Technology,
at a human
pace.

A useful folder system starts with a question: where would you look for this next month? The answer is often a project, rather than a file type. A proposal, a spreadsheet, and a set of photographs can belong together if they support the same piece of work.

1. Create an inbox for files you have not sorted yet. Treat it as a temporary stop, rather than an archive.
2. Keep active projects in their own folders. Move completed projects into an archive without changing the structure inside them.
3. Use names that explain the contents. A pattern such as “2026-09-26 — workshop plan — v02” can make versions easier to recognise.

Set aside a few minutes each week to empty the inbox. Delete only what you recognise and no longer need. Where possible, link to a shared original instead of making another copy that may drift out of date.

Start with one active project. A system you can maintain is more useful than a perfect archive you never finish.

Try sorting notifications by when they need your attention: now, later today, or the next time you open the app. Calls from a family member may belong in the first group. Recommendations and reactions will often fit in the last.

Change one app at a time. Begin with sounds and pop-up banners, then live with the change for a few days. If something important becomes harder to notice, adjust the settings again. A quieter phone should fit your responsibilities.

For shared work, agree on a channel for urgent questions and a reasonable response window for everything else. This makes focused time easier to protect because silence no longer has to mean uncertainty.

Before using a long focus session, check that essential contacts can still reach you. The aim is a deliberate rhythm: fewer interruptions, with a reliable way back to the people who need you.

Setting up a new service can feel like progress before any real work has happened. Start by describing the friction you want to remove in one sentence. “Find decisions from previous meetings” is a more useful target than “get organised”.

1. What recurring task should become easier? Choose something you can observe.
2. Could a familiar tool already do it? A simple template may be enough.
3. Can you export your work in a usable format? Try this before moving an archive.
4. Does the tool still feel helpful after a week? Include search, corrections, and handover in your trial.

Decide what success would look like before the experiment. You might spend less time preparing a meeting, or find a document without opening five folders. If there is no useful difference, returning to your earlier approach is a reasonable outcome.

Rough notes can mix observations, decisions, and unanswered questions. An AI assistant may help turn them into a first pass at a structure. Keep the task narrow so that you can compare the output with the original.

For a low-risk trial, use a short set of your own non-sensitive notes. Remove private details and anything you do not have permission to share. Ask for explicit gaps instead of allowing the assistant to fill them in.

Organise these notes into decisions, open questions, and possible next actions. Use only the supplied text. Do not invent owners or deadlines. Mark missing information as “not specified” and include the supporting note for each decision.

Read the proposed decisions first. Check that an idea discussed in passing has not become an agreement. Then verify names, numbers, and dates. If a task has no owner, keep it unassigned until a person confirms it.

The useful output is a reviewed draft you can work from. Keep the original notes alongside it so that the context remains available.

A vague request leaves many choices to the model: who the reader is, what they already know, and which details deserve attention. A small brief makes those choices explicit and gives you a clearer way to review the draft.

Draft a 150-word introduction for a guide to organising personal files. The reader is comfortable with basic computer use. Use a friendly, direct tone and one concrete example. Avoid product recommendations, invented statistics, and claims about guaranteed time savings.

Add the facts the draft must preserve. If a date, price, or feature has not been confirmed, ask for a placeholder instead of a guess. Where the text relies on external information, check the original source before using the claim.

Review for meaning before polishing the tone. Does the draft say what you intended? Has it introduced a promise you cannot make? Replace generic language with your own example, and read the final version as someone encountering the subject for the first time.

For an important message, keep the final decision and sending step with a person. A polished sentence still needs an accountable author.

Consider a personal reading list. An assistant could suggest topic labels for article descriptions you provide. This is a small enough task to inspect: the input is clear, the result is visible, and a wrong label is easy to change.

1. Choose a few allowed labels, such as design, tools, and habits. Include an “uncertain” option.
2. Try ten examples manually. Check whether the labels actually help you find something later.
3. Record the errors. Look for ambiguous descriptions and overlapping categories.
4. Only consider automation when the manual trial produces useful results consistently.

Use ordinary rules where they are sufficient. Sorting a filename by its date does not need a language model. AI is more worth exploring when the task involves varied wording, and the possible mistakes are manageable.

Keep originals, review suggestions before applying them, and avoid automatic deletion or publication. Compare the time saved with the time spent checking. The result may be a useful workflow, a smaller one, or a decision to keep doing the task yourself.