Agentic AI requires new systems thinking
Bill Doerrfeld | April 20, 2026
My
latest for InfoWorld explores emerging best practices for building agentic systems.
What actually goes into building enterprise-grade AI agents?
Excited to share my new feature for InfoWorld exploring emerging best practices for building agentic systems.
I like how Ari Weil puts it: "Building an AI agent is like constructing a nervous system."
That framing holds up. Getting this right requires coordinating a wide range of components: reasoning models, context, tools and discovery, workflows, orchestration, and security.
My piece reviews these building blocks and digs into new practices like context engineering and agent evaluation, with real-world examples from Shopify and Block.
The main takeaway? Agentic systems aren't just a new layer of automation. They require a new form of systems thinking.
Other Blog Posts
My latest InfoWorld feature explores what makes an enterprise MCP registry effective, from semantic discovery to governance and security for AI agents.
My first-ever contribution to CSO Online looks at the shifting landscape, from perimeter-based security to API security, and how CISOs are responding.
My latest feature for The New Stack looks into solutions being proposed to fix open source Slopmageddon.
My latest DirectorPlus looks at how agentic AI is reshaping platform engineering at Squarespace: less shared code and more developer experience focus.
Usage-based pricing is reshaping the API economy. Discover 5 API monetization success stories, including OpenAI, Plaid, and AssemblyAI.
Why event-driven APIs matter for AI workflows, enabling real-time data, scalable systems, and responsive agent behavior.
While hardware usually gets the spotlight in physical AI, the real differentiator won't be hardware. It'll be the models.
In the latest DirectorPlus, Workato's CTO explains how MCP-enabled integration catalyzed internal AI usage and ROI.
My latest on InfoWorld reviews MCP servers from 5 major cloud providers
How are organizations actually using agentic knowledge bases in practice? My article for The New Stack looks at six emerging patterns.
