Introducing Duet: Your agent building partner

Today, we're announcing Decagon Duet, an AI partner that enables teams to build self-improving agents. Duet ships better agents faster, identifies opportunities for improvement, and takes action across the platform, compressing days of effort into minutes.

Since we launched Decagon, we’ve had a front row seat to how leading enterprises are adopting AI to provide concierge customer experiences. Our agents have now handled hundreds of millions of customer conversations across some of the most complex industries, with measurable impact for both contact center efficiency and customer happiness.

One theme has become obvious: agent building is iterative by nature. The initial build is only 10% of the work; the other 90% is what comes after. Agents need continuous updates as new conversations reveal performance gaps and opportunities to serve customers better. That insight has shaped how we’ve architected Decagon, empowering teams to iterate quickly without requiring a vendor support ticket for every change.

Now we’re taking the biggest step yet in that vision. Instead of manually connecting insights across dashboards, testing tools, and workflows, teams can simply talk to Duet to manage the full agent lifecycle, from the first deployment to the hundredth improvement.

Duet is currently available to a limited group of customers. If you're interested in being among the first to build with Duet, we'd love to hear from you.

SaaS isn’t dead, but its adoption model is

Headlines of the “SaaS-pocalypse” are provocative, but overstated. What's actually dead is the adoption model that defined the last 20 years: complex SDKs, months of configuration with professional services, and black-box implementations that only the vendor understands. 

This approach worked because most traditional software is relatively static, but AI systems are different. They require continuous refinement, and every day spent waiting on an engineering sprint or a vendor's backlog for a simple change is a day your agent falls further behind. This is why we've been so intentional about building Decagon differently than others in our space.

To date, we've pioneered a wide range of features that make agent development faster:

• Agent Operating Procedures (AOPs) to define and control agent behavior in natural language
• Simulations to test agents against large sets of scenarios before release to production
• Versioning to experiment safely and A/B test workflow updates
• Watchtower to monitor, QA, and debug agent performance against live conversations
• Insights & reporting to diagnose key trends and prioritize improvements

Our customers have had tremendous success using these tools, but building an agent still required the technical depth to interpret data, the instincts to prioritize the right changes, and the time to act. Duet changes that equation, putting that expertise within reach of any team.

Meet Duet

Developing an AI agent should be as easy as talking to it. Duet is designed around that idea.

It distills the best practices we've learned about conversational design, agent building, and hillclimbing from hundreds of enterprise deployments. The result is a step-change in what's possible: faster iteration, broader ownership, and agents that keep improving long after the initial launch.

Build with Duet to launch better agents faster

AI has changed who can build. Two years ago, shipping changes required engineering resources. Today, tools like Claude Code enable non-technical users to develop production-ready software without depending on those teams.

Duet provides exactly that for your Decagon agent. By analyzing real customer transcripts, it identifies necessary workflows and automatically generates AOPs with detailed logic, grounded in customer needs and best practices for conversational design.

As customer behavior and business requirements evolve, Duet keeps pace. It flags where existing AOPs need updating and validates that newly built workflows are correct before they go live, giving teams confidence to ship changes faster.

Optimize with Duet to learn from every interaction

Improving an AI agent requires testing it across a wide range of scenarios. Many issues only appear under specific conditions: unusual customer intents, unexpected tool outputs, or edge cases in workflows.

Duet automatically generates test scenarios tailored to your agent's workflows, simulates conversations, stresses edge cases, and flags situations where the agent may fail.

It also goes deeper into production: Duet investigates live conversations to understand what's actually happening and tunes AOPs accordingly, surfacing guardrail gaps and driving targeted improvements.

Human QA insights feed directly into this loop, allowing teams to diagnose issues and validate fixes much faster than manual analysis alone.

Scale with Duet to act on the biggest improvements

As agents handle more conversations, they also become a rich source of customer insight. Understanding those patterns is critical for improving both the agent and the broader customer experience.

Duet helps teams make sense of this data by analyzing and generating custom reports that highlight trends in customer behavior, recurring issues, and opportunities to improve workflows. This allows teams to focus their efforts on the changes that will have the largest impact on resolution rates, operational efficiency, and customer satisfaction.

When deflection dips, Duet can identify which intents are dragging performance and explains why. When knowledge gaps exist, Duet pulls evidence directly from real conversations to suggest specific rewrites, grounding every improvement in what customers are actually saying instead of guesswork.

From reactive support to concierge experiences

Support is often the first use case for AI agents because the value is immediate and measurable, but the long-term opportunity is broader.

Once teams can iterate quickly and safely on their agents, they can begin expanding into new workflows and customer experiences, such as proactive outreach, product discovery, and in-app assistance.

The end goal is not simply to resolve tickets more efficiently, but to build an AI concierge that understands customers deeply and helps them accomplish their goals. Faster iteration—and Duet—is what makes that future possible.

‍

Reach out to us today if you’re interested in building with Duet.