The methods are still right. They just answer a different question.
The first step in any task system is the same as it has always been: stop using your brain as the main storage place for the list. A task that is written down stops being a cognitive burden — it frees up mental resource, lowers the risk of forgetting and removes the permanent internal reminder.
Then comes the organizing layer: tags that separate personal from professional, or that distinguish projects, processes and deliverables. Except that sophistication charges a price: every extra level of structure is more time spent organizing instead of doing, and finding the balance between detail and efficiency is the whole technique.
For prioritizing, the reference is still the Eisenhower Matrix, named after Dwight D. Eisenhower, who in a 1954 speech, quoting a university president, said: "I have two kinds of problems, the urgent and the important. The urgent are not important, and the important are never urgent." Stephen Covey, author of The 7 Habits of Highly Effective People, turned the line into the four-quadrant tool: do (urgent and important), schedule (important, not urgent), delegate (urgent, not very important) and eliminate (neither one nor the other).
One caution almost everyone skips is worth keeping: a task that looks trivial or not urgent can have substantial repercussions if it is neglected. Paying an electricity bill with a future due date is neither important nor urgent until the day it is.
There is also the Lazy Susan method — an analogy to the rotating serving tray — which prioritizes by current interest, by physical and mental energy, by emotional state and by readiness. Rotating types of task softens exhaustion and gives the brain time to recover. Depending on each person's schedule, that rotation makes it easier to reach states of Deep Work, a concept from Cal Newport, or of Flow, popularized by Mihaly Csikszentmihalyi.
Notice what they all have in common. They order, group and spread over time the work you are going to do. The only quadrant that talks about someone else is "delegate" — and delegating has always presupposed someone to delegate to. On most lists, that someone did not exist.
A task is not a deliverable, and that matters more now
- Deliverable
- the final product or service that results from your team's work. "Customer satisfaction survey conducted", "App prototype created".
- Task
- the action needed to produce that deliverable. "Draft the questionnaire", "code the prototype screen".
The test is short: if you cross an item off the list and nobody on the outside notices a difference, it is probably a task, not a deliverable.
The confusion is expensive in both directions. A task list with no deliverable produces motion with no visible result. A deliverable with no task produces a promise nobody can measure.
In Y Managers this distinction is structural, not advice. A deliverable's progress is computed, never typed: it is the average of the tasks' progress, weighted by each one's weight — a task with no weight counts as weight 1; a task with weight 0 is left out of the calculation on purpose. And a work plan that points at deliverables but at no task gets a red tag, "no tasks", with the explanation that pointing at the deliverable is not enough.
For a person that is a warning, and they carry on anyway. For an AI worker, it blocks the signature of the plan. This is where the task stops being a detail of personal organization and becomes the condition for an agent to be held to account: with no task pointed at, there is nothing to hold it to.
The new question: which of these should a machine be holding
Orion reads the description of your team's active tasks and sorts each one into three destinations.
- AI does the whole thing
- send an email or a message, book a meeting, create a form, extract, structure or summarize a text, set up a reminder.
- AI does the draft
- write a document, report or plan, analyze data, create content, research options — always with a human reviewing and approving.
- Human only
- a decision, an approval, a signature, on-site work, human production, and the technical work of your company's professional domain — an engineering design, a legal filing, a medical report.
When the description is too vague, the task falls into "human only". It never overestimates automation.
From there it makes an offer, and the offer comes split into two lists: what it can get moving now and what it needs an input from you for. You answer which item you want, or "all of them". Two deliberate locks: it only makes the offer when there is at least one thing it can really move forward — there is no message that just asks for a document — and at most one offer per week.
The reverse works too. Ask "what on this list can you do for me?" and it points out the tasks that are a case for AI and offers to execute or draft them.
| Destination | What the machine delivers | What stays with you |
|---|---|---|
| AI does the whole thing | the action executed | authorizing and checking the result |
| AI does the draft | text or analysis ready for review | judging, correcting and approving |
| Human only | nothing | everything |
Classification is the management decision that is left to you
Notice what happened to the "delegate" quadrant: it stopped being theoretical for people with no team, and gained one more destination for those who have one.
