Already using AI, missing the guardrails

Some rooms need convincing that AI is worth learning. This one didn't. Between them the group were already running six different tools, ChatGPT, Copilot, Gemini, Claude, NotebookLM and Perplexity, across research, job applications, revision and planning. The words they used to describe AI were practical rather than starry: useful, adaptive, convenient, fast, career survival. Their worries were just as concrete: over-reliance, personal data, bias, ethics, reliability, accountability. They had worked out on their own that nobody had taught them how to use any of it responsibly. The gap wasn't access, and it wasn't enthusiasm. It was judgement, knowing when an output can be trusted, and who is accountable when it can't.

• A hands-on session for a small LUBS student group
• Live polling on current AI use, confidence and concerns
• The difference between generative AI and AI agents
• Responsible use, reliability and accountability in practice
• Use case thinking for work, study and personal productivity
• Applied examples tied to careers and business tasks

The session opened with a poll rather than a pitch. Twelve students answered the first question; 17 took part in polling across the session, the smallest of the four LUBS blocks, and the most fluent. What came back wasn't scepticism. It was a group describing AI in terms of career survival, and then listing over-reliance, bias and accountability as the things that worried them. They were describing a tool they already depended on and had never been given rules for. So the session led with the rules. Where generative AI ends and an agent begins. What reliability actually means when an output looks confident and is wrong. Who is accountable when a tool you didn't build makes a decision you signed off. Then the room split into three teams. Three hours, Microsoft Copilot, one working agent each, presented back to their peers at the end. What they chose is worth noting. Stocky, a stock market dashboard. Travel Trevor, a scheduling agent with explicit rules around what it was allowed to do. Stuey the Student House Hunter, for students trying to find somewhere to live in Leeds. Personal problems, not textbook ones. And a group that had spent the session on reliability and accountability picked financial markets as the thing to build, the highest-stakes subject on the board.