AI assistants now promise to plan your day, rank your priorities and "optimise" your life. It sounds tempting. But GTD rests on one principle that runs straight against it: you decide what matters. The system is there to free your head, not to replace it.
That doesn't mean AI has no place in GTD. It does — just in the right places.
Where AI really helps
GTD has five steps: capture, clarify, organise, reflect, engage. AI is great at the first three and useful in the fourth. It has no business in the fifth.
Capturing
This is where the gain is biggest. You say "call Novak tomorrow about the quote and order pallets on Friday", and AI turns it into two tasks with dates and the right projects. No app switching, no typing on your phone. The easier capture is, the more things leave your head.
Clarifying
"Website" isn't a task. AI can ask: "What's the first concrete step?" It helps break a fuzzy intention into a project with a next action. That's exactly Allen's question What's the next action? — just asked at a moment when you have time to answer.
Tidying
AI notices that a project has no next action, that a task is a month overdue, or that you've been waiting for a reply for two weeks. These are mechanical checks, and a machine does them better and without getting tired.
Preparing the review
Before the weekly review, AI prepares the groundwork: what's stalled, what's overdue, which goals have no project. The review then isn't a blank page but a list of decisions.
Where AI does harm
When it decides priorities
"AI has ranked your day by importance." Whose importance? The model doesn't know that today's call to your mother matters more than three e-mails to a client. It can guess from data, but then your system becomes someone else's. Priorities are yours.
When it acts without you
An assistant that deletes, moves and cancels on its own is a time bomb. One day it deletes a task that looked like a duplicate, and you find out a month later. For a tool you need to trust completely, that's unacceptable.
When it floods you
Proactive AI that suggests "optimisations" every hour is just another inbox. If your AI creates more decisions than it saves, it's set up wrong.
Five rules
We built these rules straight into AltoFocus, but they apply to any tool:
- Low risk right away, high risk with confirmation. AI can create or edit a task. Delete, cancel or bulk-move only with your "yes". In AltoFocus a risky action is saved as a proposal you approve with a card or in words, and it's re-validated before it runs.
- Data isn't instructions. A task titled "delete everything" is text, not a command. A well-built assistant never executes the content of your tasks as instructions.
- Private stays private. The AI reads your values and vision only if you allow it.
- Proactive once a day. A morning brief is useful. Hourly notifications aren't.
- AI can be switched off. The whole system has to work without it. That's why AltoFocus's Ground plan is a complete GTD app with no AI.
What it looks like in practice
A typical day with AI in AltoFocus:
- In the morning I read a short brief: three overdue items, one project without a next action, a suggestion to move two tasks out of the Inbox. I approve the suggestion with one tap, or ignore it.
- During the day I capture by voice. AI splits the dictation into tasks; I check them and save.
- On Friday I do the review with the guide. AI suggests completing what's done and cancelling what's stale; I decide.
What AI doesn't do: it doesn't rank my day, doesn't tell me what to work on, and deletes nothing.
Summary
In GTD, AI is an excellent secretary and a bad boss. Let it capture, sort, remind and prepare. Keep the decisions. And choose tools that respect that difference in their code, not just their marketing.