Decagon Missions: AI Agents That Work for Weeks

Neon cyber illustration of a glowing agent timeline stretching across a calendar, with the words "WEEKS, NOT MINUTES."

Most AI agents have the attention span of a single chat. You ask, it answers, the conversation ends. But real customer problems do not work like that. Getting an MRI approved takes days. A small loan takes a week. A prescription refill can bounce between a patient, a doctor, and an insurer for a month.

Decagon, a company that builds AI agents for customer service, introduced Missions on 12 November 2026 to close that gap. A Mission is an AI agent given one objective and the time to finish it — days, weeks, or months — across many conversations, systems, and people.

From one chat to one outcome

Today's customer-service agents are good at single interactions: answer a question, reset a password, check an order. Missions change the unit of work from the conversation to the outcome. The agent keeps going until the job is done, even if that means coming back tomorrow, or next week, with new information.

Decagon gives three examples of the kind of journey a Mission can run. Getting prior authorization for an MRI, which means gathering documents, contacting the provider, and waiting on an insurer. Approving a small loan, which means checking details, asking follow-up questions, and reaching a decision. Qualifying a sales lead over several touchpoints instead of one call.

Five things that make Missions work

1. One objective, carried across conversations

A Mission is defined by its outcome, not by a single chat transcript. The agent remembers what it is trying to achieve and picks the work back up in the next conversation, with the next person, on the next system.

2. Agent Operating Procedures

Every Mission is built as an Agent Operating Procedure, or AOP — a structured plan the agent follows and can be tested against. It is the difference between an agent improvising each step and an agent running a procedure a business has actually designed and approved.

3. Browser Actions for back-office work

Missions extend beyond conversations. Through Browser Actions, an agent can do back-office work such as completing forms on a portal — the unglamorous clicking that usually waits for a human with a login.

4. Cases and Memory

Decagon's cases and Memory shape what the agent does next. What happened in the last conversation, and what the business already knows about this customer, inform the next step instead of starting from zero each time.

5. Optimisation for the whole journey

Decagon identifies attribution as the central difficulty when automated work stretches across many interactions: which step actually caused the outcome? Its Duet and Duet Autopilot tools aim recommendations at the full customer journey, not at single chats, so teams can see and improve the whole path.

The voice upgrade that came with it

Missions arrived alongside Alto, a new voice model from Decagon Labs announced on 11 November. Alto powers the conversational layer of Decagon's Voice 3 architecture, deciding on every turn whether to answer immediately or wait for the reasoning model to confirm an action before speaking.

Decagon says Alto lifted resolution rates by more than five percentage points on live customer calls, with average responses in about 700 milliseconds, after testing across more than 90 million simulated calls. Those are company-reported figures: Decagon did not identify the customers, the sample size, the baseline rates, or the measurement period, so treat the lift as a claim rather than an independent result.

What to watch

Long-horizon agents raise questions that single-chat agents do not. Who is responsible when an agent takes a wrong step on day nine of a two-week process? How does a customer check what an agent has done on their behalf while they were not watching? And how does a business audit a decision that emerged from twenty small steps across four systems?

Decagon's answer, in structure at least, is the AOP: a written procedure, plus cases and Memory that record the path taken. Whether that is enough in practice is something only real deployments will show. Pricing for Missions has not been disclosed either.

Key takeaways

  • Decagon launched Missions on 12 November 2026: AI agents that pursue one customer outcome for days, weeks, or months.
  • Examples include MRI prior authorization, small-loan approval, lead qualification, and prescription refills.
  • Every Mission is built as an Agent Operating Procedure, with Browser Actions for back-office work and cases plus Memory to carry context forward.
  • Decagon's Alto voice model, launched 11 November, decides on each turn whether to answer or wait — with a claimed 5+ point lift in resolution rates on live calls.
  • Pricing is not disclosed, and the Alto lift is a company claim without published sample size or baseline.

Conclusion

The agent race is moving from faster answers to finished jobs. Missions are Decagon's bet that the next useful unit of AI work is not the chat but the outcome — the approval, the refill, the decision — however many days and systems it takes. If the procedures hold up and the audit trail works, that is a genuine step forward for customer service. Watch the first real deployments: the proof will be in the outcomes, not the announcement.

FAQ

What are Decagon Missions?

Missions are AI agents from Decagon that manage a customer outcome across many conversations, systems, and participants, over days, weeks, or months, instead of ending with a single chat.

What jobs can a Mission handle?

Decagon's examples include prior authorization for an MRI, small-loan approval, lead qualification, and prescription refills — multi-step journeys involving several people and systems.

What is Decagon Alto?

Alto is a voice model from Decagon Labs, launched 11 November 2026. It controls the conversational layer of Decagon's Voice 3 architecture, deciding on each turn whether to answer immediately or wait for the reasoning model.

Researched with AI assistance and reviewed for accuracy.

Sources

Follow on WhatsApp

Follow Daily Tech & AI on WhatsApp for the latest tech and AI news, every day: Daily Tech & AI on WhatsApp

Comments