Getting Things Done — GTD, from David Allen’s 2001 book — is built on one idea: your head is for having ideas, not holding them. Get every commitment out of your brain and into a trusted system, and you stop burning energy on remembering and start spending it on doing.
The five steps
GTD is a loop of five moves:
- Capture — collect everything that has your attention into an inbox, the moment it occurs to you.
- Clarify — process each item: is it actionable? What’s the next physical action? If it takes under two minutes, do it now.
- Organize — put actions where they belong: a project, a context, a “waiting for,” a someday/maybe list, or a calendar slot.
- Reflect — review your lists regularly (the weekly review is the heartbeat of GTD) so the system stays trusted.
- Engage — just do the work, confident the list is complete.
The friction in GTD has always been the middle steps. Capture is easy; clarify and organize are where most people’s systems quietly collapse, because turning “email Sarah re: the thing” into a properly filed, next-action-worded task is tedious.
Where an AI assistant fits
This is exactly the part a good AI to-do list can take off your hands. In Lume, you capture in plain language — “follow up with Sarah about the Q3 proposal Friday” — and the assistant does the clarify and organize steps: it titles the task as a next action, sets the date, and files it in the right project. Capture stays a one-liner; the filing happens for you.
For the weekly review, you can ask the assistant to surface what’s stalled — “what’s been sitting in Anytime for two weeks?” — instead of scrolling every list by hand. And for the rare GTD action that’s actually scoped work, you can hand it to an agent rather than do it yourself.
A lightweight GTD setup in Lume
- Use the Inbox for capture; clarify into projects during your review.
- Word every task as a next action (“draft”, “email”, “call”).
- Use Someday/Anytime for non-urgent maybes.
- Run a weekly review with the assistant to keep the system trusted.
GTD works because the system is complete enough to trust. The faster capture and filing are, the more complete it stays — which is the whole reason to put an assistant behind it. See the AI assistant or just start capturing.