Getting closer to the build: How I helped create agent deployment engineering

For a long time, product management was the closest role I could find to the work I wanted: sitting with customers, understanding messy problems, and turning them into something a team could build.

My role was to orchestrate that process. Then generative AI made me a faster, more capable orchestrator. I could prototype ideas, sharpen specs, and extract insights on my own.

But even with that expanded range, I still wanted my hands on the solution itself—to build it end-to-end and stay with it in production.

That search led me to Decagon, where I joined as a Senior Agent Product Manager (APM). As an APM, I owned the broader deployment: understanding the customer’s goals, coordinating work across teams, and keeping everything moving toward launch.

That breadth was what drew me to the role.

On our most complex deployments, however, we saw that the two sides of the work demanded different kinds of depth. The customer relationship and agent requirements required sustained partnership. The agent build required sustained technical focus: continually designing, testing, debugging, and improving its behavior.

Why we created Agent Deployment Engineering

That realization led us to create Agent Deployment Engineering: a team of customer engineers who work alongside APMs and customers while owning the technical agent build.

Creating the ADE function gives each discipline room to go deeper. APMs lead the customer relationship and broader transformation, keeping the work tied to the right outcome. ADEs translate that customer context into a production-ready agent.

At its core, Agent Deployment Engineering exists to make agent building a center of excellence at Decagon.

What ADEs own and how the work compounds

In practice, ADEs own how customer requirements become agent behavior on Decagon’s platform.

Working with APMs and customers, they translate business goals, policies, and operating requirements into technical decisions. They build the instructions and tools that govern the agent, test them against real conversations, diagnose unexpected behavior, and improve performance after launch.

The work is highly creative and iterative.

Every design requires constant tradeoffs among reliability, latency, precision, maintainability, and speed. The right answer is rarely obvious. ADEs have to uncover the problem beneath the requirement, iterate on a solution, and prove that it works.

Because ADEs work deeply across real-world deployments, they see the same design decisions, failure modes, and product constraints recur. Those patterns create leverage in two directions.

First, every deployment strengthens the agent-building craft. ADEs turn recurring lessons into reusable architecture patterns, tools, and testing practices, so the next builder starts further ahead.

Second, every deployment strengthens the product. ADEs bring recurring platform and integration gaps to the broader product and engineering teams, turning them into shared capabilities.

The result is a compounding loop: each deployment makes the next agent easier to build and the Decagon product more capable.

The people who thrive as ADEs

The people who thrive here are technical thinkers who are energized by hard customer problems and comfortable navigating ambiguity.

They are tinkerers who cannot stop building and testing. They have strong debugging instincts and enjoy reasoning through unfamiliar systems. When something fails in production, they want to trace the behavior to its source and fix it.

They also want to work directly with customers.

They can move from a technical investigation into a customer conversation, uncover the problem beneath the initial request, and explain the tradeoffs behind their solution. Above all, they have the grit to own the outcome. Customer requirements change, technical constraints emerge, and the first solution rarely survives contact with production.

The people who thrive here stay with the problem until it is solved.

Agent Deployment Engineering is still an emerging craft. We are looking for engineers who want to help define how production agents should be architected, tested, deployed, and improved.

If that combination of technical depth, customer proximity, and end-to-end ownership sounds like the work you have been looking for, come help us build it. Explore careers at Decagon.

‍