Welcome to the O'Dovey AI blog.
O'Dovey AI is a professional services firm of forward deployed engineers. We embed with your team, build the agents and platforms your organization needs, deploy them into your environment, and make sure your people can run them. This blog is where we share what we learn doing that work.
What We Do
Most organizations have seen an impressive AI demo. Far fewer have agents running reliably against their real systems. Getting from one to the other is the work we focus on, and we help in three ways:
- Build & Deploy — Our engineers design, build, and deploy agentic solutions in your environment: agentic platforms with governed model access, routing, policy, and observability; agents and agentic workflows, from structured automation to autonomous, multi-step agents; custom agent harnesses your organization owns and controls; and the integrations that connect agents to your data and systems.
- Enable — We pair with your engineers on real work through co-build engagements and technical advisory, so the skills stay with your team after we leave.
- Train — We run corporate training on building agents, from executive briefings to hands-on engineering bootcamps.
We're vendor-neutral. We don't resell models or platforms, and we design so you can change them later without starting over.
What to Expect Here
We'll write about what it actually takes to get agents out of the demo stage and into production:
- Harness Engineering — Every agent is a model inside a harness. We'll dig into the input, context, model interface, output, and control flow that turn a model into something that does real work.
- Agentic Platforms — The shared foundation for building and running agents across an organization, and the decisions that make the second agent far cheaper than the first.
- Agents and Agentic Workflows — Patterns for everything from workflow-driven automation to autonomous agents, including where humans belong in the loop.
- Production Readiness — Persistence, security, observability, resilience, performance, and evaluation: the questions every production agent has to answer.
Teaching in the Open
We believe the fastest way to raise the bar for agentic AI is to teach in the open. Our Harness Engineering curriculum is free and open source: it builds an agent from scratch, one primitive at a time, evaluates it, and gets it ready for production. Our first post, Harness Engineering: Building the Body of an AI Agent, walks through the ideas behind it.
Stay Connected
Whether you're a technology leader deciding where agents belong, a platform engineer standing up shared infrastructure, or a developer building your first agent, you'll find something useful here. And if you're ready to put agents to work in your organization, let's talk.
Welcome aboard.