DEV Community

News
News

Posted on Edited on

Decagon Missions runs AI customer outcomes across days and weeks

Infographic cover

๐Ÿ“… November 13, 2026 โ€ข Source: Decagon

Decagon introduced Missions on November 12, 2026, per the company's blog post from product manager Gunguk Kim and member of technical staff Jian Yang. Missions are long-horizon agentic tasks: the agent coordinates work across systems, tracks statuses and engages several people until the outcome is reached, over days or weeks rather than a single exchange. Coverage from unite.ai frames the launch the same way, as Decagon expanding its agents into end-to-end customer workflows.

Setup happens in plain language, per the company: a team states the goal in ordinary words, names the personas involved and adds rules for how to proceed. The agent then carries that goal as a north star and selects the next step on its own until the result lands, whether the horizon runs to days, weeks or months. Orchestration is multi-persona: the agent connects threads across participants, chooses a relevant channel for each of them, and any step that depends on a third party puts the mission on pause until that external side completes its part.

Browser Actions push the work into the back office, for example filling out forms, per the company. The worked example is a prescription renewal: the patient receives a notification, a physician is engaged to authorize the renewal, claims are re-run inside the payer's system, and dispensing details are confirmed back to the patient. Other examples named by the company and by unite.ai include prior authorization for an MRI, approval of a small loan, and lead qualification for sales teams.

Every mission is written down as an Agent Operating Procedure, or AOP: the team writes, edits and tests the logic itself, every agent decision stays visible, and parts of the configuration can be tested before anything ships, per the company. Context comes from two layers โ€” the mission case, a dynamic record of progress and new information, and Memory, which holds preferences such as not calling a customer after 3 PM.

Before each action the agent re-checks the case and Memory and adapts when a customer changes the sequence, for example by booking a follow-up over email on their own, per the company. Duet Autopilot, a Decagon product, adds recommendations for improving the whole route rather than individual interactions. The market framing comes from analyst Futurum: in the segment Missions addresses, 56.5% of enterprise decision-makers name support and customer experience as the top GenAI use case, per Futurum.

Key figures

IMPACT: For CX teams the practical shift is from tracking statuses and handoffs toward improving each step of a mission, since the agent carries the thread across systems and people over days. Visibility into the AOP doubles as an audit trail for agent decisions. The company itself notes the approach depends on integrations and human oversight, so rollout quality will vary from one deployment to the next.

Sources:
https://decagon.ai/blog/missions
https://unite.ai/decagon-launches-missions-for-long-running-ai-customer-workflows
https://futurumgroup.com/insights/decagon-missions-can-purpose-built-orchestration-win-enterprise-cx

IMPACT checklist


Continue the conversation: https://t.me/AIAutomationArchitecture ยท Bluesky

AI Automation

Top comments (0)