John Berryman, an AI consultant at Arcturus Labs, led a workshop on building workflow agents in 30 minutes, according to O'Reilly Radar. The session defined an agent as a large language model operating within a structured workflow environment.
Berryman has spent years shipping AI products and helping teams deploy them at scale. His expertise spans the gap between research prototypes and production systems, where most AI failures occur. The workshop format suggests a practical, hands-on approach rather than theoretical foundations.
The "Zero to Agent" framing indicates the goal: getting developers from no agent experience to functional implementation quickly. This reflects a broader industry trend. AI agents have moved from research curiosity to business necessity. Companies need developers who can build and deploy them immediately, not spend months in lengthy training programs.
Workflow agents differ from chat-based AI systems. They execute multi-step tasks autonomously, make decisions based on available tools, and integrate with existing business processes. A workflow agent might approve expense reports, coordinate team scheduling, or route customer inquiries to specialists. The intelligence comes from the language model, but structure comes from the workflow system.
The 30-minute constraint matters. It signals that agent building is becoming accessible. Complex AI work once required specialized teams. Now, the tooling and frameworks have matured enough that competent developers can prototype working systems in under an hour. Frameworks like LangChain, CrewAI, and Anthropic's native tools have reduced friction dramatically.
O'Reilly's promotion of this content reflects market demand. Developer audiences increasingly need practical agent skills. Companies hiring engineers expect them to understand how to wire LLMs into production workflows, handle failures gracefully, and integrate with existing systems.
The session likely covered essential elements: defining agent goals, selecting tools, handling errors, and testing behavior in constrained environments. Berryman's production