Building Your First AI Agent: Tools, Memory and Guardrails
Most people have used AI as a conversational assistant. AI agents go a step further: they plan, decide which tools to use, take actions and work through multi-step tasks with limited supervision. This masterclass moves beyond the concept and shows, in practical terms, how an agent is actually put together and what makes one reliable enough to trust.
This masterclass will guide you through a live build. Starting from a blank project, we create a small working agent, give it tools such as web search, data lookup and messaging, add memory so it can carry context between steps, and then deliberately push it until it fails. Along the way, you’ll see where agents deliver real value, where they go wrong, and which guardrails keep them safe in a business setting.
Learning Outcomes
By the end of the session, participants will be able to:
- Understand the agent loop: Learn how agents reason, choose tools, act and observe results, and how this differs from a single prompt and response.
- See an agent built live: Watch a working agent assembled step by step, including tool definitions, memory and a simple orchestration loop.
- Recognise failure modes: See common problems in action, such as tool misuse, looping, runaway costs and confident but wrong actions.
- Apply guardrails: Learn practical controls including human approval steps, permission scoping, logging and cost limits, and how to decide which tasks are suitable for agents.
Target Audience
This masterclass is designed for hands-on practitioners, including developers, data engineers, analysts and architects, who already use generative AI tools confidently and want to go further.
It is also relevant to technical managers who need to understand what agents can realistically do and the risks they introduce.
Basic coding knowledge will help practitioners get the most from the demos but is not required to follow along.