The work changed in nature. It is no longer ordering thirty items by urgency. It is running your eyes down the list and deciding three things: what leaves your hands entirely, what comes back as a draft for your judgment, and what should never have left.
That decision is managerial, not operational. Classifying too far up costs rework, and a badly done review costs more than having done the work yourself. Classifying too far down costs you your time — you keep work on your desk that did not require your judgment.
And the classifier does not decide for you: the proposal is its, the call is yours. A task can be technically "AI does the whole thing" and still be yours, because whoever receives that message needs to know it was you who wrote it. No classifier sees that.
A rhythm that lasts, not a sprint that burns you out
The force behind the discipline of keeping up any of these methods is the satisfaction of completing tasks on a list. But, despite the continuous completion, the constant inflow of new tasks produces the sensation of not moving forward — a cycle of frustration and demotivation.
What sustains it is the constant recognition of progress. Assessing daily the quantity and the quality of what was accomplished works as an incentive, a recognition of the effort.
The same technique shows up among long-distance runners. A study published in the journal Motivation and Emotion found that, instead of concentrating on the finish line, focusing on a set point ahead — a building, a tree — makes the distances feel shorter, which leads athletes to move faster and reduces the sensation of effort.
And here comes the trap. Setting a daily or monthly target for the number of tasks completed is not a prerequisite for success. Because complexity and duration vary so much from one task to another, that metric can be misleading.
That is why the product counts cycles, not tasks. When you build a work plan you choose the feedback frequency — weekly, biweekly, monthly, or a single cycle covering the whole period — and that is where the duration of the cycle comes from. What you write in "What was done", on each deliverable, is what becomes the cycle report — there is no separate form.
The deadlines are short on purpose: five days for the report after the cycle ends, ten for the manager's feedback, five to request a review. With weekly cycles, the goal is for each lap to close before the next one ends — feedback that arrives months later no longer changes anything.
And work that repeats has its own treatment. A deliverable marked as recurring is born with a cycle clock on. Within the cycle, each step is worth a slice proportional to its weight; at the end of the lap, the real percentage is registered in the history and the steps reset. That is why a process's progress does not leak from one month into the next. A process never reaches 100% — and pretending it does is exactly what turns constant work into a permanent feeling of debt.
What this does not solve
The classifier reads descriptions, not reality. It decides from the text someone wrote in the task. A vague description becomes "human only", which means the tool errs on the side of giving you back more work, not less. It is the right error to make, but it is an error — and its input is your writing.
It does not know what matters. Urgency and importance still come from you. The quadrants have not been automated, and the electricity bill with a future due date is still exactly the kind of thing no classifier catches.
"AI does the draft" is not "done". By definition, there is a human reviewing and approving. If you approve without reading, that task's real destination became "AI does the whole thing" — and the one who decided that was you, not the system.
Very technical teams will see almost everything as human. If your team's work is engineering, legal or healthcare, the classification will come out "human only" on almost everything and the offer to get things moving never arrives. That is the correct behavior, not a failure: it would rather stay quiet than offer to do what it does not know how to do.
The count of automatable tasks is a snapshot of the moment. It appears inside each report, but it was left out of the maturity series on purpose: it is only known at today's value, so two periods would display the same number and the comparison would be false.
A work task and a personal list are not the same thing. A work task belongs to a deliverable. If you ask for something without mentioning any deliverable, it becomes an item on your personal list — when in doubt it asks, but it is worth knowing the rule.
There is no method that works for everyone
That is still true. Managing tasks is a dynamic process, one that requires continuous adjustment according to each person's needs and circumstances, and the arrival of AI has not changed that part.
What changed was one thing only: part of your list now has somewhere to go. That does not come for free — deciding where each item goes is management work, and the judgment is still yours, including about where the machine got the count wrong.
The list will keep growing. The good question is no longer how many items you crossed off today. It is how many of the items still in your hands really needed to be there.