World's Fair 2026
Build Systems, Not Code - Angie Jones, Agentic AI Foundation
Angie Jones , Agentic AI Foundation
Overview
This talk argues that building agentic AI systems requires the same core engineering discipline as traditional software development, just with different primitives. Instead of focusing solely on using agents to write code, the emphasis shifts to architecting complex agentic systems. This approach allows engineers to leverage their existing skills in systems thinking, workflow design, and modularity, recapturing the thrill of building by operating at a higher level of abstraction.
Who should watch
- AI Engineers
- Product Managers
- Builders of AI-powered applications
- Those interested in the practical application of agentic AI
- Engineers seeking to apply traditional software engineering principles to AI systems
Key takeaways
- Building agentic systems involves architecting the overall environment, not just prompting an agent to write code. Treat agents as components with defined boundaries and responsibilities.
- Workflow design is crucial for agentic systems, defining the path an agent takes to achieve a goal, including gathering information, weighing criteria, and determining outcomes like stopping, retrying, or escalating.
- Decomposition is essential to avoid monolithic prompts; break down complex agent tasks into smaller, manageable pieces with distinct jobs.
- Separation of concerns applies to agents, determining where specific logic (e.g., data normalization, calculations, complex research) should reside, whether in prompts, skills, scripts, or sub-agents.
- Modularity, through reusable agent skills and sub-agents, allows for code reuse across different agents and workflows, similar to libraries in traditional software.
- Algorithmic thinking dictates when to use code versus an agent: use code for deterministic tasks with exact answers (like calculating commute times) and agents for tasks requiring interpretation, judgment, or reasoning over messy input.
- Structured output and defined contracts are vital for agentic systems, enabling reliable communication between system components and making agent memory queryable for downstream processes.
- Designing for state management and idempotency is critical, as agentic systems must handle failures and retries gracefully without unintended side effects, often by logging actions to memory.
Notable quotes
*The thrill is still there. When you're building agents, not just using them to write code, you start getting into architecting agentic systems.*
*An agent is not the system, it's part of the system. And that system has files and tools, humans, even other agents.*
*If a task has an exact answer, reach for code. If it needs interpretation or judgment, that's when you can get the agent to do it.*
Unofficial community note. Prefer the recording for nuance.