Decagon launches missions for long-lived AI customer workflows – Unite.AI

Decagon introduced Missions on November 12, 2026, a product designed to enable AI agents to manage customer outcomes across multiple conversations, systems and participants. Agents can pursue a defined goal for days, weeks, or months instead of finishing their work with a single interaction.

Decagon cites prior authorization for an MRI, small loan approval, and lead qualification as examples of multi-step journeys that can generate significant revenue. For customer experience teams, the intended change is to replace manual status tracking and handoff coordination with continuous journey refinement.

Orchestration between conversations and back-office systems

A team begins a mission by stating the objective in plain language, naming the characters the agent should engage, and adding directions. The agent then chooses subsequent actions around that goal. It can connect threads involving different people, contact each participant through an appropriate channel, and wait until a third party needs to complete a step before the process can continue.

The product goes beyond conversations. Through browser actions, an agent can perform back-office tasks such as completing forms. Decagon’s prescription refill example begins with a patient notification, continues with provider awareness when authorization is required, and includes re-running claims on a payer portal. The process ends when the agent confirms the pickup details with the patient.

This example illustrates the broader business role of the product: a single service outcome may require customer communication, provider coordination, and payer system activities. Missions assign these tasks to a goal-directed workflow instead of treating each interaction as a separate job.

Chance and memory shape subsequent actions

Each mission case keeps a record of progress, including completed steps and new information that could alter the plan. The agent uses the case to select the next action, organize proactive contact, and preserve context for subsequent work. Before taking action, double-check both the case and the client’s memory to determine whether the existing plan is still a good fit.

Memory provides preferences and relevant context from other conversations. Decagon examples include honoring a customer’s request to not receive calls after 3pm and accounting for a parallel interaction where the same customer is onboarding a new product while resolving an unrelated issue.

The same double-check applies when a customer changes the expected sequence. If an agent has qualified a lead and arranged a follow-up, but the customer self-schedules that follow-up via email, the agent changes its course. The design therefore separates two types of context: the Mission case tracks progress towards the business outcome, while Memory keeps individual preferences and simultaneous interactions visible.

Optimization solves long workflow attribution

Decagon identifies attribution as a central difficulty when automated work extends across many interactions. If a prescription refill is delayed, teams may need to determine whether the problem originated in the initial contact, provider transfer, or insurance verification. This makes it difficult to isolate the interaction or transfer that needs to be changed.

Each mission is constructed as an Agent Operating Procedure, or AOP. Teams can write, modify, and refine that logic, inspect agent decisions, and test individual parts of the configuration to identify what needs improvement. Duet and Duet Autopilot provide targeted recommendations across the entire customer journey, extending optimization beyond isolated interactions.

Launch consequently links execution and diagnosis into a single product model: Agents advance an objective across participants and systems, while case-, memory-, and AOP-based testing provides teams with the tools to review and revise the path taken. For companies considering automation long-term, the open question is whether such controls can reliably highlight weak steps and improve outcomes as the number of interactions grows.