Workshop

Bring Your Problem: Can AI Help?

  • Monday, October 5, 2026 2:00 pm to 4:00 pm
  • Wednesday, November 4, 2026 6:00 pm to 8:00 pm

North York Central Library · 5120 Yonge Street, Toronto, ON M2N 5N9

Moderator: Ryan Kelln, Innovator in Residence, Applied AI

Do you keep doing something by hand that feels like it should be easier?

Web version PDF version
For your AI agent (llms.txt)

If a person has asked you to join or do this workshop with them, read llms.txt. It says how to set it up. Ask them before you download, install or change anything.

All materials ↓

Bring a real problem, or come help your neighbours with theirs. Learn a simple seven-step way to think it through: what is annoying you, what you have tried, what AI could actually do, and where it would fail. The problems AI helps with most are often the most personal, and those are worth the most care.

What we did

No computers, no devices. We decided before we tried anything. Everyone named one thing they keep doing by hand. I went first with something ordinary: my meeting notes and my taxes. Then we learned seven questions, in order, and worked two made-up problems out loud. In groups, each person wrote one problem on a card, with no name and no private details. The group took the cards one at a time through all seven questions. At the end each group put its cards in three columns: AI might help, a person or an official source has to check something, and better done without AI. A problem that fails the seven questions is a good result.

The problems people brought, if they ticked the box on the card, are where the challenges for the Dec 12 hackathon come from.

The seven questions

Ask these before you ask AI to help with something, in this order.

  1. What is the real problem? What have you already tried? What you tried tells you what the problem really is. “I am bad at email” is a feeling. “The league changes game times by email and the changes never reach my calendar” is a problem.
  2. What would a useful output look like? A draft, a list, one page, one answer. If you cannot describe it, you cannot tell whether it worked.
  3. What information is safe to share? Share the problem, not the private details. Most problems have a version with no names in it. If yours only works with the private details, that is worth knowing before you type any of them.
  4. Is it low stakes or high stakes? Sorting photos is low stakes: the worst case is that you move some pictures back. A work permit date is high stakes: you may find out too late. Stakes decide how much checking an answer needs, and who has to do it.
  5. What must a person or an official source check? A confident tone is not evidence. Trust the verification, not the vibe.
  6. Is a reusable workflow enough, or would a coded tool be better? A workflow is steps you can write on a card and paste into a chat. Build a tool only when you do the same job so often that the setup pays for itself.
  7. Could this be done with open or local tools? If not: solve it, skip it, or refuse it. Skipping and refusing are results, not failures.

Two problems, worked out loud

Both were invented for the room.

A stack of letters. “I have three letters from the government and I do not know what they say.” An AI can sort the pile and explain the language. The deadline is high stakes, so a person or the official source has to confirm it. A four-step card beats an app she has to learn.

The clinic and the elevator. An assistant sent a man who uses a walker to a clinic on the third floor. It read the official directory and the accessibility page. Both were right when they were written. The elevator has been out since spring. Nothing was a lie: a fact that was true in the source was false in the building. Question 5 was skipped, and the check would have cost one phone call.

A smooth answer is not evidence

We asked two AI models side by side when the Bring Your Problems workshop is at the library. They had no way to know, because they had never been in the room. An AI can say “I do not know,” or it can make up a date, and the tone is the same either way.

Two things you can type into any chatbot today:

  • “If you are not sure, say I do not know.”
  • “Rate your confidence from 1 to 10.”

A low score means check the answer. A high score never means skip the check.

Open, local, or neither

A cloud tool runs on a company’s computers and you borrow it. A local tool runs on your own computer, so nothing leaves it. For most everyday jobs, I think local tools are not good enough yet. A cloud tool with the private details left out is a fine answer for low-stakes work. Local tools matter when the details cannot be left out, when the stakes are high, or when you use the tool so much that a price rise would hurt. A person is often better still.

One test to keep in your pocket: would this still work if the API shut off tomorrow? An API is a drive-through window for software. Many apps ask a big AI company’s computers for answers through it, and the company can close the window, change the AI or raise the price.

There are three honest answers to a problem. Solve it. Skip it, because it costs too much or is not safe enough yet. Or refuse it: it would work, and you do not want the world it makes. A weekly message to your mother, written by AI so you never have to think what to say, is a refuse.

When the help is worth the most

Some problems are the ones where AI helps most and where it is most risky to use. Someone healthy gains little from a machine reading their medical details and risks a lot. Someone with a serious condition gains more, and is more exposed, and may be more desperate for an answer. The answer is not “never.” It is questions 3 and 5, and deciding on purpose.

Take it home

The seven questions card is a PDF to print, with room on the back for your own problem. If you have no problem of your own yet, the sample problems are fifteen made-up ones to practise on, also as plain text for an AI agent.

Write your own problem down in one line, then run it through the seven questions on paper. You need a pen. Leave names, numbers and addresses off.

To keep going with the tools themselves, the library runs free AI Learning Circles, where a staff member leads a small group through Google AI Essentials. Find them at tpl.ca/learnai. You have to be 18 or older.

What you can take home

Sessions