New offerings target voice interactions, personal AI agents, industry-specific workflows and enterprise knowledge for customer experience

Conversational AI company Decagon has announced four new products and expanded partnerships at its second annual Decagon Dialogues conference, as the company looks to extend AI agents beyond traditional customer support into broader customer journeys.

The announcements include Voice 3, Personal Agent Gateway, Agent Modules and Duet Apprentice, with each product addressing a different aspect of deploying AI agents within customer-facing operations.

Decagon said the products are designed to address a gap between general-purpose frontier AI models and the business-specific knowledge, policies, systems and workflows required to deliver customer experiences.

Voice 3 Brings Parallel Conversation and Task Execution

Voice 3 combines a new duplex architecture with Chord, a voice model developed by Decagon Labs.

The company said the architecture allows an AI agent to handle conversation and task execution simultaneously. Rather than placing customers on hold while a task is completed, the agent can explain its progress, respond to follow-up questions and continue working in the background.

Chord is designed around the characteristics of live conversations, including natural pauses and changes in speaking pace. Decagon said the system can slow down when communicating information such as confirmation codes and return to a more conversational pace during other parts of a call.

Personal Agent Gateway Connects With AI Assistants

Personal Agent Gateway is designed to identify when an AI personal agent, rather than a human customer, is interacting with a business.

As personal AI assistants increasingly perform tasks such as booking, purchasing, cancellation and negotiation, Decagon said businesses will need ways to authenticate and manage these interactions.

The gateway can use business and platform signals to identify personal agents and route them through a dedicated channel. Decagon said businesses can define what actions an agent is permitted to perform, while sensitive transactions can require approval from the human customer behind the personal agent.

Agent Modules Extend AI Across Industry Workflows

Agent Modules are designed to extend customer-facing AI agents into workflows beyond traditional support.

The modules target industries including financial services, travel and hospitality, healthcare, retail, technology and telecommunications. They package the orchestration and requirements needed for specific customer journeys rather than requiring businesses to build each workflow from the ground up.

Decagon said modules also include shared management capabilities, allowing different teams to manage portions of an agent with separate access controls while administrators maintain visibility across the broader customer relationship.

Customer memory can also carry between conversations and channels, allowing information from previous interactions to inform subsequent journeys.

Duet Apprentice Helps Agents Learn From Employees

The fourth launch, Duet Apprentice, is designed to help AI agents learn from existing company knowledge and experienced employees.

The system can analyze internal resources and escalated conversations handled by experienced representatives. That information is used to develop Agent Operating Procedures (AOPs), run simulations and analyze customer conversations.

Duet can also monitor discussions in Slack and Microsoft Teams to identify changes in policies and business plans. Teams can tag Duet in conversations to request updates to procedures, while the system can ask for clarification when information is ambiguous.

Decagon said knowledge learned through Apprentice remains within the customer's workspace and is not used to train models outside that environment.

Customer Experience Leaders Join Decagon Dialogues

The conference also featured executives from retail, healthcare, travel, hospitality, insurance, finance and technology organizations using Decagon.

Speakers included Eric Phillips, Senior Vice President and Chief Digital Officer at Delta Air Lines, who joined the opening keynote to discuss AI and airline customer experience.

Representatives from American Airlines, including Vice President of Digital Customer Experience Samuel Liyanage and Vice President of Commercial Technology Misty Shafer Sterne, discussed AI-powered customer journeys.

Other speakers included Adam Terk, Vice President of Operations at Carrot, and Diana Rainesh, Senior Vice President of Strategy & Analytics Support & Operations at Ticketmaster, who discussed applications of AI across healthcare and product discovery.

Databricks Partnership Connects Enterprise Data to AI Agents

Decagon also announced a partnership with Databricks focused on bringing governed enterprise data into customer conversations.

Customer information such as order status, entitlements, product usage and payment history can reside within the Databricks Data + AI Platform. Under the partnership, Decagon agents can access that information directly rather than requiring businesses to create separate copies through traditional ETL pipelines.

Conversation data can also flow back into Databricks, adding information such as customer intent, sentiment and root-cause insights.

Decagon has joined Databricks' Built on program and plans to route model calls through Unity Gateway, extending governance, lineage and audit capabilities to agent activity. Decagon is also a launch partner on Databricks Marketplace.

Plaid Partnership Targets Financial Services Workflows

Decagon is also partnering with Plaid to allow AI agents to handle financial tasks that require access to customer accounts.

Many financial service interactions currently require customers to leave a support conversation and authenticate through their bank before returning to the original workflow. The partnership is designed to reduce that friction by allowing Decagon agents to work with permissioned financial account information within the conversation.

Potential use cases include completing claims, recovering failed payments, managing applications and handling disputes while maintaining context from previous interactions.

Plaid provides consumer-permissioned account access, while Decagon's existing agent guardrails remain in place for financial services customers.

Together, the product launches and partnerships reflect Decagon's broader strategy of positioning AI agents as a unified customer-facing layer that can combine conversation, business data, specialized workflows and operational actions across the customer journey.