneous snapshot still owed) — Wayback 13 August 2026
• Decagon Security page — certifications, encryption, data practices: https://decagon.ai/security (checked 26 June 2026 — contemporaneous snapshot still owed) — Wayback 21 August 2026
• Decagon Duet Autopilot announcement (blog): https://decagon.ai/blog/autopilot (checked 26 June 2026 — contemporaneous snapshot still owed) — Wayback 14 August 2026
• Notion Developer Platform news item (secondary, AID internal reference) — confirms Decagon named as integration partner in Notion 3.5 External Agents API: https://www.theaidirectory.com.au (internal reference only)
• USD/AUD conversion: USD→AUD 1.4045 (ECB reference rate, 18 September 2026), shown ex-GST per our GST rule. AUD figures are our conversions of an estimated USD range, not prices Decagon quotes — they are estimates twice over.

The AI Directory Australia is an independent editorial publication. We don't accept payment for positive reviews. Read our Methodology and Affiliate Disclosure.