Session
Anatomy of an AI Assistant
Friday 24 July
14:00 – 15:00
Educator 1
Back to agenda
Chatbots wait for a prompt; agents wait for a goal. This talk is a deep dive into the transition from passive AI to active, autonomous partners. I’ll demonstrate a personal assistant agent I built to handle my daily admin and explain the engineering that makes it work.
We will deconstruct the "Agentic Stack" piece by piece, focusing on how to build a reliable system without relying on opaque, one-click solutions.
System Prompting for Autonomy: Crafting the "DNA" that keeps an agent on-task without human hand-holding.
Tools and Sub-Agents: When to give your agent an API and when to give it a "helper" agent.
Multi Memory Types: What should you store where, and how do you manage context around the current task.
The Agentic Loop: Wrapping everything in a loop to bring proactivity to LLMs.
Stop using AI as a chat and start building it into a 24/7 proactive helper.
We will deconstruct the "Agentic Stack" piece by piece, focusing on how to build a reliable system without relying on opaque, one-click solutions.
System Prompting for Autonomy: Crafting the "DNA" that keeps an agent on-task without human hand-holding.
Tools and Sub-Agents: When to give your agent an API and when to give it a "helper" agent.
Multi Memory Types: What should you store where, and how do you manage context around the current task.
The Agentic Loop: Wrapping everything in a loop to bring proactivity to LLMs.
Stop using AI as a chat and start building it into a 24/7 proactive helper.
