The axis that matters: proactive vs reactive
Task breakdown is solved — every AI planner can chop a task into steps, and the honest review of that skill is in task breakdown. The unsolved problem is the same one that kills every ADHD system: who reminds you when you forget? A planner that waits for you to return is a documentation tool wearing an assistant costume — the same failure as the prompt-based ChatGPT pattern, scaled up. What separates the category is proactivity: resurfacing items, firing at start times, noticing silence, and tracking follow-through without streaks or guilt.
The honest category breakdown
General chat LLMs (ChatGPT, Claude, Gemini)
Excellent brainstormers and breakdowners, zero memory duties — they never text you back if you stop asking. The “you have to prompt it every time” pattern is covered in our ChatGPT guide.
AI task managers (neurolist-class)
Break large tasks into time-boxed steps with estimates, organised into a planner. A genuine upgrade on manual breakdown — but most are still pull-based: open the app, follow the list. Resurfacing depends on you.
AI scheduling/calendar agents (Reclaim, Motion class)
Auto-block focus time and re-plan around events — real proactivity on the calendar axis. Limits: they re-plan your day around what you give them, and the “giving” (keeping the calendar true) is your job; see the tools directory for friction scores.
Managed AI partners (the rare, genuinely proactive end)
A small and new end of the market: the agent lives in a channel you already use (chat), holds your loops, breaks them into steps, and reaches back out when something slips — on its own, with no streaks and no shame. This is the shape of the “partner, not assistant” claim, and it is the category to watch. If you are comparing it against chat-LLM setups, the difference is exactly the proactivity axis above.
Buying criteria
- Ask about the miss: “What happens when I stop checking in for a week?” A system with no answer to that is storage.
- Check the guilt mechanics: streaks, overdue badges, and red counters predict abandonment (planner guide).
- Match the layer to the failure: breakdown → any AI planner; remembering → proactive only; presence → body doubling.
- Prefer the channel you already live in — a tool that needs its own dashboard is one more dashboard to forget.
The verdict
AI planners help the moment they are used, and the best of them genuinely automate breakdown. The category’s frontier — and the claim to check before paying — is whether the system still works while you forget about it. That single axis separates a prompt machine from a partner in the load.