Introducing Missions: Own end-to-end outcomes

Most AI agents are built for a single conversation. But the customer outcomes with the biggest ROI often take multiple touchpoints to achieve. Gaining prior authorization for an upcoming MRI, approving a small loan, or qualifying a lead are all multi-step journeys that drive meaningful revenue.

Today, we’re introducing Missions to drive customer outcomes from first contact to final confirmation. Missions power long-horizon tasks. They coordinate the agent across systems, monitor status updates, and engage multiple people as needed to move the work forward over days or weeks. 

Missions change how teams deliver customer experience. Instead of tracking statuses and managing handoffs, they can focus on improving every step of the Mission to optimize the goal outcome. Teams define the objective once, the agent sees it through across unique customer interactions, and together they refine the journey over time.

With Missions, teams can:

• Drive end-to-end outcomes with intelligent orchestration across multiple parties, conversations, systems, and days
• Ensure the best next step each time, with user memory and Mission context
• Optimize every Mission with logic built on AOPs and designed for continuous with Duet Autopilot

Drive end-to-end outcomes 

Missions orchestrate multi-step workflows that carry a business goal all the way to completion. Setup is simple: state the goal in plain language, name the personas to engage, and add any guidance to follow. The agent treats that goal as its north star, choosing the best next step at each point until the outcome is achieved, whether that takes days, weeks, or months.

Most goals involve more than one person. Missions let the agent connect threads across multiple personas and reach each one on the most relevant channel. When the best next step isn’t a conversation at all, Missions extend into back-office work, such as completing a form using Browser Actions. When a third party is required to complete a step, the agent patiently waits for completion before triggering the next best action.

For example, healthcare partners have turned to Decagon to drive prescription refills from beginning to end. This starts with patient notification, then the agent proactively engages the provider when a refill needs authorization, re-runs claims directly on a payer portal, and concludes by confirming the pickup details with the patient to complete the outcome. 

Ensure the best next step each time

To inform every touchpoint, the agent needs context on where it’s been, where it’s headed, and any details that make this particular case unique. Each Mission supplies that context dynamically, so the agent consistently chooses the best next step toward the goal for each individual it serves.

Mission cases act as a living record of the agent’s progress toward its goal. It captures the steps the agent has taken and any new information that could shape what happens next. The agent uses the case to set its next step, schedule proactive outreach, and save context that helps it keep moving toward the outcome.

While Mission cases keep the agent on track, Memory makes each touchpoint personal. Memory includes customers’ preferences, such as not being called after 3pm, and any context from other conversations that the customer is having with the agent. A customer might be engaging with a Mission to help them onboard onto a new product while also troubleshooting an unrelated product issue, and the agent can account for both while acting as a consistent and cohesive concierge.

Customers don’t always follow the plan, so the agent doesn’t either. Before each next step, the agent rechecks the Mission case and the user Memory to make sure its plan still fits. Say the agent has qualified a lead and scheduled a follow-up. If the customer emails beforehand to schedule the follow-up themselves, the agent adjusts the course accordingly.

Optimize every Mission

Optimization is both the biggest challenge and the biggest opportunity with long-horizon agents. As interactions grow longer and more complex, teams face an attribution problem: when an outcome falls short, it’s hard to pinpoint why. If a prescription refill is delayed, was it the initial outreach, the provider handoff, or insurance verification? 

We've designed Missions with ease of iteration as a first principle. Every Mission is built as an Agent Operating Procedure (AOP), so teams can write, edit, and refine the logic themselves. They maintain full visibility into every decision the agent makes and can test each part of the configuration to understand what’s working and what could be improved.

With Duet and Duet Autopilot, teams can continuously improve not just individual interactions, but the entire customer journey. With every recommended adjustment, teams drive measurably better outcomes for their customers. 

If you’re interested in owning end-to-end outcomes for your customers, book a demo.