Two Workshops, One Day: Parents in the Morning, Kids in the Afternoon
On May 30, 2026, a Hong Kong secondary school hosted a parent workshop about AI. Solo delivery in Cantonese, funded by the school's Home-School Cooperation Grant, school hall in Tseung Kwan O, folding chairs, a projector. About a hundred parents. Their questions arrived before the session through a registration form, and reading through them was more telling than any industry report on AI anxiety.
The questions weren't about tools. They were about their children.
"How do I know my child isn't just letting AI think for them?" "Is it too late for me to understand this?" "What career should my child pursue if AI can do everything?" These were parents asking questions they didn't have a good person to ask -- not their children, because the power dynamic runs the wrong way, and not the internet, because the internet either tells you everything is fine or everything is collapsing, and neither helps on a Saturday morning when you're wondering whether your twelve-year-old's homework is still their own.
What I actually tell parents
The session was built around two layers. The first is for the parents themselves, and it's the layer most AI-in-education discussions skip: before you can guide your child's AI use, you need to have used the tools yourself. Not hypothetically. Open NotebookLM. Upload a document you already have -- a work report, a recipe, anything real. Ask it a question. See that it works. Then ask it something it should get wrong, and watch it fail. That ten-minute exercise does more to calibrate a parent's understanding than an hour of explaining what large language models are. The parent walks out with a direct experience of what the tool can do and where it breaks -- not reassurance, not "it'll be fine," but a specific encounter with the thing being, actually, a tool. Imperfect and still useful.
The second layer is harder: how do you guide a child's AI use when nobody -- not the schools, not the industry, not the people building the tools -- has fully figured it out? Pretending otherwise would have lost the room, so I didn't. But "nobody has it solved" is not the same as "there's nothing to do." The working approach: treat it the way you'd treat any tool a child has access to that's more powerful than their judgment. Use it alongside them. Ask them to show you what they made and what they changed. The children who describe what they want, evaluate what they get, and adjust their language until the output matches -- those children are building a skill the corporate world calls "prompt iteration." They just do it naturally when they care about the result.
What the afternoon showed
That same day, twenty children -- ages nine to fifteen, from low-income families -- built AI-generated posters at a the food manufacturer CSR workshop. No homework pressure. They described what they wanted, got something different, tried again with different words. The gap between what they described and what the tool produced wasn't frustrating the way it is for adults -- it was a problem to solve, and they solved it by adjusting until the poster looked right.
The parents' questions that morning centered on protection. The children's experience centered on access and time with the tools. The comparison isn't quite fair -- children don't have mortgages or careers to protect, and the parents' fears are rational, grounded in real economic uncertainty. A two-hour workshop doesn't dissolve that.
What it can do is give a parent their own hands-on encounter with the tool, and then give them language for the conversation with their child. The language is simpler than most people expect: "Show me what you made. What did you change? What did the tool get wrong?" Those three questions work whether the child is nine or fifteen. They're not an AI-in-education framework -- they're what you'd ask about any first draft.
