Workshop · W05

Get Things Done with AI: Build Your Own Productivity System

  • Wednesday, October 28, 2026 6:00 pm to 8:00 pm
  • Monday, November 23, 2026 4:00 pm to 6:00 pm

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

Moderator: Ryan Kelln, Innovator in Residence, Applied AI

What if your to-do list could notice what has been stuck for weeks?

A personal productivity system built around an AI agent and a weekly review. It loads your goals, counts what actually got done, flags what keeps slipping, and drafts next week’s plan. Support without cheerleading. Try it with a practice set of goals and leave with the plain-text templates.

What we did

We used a made-up person, Dana, who has a part-time job and an evening course. Her folder holds her goals, eight weeks of lists and a work log. We asked an AI for help with one stuck item, ran her weekly review, changed how the AI talks to her, and fitted next week to the hours she really had. Then everyone filled in a paper sheet about their own week. Nothing personal went into a computer.

The review, in four jobs and an optional fifth

A weekly review is one conversation with four jobs:

  1. Count what got done, and celebrate it. Say which things, not “a good week”.
  2. Ask why the rest didn’t happen. Too big, too boring, waiting on someone, or something else?
  3. Change the list or the habit, so next week can work.
  4. Fit next week to the hours you really have, and break big things into pieces that fit.

And an optional fifth job: add your must-haves. These are requirements for certain kinds of tasks, so next week has them on it. In the club, everyone had something fun and something health related. Then some people added their own, such as time with family or friends, or the chores that keep a home running.

Take it home

The weekly sheet and take-home card (PDF) is the paper from the workshop. The front is the sheet you fill in. The back is a card with a first review you can do on paper, the four jobs as a checklist, and the questions to ask before you give any AI your week.

You can do all of this with a pen and no AI. An AI helps most when there is already a list and a log for it to read.

Four places to keep your system

  1. Paper or a text file. A list and a log you keep yourself. Free, and it outlasts any app. Ask yourself: will I open it each week?
  2. An app with a review built in. We used Mindwtr. It is free and open source, works on Windows, Mac, Linux, iPhone and Android, works offline and needs no account. It has a guided weekly review screen. You can sync it between devices if you want to. You do not need to. It is a young project and it changes about every week, so read what a new version does before you update. Its guides are at docs.mindwtr.app.
  3. An agent that reads your files. An AI with a few notes on how to work with you, tools to read your list, and a log of what happened. In the workshop it was Claude in the Claude desktop app, which needs an account and may cost money. Open source agents work with any AI you choose, and take more effort to set up. I use and recommend Hermes Agent. OpenClaw and OpenCode are others.
  4. A paid assistant that checks your apps on its own and nudges you. The big AI companies all sell one now. Prices change, so check before you pay. Three questions first: what can it read, what can it do without asking, and what happens to my list if I stop paying?

They stack. Start at the top and move down only when the one above stops being enough.

The helper skill

The AI in the workshop followed a skill: a set of plain text instructions it reads when a job starts. This one has two jobs, getting unstuck on one thing and the weekly review. You can read it and change it.

It expects three things in a folder: goals.md (your goals and your free hours), a weeks/ folder with one file per week named for its Monday, and worklog.md (a short dated note of what you really did). Dana’s files had that shape. The skill reads only those files and writes only one new file, review.md.

If you try it with an AI, it is a starting point and not a finished product. Make it yours:

  • Read what it says. Pick one thing, like a date or “you finished this”, and find it in your own files. It may be a little wrong. Finding that is the most useful part.
  • Change the line about how it talks to you. These AIs lean toward agreeing with you, so an instruction like “be tough on me” only nudges it.
  • After each review, ask: how could the skill be better, given how this went?

Habits from the GSD club

For years I ran a weekly review with a small group of people. These are the rules we kept, because the reviews went wrong without them.

  • Write down what you did before you look at the list. Then compare. The gap is the review.
  • Trust the log over the boxes. A week can look empty on the list while a lot got done, and a ticked box can be wrong.
  • Size the week to your hours. Add them up and leave about a fifth spare. Nothing on a day off.
  • Park things instead of piling them up. A long list makes every review feel like failure. Keep parked things as plain bullets, not boxes, and never call a parked thing a failure.
  • A date is not a task. Anything due in the next two weeks gets its own task with a time estimate.
  • Break big things into pieces that each fit one slot. Make the first piece ten minutes or less.
  • Change the plan, not the person. If a habit keeps missing, ask what is in the way of that slot.
  • Always something fun, and something healthy. Add your own must-haves on top of those.
  • A skipped review is a restart. Look at the log and begin again.
  • Have someone to answer to. The two biggest reasons people liked the review were accountability and time set aside to think. A friend works. An AI works if you make it insistent. When I swapped most of my human reviews for an AI I missed my first week in fifteen years, so I still meet people once a month.

Be careful with what you type

Dana is made up, so nothing in the workshop could hurt anyone. If you use an AI with your own list, goals or log, that text goes to the AI company. Some keep it. Read the provider’s terms first, leave out anything you would not want kept, and decide what the AI may never touch, such as your email or your bank. That is line 4 on the paper sheet.

What you can take home

Sessions