AI Models & Platforms

Decagon Expands AI Concierge Platform With Four New Releases

Add Unite.AI to your preferred sources on Google

Decagon introduced four products at its Decagon Dialogues 2026 event on October 1, 2026: Voice 3 with a new speech model called Chord, Personal Agent Gateway with a companion authorization protocol, Agent Modules, and the Duet Apprentice beta. Co-founder and CEO Jesse Zhang and co-founder and President Ashwin Sreenivas detailed the releases in a company announcement.

Enterprises first deployed Decagon’s agents to resolve support tickets, and the company said those agents now qualify leads, onboard customers, collect payments, and grow relationships. The new releases are organized around what Decagon calls an AI concierge that meets customers however they choose to engage, including when a customer sends a personal AI agent to get something done on their behalf.

Voice 3 and the Chord Speech Model

Voice 3 ships with Chord, the first voice model from Decagon Labs, post-trained specifically for customer conversations. Decagon said the model shapes speech phrase by phrase, slowing down for details like phone numbers and confirmation codes before returning to a conversational pace. Existing brand-customization controls carry over, including voice selection, pronunciation of business-specific terms, and delivery tuning.

Voice 3 also runs on a new duplex architecture with two layers operating in parallel: a low-latency conversational model handles listening and speaking, while a more powerful model manages reasoning, tool calling, and guardrail enforcement behind the conversation. The agent processes incoming audio even while speaking, so it talks through a casual “mhm” but yields on a genuine interruption, and it can narrate progress and answer follow-up questions during a long-running task rather than putting the caller on hold.

Decagon reported that in a blind test across three pairs of recordings, each pairing a real person with the same voice run through Chord, on average approximately 90% of listeners could not identify the human. The company said Chord is trained on licensed data and consented voice talent, never on customer-owned data.

Voice 3 supports more than 70 languages, detects the caller’s language, and switches automatically even when callers move between languages mid-sentence, according to the company, which said every language is validated by native speakers before it ships. Christian Niedworok, Lead of Digital Service Communication at Deutsche Telekom, said in the Voice 3 announcement that “the voice sounds like it’s actually listening, and it keeps the conversation moving instead of going quiet while it works.”

Personal Agent Gateway and the PACT Protocol

In the Personal Agent Gateway announcement, Decagon noted that in the month leading up to the event, Meta launched Muse, OpenAI introduced dots, and Instinct started placing phone calls for users. The company described these as personal agents that book, buy, cancel, and negotiate on their owners’ behalf, and said they are already contacting customer support.

The gateway combines two capabilities. Personal agent detection flags likely personal agents across chat and voice using signals from the business and the platform, with each business deciding what happens next and what those agents can access. A dedicated personal agent channel sits alongside chat, email, and voice with separate Agent Operating Procedures (AOPs), so the same request can follow a different workflow depending on whether a person or an agent is asking.

Permissions live inside the AOPs: businesses define the scopes an agent can request and mark which sections require each one, and those sections are never exposed to an agent lacking the required scope. Decagon described an airline example in which a traveler’s personal agent requests permission to view and rebook flights but the traveler approves only viewing; the airline’s agent shares earlier flight options but cannot make the change until the traveler authorizes rebooking.

Decagon also introduced PACT, short for Personal Agent Consent & Trust, a protocol it said lets a person’s agent act on their behalf under permissions the service defines and the person grants. PACT builds on the Agent2Agent protocol, which covers how agents find each other and exchange messages, and adds delegated authorization built on OAuth 2.0 so an agent can prove which person it represents and what that person allowed it to do. The specification is available, and Decagon said it is working with personal agent providers and enterprise customers to shape it.

Agent Modules and Duet Apprentice

Agent Modules combine infrastructure, analytics, and campaign orchestration for customer journeys beyond support, including lead qualification, onboarding, and collections, across financial services, travel and hospitality, healthcare, retail, and telecom. The customer’s team defines the business logic while Duet draws on the company’s SOPs, policies, integrations, and knowledge sources to translate processes into agent behavior. In collections, for example, teams set disclosure language, approved payment plan tiers, and conditions that require a handoff to a human.

Before launch, Simulations test agent behavior against personas with varying intents, such as a customer disputing a balance; once live, module-specific analytics track journey outcomes such as recovered payments, not just whether a conversation was resolved. Campaign Composer sequences touchpoints across channels and time, from an SMS payment reminder to a follow-up call and, when needed, a handoff to a collections representative, with each step beginning from prior interaction history. A shared management layer supports multiple instances with separate access controls per team while analytics aggregate across instances, and memory carries across sessions, channels, and instances, according to the Agent Modules announcement.

Duet Apprentice is in beta, with general availability planned. Decagon said Duet arrived earlier in 2026 as an agent-building partner and that it writes more than 70% of the AOPs running on the platform. With Apprentice, Duet reads onboarding documents and process wikis in Notion, Confluence, SharePoint, or Google Drive without migration, and learns from escalated conversations handled by experienced representatives. Plugins for Slack and Microsoft Teams let Duet follow discussions as policies change, draft AOP updates when tagged, and ask for clarity in the same channel when a thread is ambiguous. Decagon stated that what Duet learns belongs to the customer, stays in the customer’s workspace, and is not used to train models outside it.

Anne Young, Senior Manager, AI & Process Automation at Questrade, said in the Duet Apprentice announcement, “With Duet, we confidently launch new features on a weekly basis.” Jenn Palk-Cogley, Head of CX and User Operations at Perplexity, said Duet lets her team build and iterate on agent logic without pulling resources away from the roadmap.