GTD + AI: The Two Slowest Steps Just Got 10x Faster
Clarify and organise are where GTD quietly dies. Those are exactly the two steps a model is good at.
Getting Things Done is one of the most complete productivity systems ever written down. It is also one of the most expensive to maintain: an active knowledge worker can spend one to two hours on the weekly review alone.
Inbox processing is where most practitioners fall off. AI doesn't change the system — it removes friction from its most labour-intensive steps. Using AI here isn't abandoning the methodology; it's the first thing in twenty years that makes it cheaper to run.
Where AI helps across the five steps
Capture: barely at all. Fast capture is already a solved problem, and adding a model to it just adds latency.
Clarify and organise: substantially. “Is this actionable? What project does it belong to? What's the next action?” are exactly the questions a model answers well from a short description.
Reflect: usefully, by surfacing stalled projects and missing next actions. Engage: not at all — doing the work stays yours.
AI as a clarification partner
A captured line like “figure out the new office situation” isn't clarified. It's a blob with a verb in it.
Run the slash command and you get something to react to: this is a project, call it Office Setup, and the next actions are research co-working spaces nearby, calculate the current commute cost, book visits to the top three. Ten minutes of mental effort compressed into thirty seconds — most of which you'll keep.
AI doesn't change GTD. It makes the two steps everyone skips cheap enough to actually do.
Next actions on demand
The hardest discipline in GTD is making sure every active project has a concrete next action. It's also the one that decays fastest under pressure.
“Hire a senior engineer” expands to: review the job description, post to three boards, list five companies to source from, book the recruiter call. Not perfect, but far better than an empty next-action field and a vague sense of dread.
The weekly review, accelerated
The review is where the accumulated debris of the week gets processed, and it's the ritual people skip first.
AI can flag projects with no next action, surface tasks that haven't moved in thirty days, and point out items that look like duplicates of each other. You still make the calls; you just stop doing the scanning by hand.
Building the initial hierarchy
New practitioners face a specific version of the blank page: what are my areas of responsibility?
Describe your life in one sentence — “software engineer, manage a small team, two kids, trying to get healthier” — and you get Work (IC and management), Personal Development, Family, Health as top-level areas. Nest projects and next actions underneath, with the same help on decomposition where you need it.
Takeaways
- Clarify and organise are the highest-leverage steps for AI.
- Reacting to a suggested next action beats staring at an empty field.
- Let AI scan for stalled projects during the weekly review.
- Generating the initial area hierarchy removes the setup barrier.
Taskpend is a nested outliner with mirrored tasks, six views and AI agents that execute. Free for individuals.