At Decagon, Agent Operating Procedures (AOPs) are at the core of how we build AI agents. Much like Standard Operating Procedures for human agents, AOPs are structured, natural language instructions that power AI workflows with enterprise-grade reliability and precision. The AOP approach empowers CX teams to quickly shape agent behavior while allowing technical teams to maintain control of their integrations, tools, and guardrails, driving faster iteration without sacrificing stability.

Today, we’re launching AOP Copilot to provide every Decagon user with an intelligent assistant that takes AOPs’ speed and usability even further. With just a natural language prompt, AOP Copilot generates production-ready workflows in seconds, streamlining the entire journey from idea to deployment and making it easier than ever to both build and scale high-quality AI agents.

Why we built AOP Copilot

Many organizations already have extensive Standard Operating Procedures (SOPs) that codify their most critical customer support workflows. Because AOPs are written in a similar style to SOPs, Decagon’s approach already makes it straightforward to translate this institutional knowledge into AI-powered workflows.

But as adoption has grown, we saw an opportunity to accelerate that translation process. Instead of manually reworking SOPs into AOPs, AOP Copilot can ingest those instructions and instantly generate a structured workflow for the AI agent. This dramatically shortens the path from documentation to deployment, allowing teams to cover more scenarios with less effort.

AOP Copilot is also a powerful companion for teams starting from a rough idea rather than a formal SOP. By generating a first-pass AOP skeleton, it provides a solid foundation that can be further refined and extended. Whether starting from detailed playbooks or quick concepts, Copilot helps every team move faster while maintaining the precision and guardrails that AOPs are known for.

How it works under the hood

At its core, AOP Copilot understands the full library of workflow step types and how they connect to form an AOP. With this context, it assembles workflows that are validated for structural soundness so the resulting AOPs are ready to be directly run in production.

We’ve also tuned AOP Copilot to consistently apply AOP writing standards, making sure every generated workflow aligns with Decagon’s best practices for clarity, modularity, and performance. This means users don’t just get a functional draft, but a structured AOP that’s been designed to scale, adapt, and perform under real-world conditions.

AOP Copilot is embedded directly within the Decagon UI, so users never have to leave the environment where they already manage their agents. Drafting, revising, and deploying an AOP all happens in the same place, shortening the feedback loop and keeping teams focused on improving outcomes rather than juggling tools.

What’s next

AOP Copilot is just the beginning of how we’re accelerating AI workflow creation at Decagon. Next, we’re focused on expanding its capabilities for debugging and refinement, proactively surfacing where logic may break down and recommending targeted fixes. These updates will make it even easier for teams to maintain reliable, high-trust workflows as they scale.

As these capabilities roll out, AOP Copilot will evolve from simply accelerating workflow creation to helping teams sustain and scale excellence in their AI agents. It’s another step toward our vision of giving every organization the tools to deliver concierge customer experiences that are fast, transparent, and reliable at scale.