Agentic engineering

Overview

Agentic engineering is the practice of building software systems with the help of agents that can plan, write, and execute code with a meaningful degree of autonomy. In the newsletter, the term is used primarily in the context of coding agents and the operational patterns that emerge when software development shifts from purely human-driven workflows to human-and-agent collaboration.

For AI Product Managers, agentic engineering matters because it changes both how products are built and how agent-powered products should be governed. It introduces new leverage in development speed and experimentation, but it also creates new failure modes around review, execution safety, autonomy boundaries, and behavior management. As a result, PMs need to think not only about capability, but also about controls, anti-patterns, and the design of systems that keep agents useful without making them reckless.

Key Developments

• 2026-03-05 — Simon Willison highlighted anti-patterns in agentic engineering, especially the risk of agents producing or pushing unreviewed code to collaborators.
• 2026-03-15 — Simon Willison discussed agentic engineering in a fireside chat at the Pragmatic Summit in San Francisco, hosted by Eric Lui from Statsig, signaling growing industry interest in the topic.
• 2026-03-16 — Simon Willison published a definition-oriented piece explaining agentic engineering as software development assisted by coding agents that can both write and execute code.

Relevance to AI PMs

• Define autonomy boundaries clearly. PMs need to decide which tasks an agent can propose, execute, or ship autonomously versus which require human approval. This is foundational to reducing risk while preserving speed.
• Design review and safety workflows. Agentic engineering makes code generation only part of the story; execution rights, testing requirements, and human review gates become product decisions as much as engineering decisions.
• Manage agent behavior as a product surface. Anti-patterns such as over-delegation or allowing unreviewed outputs to affect shared systems show that agent behavior management should be treated like UX and policy design, with explicit expectations, fallback paths, and monitoring.

Related

• Simon Willison — A key voice in defining and discussing agentic engineering, especially through practical writing on coding agents and anti-patterns.
• Coding agents — The most direct enabling technology behind agentic engineering; these agents can write and often execute code.
• Pragmatic Summit — A venue where agentic engineering was discussed publicly, indicating broader practitioner interest.
• Statsig — Connected through Eric Lui's hosting of the Pragmatic Summit fireside chat on agentic engineering.
• Anti-patterns — Closely linked because much of the practical conversation around agentic engineering focuses on what not to let agents do without oversight.