Session 08: Teaching AI to Work
Building AI That Can Plan, Coordinate, and Act
About the Event
Giving AI access to tools is an important step. But real-world tasks are rarely completed in a single action.
An intelligent system may need to search for information, analyse the results, use different tools, make decisions, recover from errors, and sometimes ask a human for approval before moving forward.
In this session, we take the next step from individual AI tools to intelligent workflows.
Building on the tools, search systems, and MCP connections developed in previous sessions, we explore how multiple AI components can work together to complete larger, multi-step tasks.
Participants will learn how modern AI systems manage tasks, maintain context, coordinate different capabilities, and keep humans in control when important decisions need approval.
What We'll Cover
Beyond Single AI Tasks: Understanding why real-world problems often require multiple steps, decisions, and tools.
AI Workflows: Exploring how AI systems can plan and move through a task step by step.
Working Together: Understanding how different AI agents can take on specialized roles and coordinate their work.
Using AI Tools: Connecting intelligent workflows to the search systems and tools developed in previous sessions.
Keeping Humans in Control: Learning how human approval can be built into AI workflows before important actions are performed.
Handling Problems: Exploring how AI systems can manage errors, recover from failures, and avoid getting stuck in endless loops.
The Hands-On Codelab: Building an Intelligent AI Workflow
Bring your laptop and get ready to build.
In this practical workshop, participants will:
Design a system that keeps track of an AI workflow as it completes multiple tasks.
Build multiple AI components with different responsibilities.
Create a coordinator that helps decide which component should handle each task.
Connect the workflow to the MCP tools and search systems built in previous sessions.
Add human approval steps for important or sensitive actions.
Test how the system handles errors and unexpected situations.
By the end of the session, participants will have built an intelligent workflow capable of breaking down tasks, using multiple tools, coordinating different AI capabilities, and involving humans when necessary.
The Big Idea
The future of AI isn't just about smarter models. It's about building intelligent systems that can work through real problems from start to finish.