More tools do not make unclear people clearer

The sentence comes from one of our sessions and it drew objections there. Fair enough, because at first it sounds like scepticism about technology. It is not. I automate wherever I can and have scripts written for me that used to cost me two evenings.

The point is a different one, and it is arithmetical.

The costs shift, they do not disappear

Writing a decent email used to cost ten minutes. Today it costs thirty seconds, because a model does the phrasing.

Reading that email, understanding it, deciding what it means for me and phrasing a next action out of it costs exactly as much as it did three years ago.

That is the whole mechanism. Costs on the sending side have fallen, costs on the receiving side have not. And because whatever gets cheaper gets done more, the volume rises. Your inbox does not get lighter because you use AI, it gets heavier because everyone else does too.

Anyone who had no process for turning input into decisions now gets the same problem at a higher frequency.

The summary fallacy

A second effect is subtler and therefore more dangerous.

When an AI condenses a mail thread with fourteen messages into five sentences, it feels like processing. But it is only reading, in short. The actual work is not knowing the content, it is deciding what it means for my commitments. Whether I have to do something. What exactly. By when. Whether I want to at all.

That decision stays entirely with me. It just feels done after the summary, because the sense of having an overview is there. That is why people with AI assistance have often read more and clarified less.

What AI can genuinely do in self-management

There is a clear line, and on one side of it the benefit is considerable.

Capturing. A voice note while out and about, automatically transcribed into the inbox. The hurdle between thought and capture is where GTD most often breaks in practice, and that hurdle can be lowered technically.

Extracting raw material. Pulling everything that sounds like a commitment out of a long thread or a meeting note. Important: the result lands in the inbox, not on a project list. It is material, not a decision.

Preparing the weekly review. Tasks completed this week, calendar entries, projects without a next action, waiting-for entries older than ten days. A script can gather that, and you start the review thinking rather than searching.

A sparring partner for thinking things through. In brainstorming on a stuck project, quantity is what counts and evaluation comes later. That is exactly where a model is useful, because it gives you twenty approaches of which three are any good. The selection stays with you, but the selection was never the expensive part.

Opening up reference material. Making what you have filed searchable. Unspectacular, but it genuinely saves time.

And on the other side of the line stands exactly one thing that no model takes over.

Only a person can make a commitment

GTD does not manage tasks, it manages commitments. The reason unfinished things circle in your head is not that they are unfinished but that you committed to them without deciding what becomes of them.

That decision cannot be delegated, not even to a very good model. Whether a project is worth it. Whether you withdraw a commitment rather than postpone it. Whether you tell a colleague you will not manage. That is the heart of the matter, and it consists of decisions that have consequences for you.

A model can help you make them better prepared. Making them is yours.

Commitment has a price, and software has none

Ignoring a push notification costs nothing. Standing up a running partner who is outside your door at six in the rain costs something.

This is not a motivation topic but plain economics. Intentions hold when breaking them has a cost, and social costs are the most effective ones available in daily life. That is why study groups, training partners and weekly sessions with the same people work, while reminder apps get swiped away after eleven days.

The same applies to habits in self-management. The weekly review you plan to do alone on a Friday gets skipped. The weekly review three colleagues meet for at one o’clock on a call happens.

Self-management has long been a team sport

The larger part of what is on your lists depends on other people. You are waiting for an approval, you need information, you delegated, somebody delegates to you.

That makes the quality of your system more than your problem. If you have precisely phrased next actions on your list and your team passes around topics without actions, you spend your time clarifying on behalf of others.

This is exactly the point at which individual work hits a limit. One person with a good system inside a team without shared vocabulary is well organised and still blocked constantly.

What actually changes with a shared language

Culture is a soft term, so let us be concrete. When a team uses the same terms, four things change visibly:

The question “what is the next action and who does it” gets asked at the end of every meeting, without anybody having to facilitate it.

Mails contain an action in the text, not just a topic. The difference between “FYI” and “I need your approval by Thursday” is made explicit rather than guessed.

Delegated tasks are tracked by both sides, once as a task and once as a waiting-for. That ends chasing out of nowhere.

And there is a difference between “I’ve seen it” and “I’ve clarified it” that can be spoken out loud without sounding like an excuse.

Internal trainers, and when they achieve nothing

More and more companies are training their own people rather than buying trainings in. The reason is not price but context. Someone who knows the internal processes, the tool landscape and the quirks of how people communicate can bring examples that are true for the people in the room. Nobody from outside has that context on day one, and an AI does not have it at all.

Two qualifications belong with that, otherwise it sounds like a sales pitch.

Without a time budget an internal trainer is a job title. Give somebody the role on top of a full-time job and they will run two workshops and then never another. The role needs hours in the calendar, otherwise it belongs in the category of well-meant.

Without leadership nothing happens. If tasks are handed out by chat at eleven at night with an expectation of immediate replies, no shared vocabulary will help. People then learn a language in which they can describe why their system does not work. That is worse than before.

What holds up in practice is a small recurring session. With us that is a weekly one; in companies half an hour every two weeks with the same people works. Not as a class but as a place where you can say what did not work this week. Same mechanism as the running partner.

Common questions

Can AI take over my self-management?

Partly. Capturing, pre-sorting, extracting raw material and preparing the weekly review can all be handed over well. The decision about whether you take on a commitment, keep it or withdraw it stays with you, because it has consequences for you.

Why does AI make the flood of information bigger rather than smaller?

Because AI lowers the cost of creating, not of processing. Writing a mail takes seconds; clarifying it takes as long as ever. The load shifts from sender to recipient.

Why is an AI assistant not enough for new habits?

Because ignoring a notification has no cost. Commitment comes from social costs, meaning from people who notice. That is why fixed sessions with the same people work better than reminder apps.

What does an internal trainer bring to a company?

Shared terms, examples from their own context, and a recurring session that keeps habits alive. The prerequisites are a real time budget for the role and a leadership team that follows the rules itself.

---

If you use AI in your self-management, look at where it makes decisions easier and where it merely gives you the feeling of having decided. The difference lies in whether something ends up on a list afterwards.

Where the line runs in your organisation depends on your processes and your works agreement. I clarify that under automation and AI. Write to me.