Get started in one week

There is no need to build everything at once. Here is the order that works, one step per day.

Validation loop

Figure. The principle behind the whole method: the AI prepares, you validate.

Day 1: safeguards

Write your rules in Claude’s preferences: never overwrite, never delete, never send, never write to the calendar without approval, no identifiable patient data. → Setting safeguards

Day 2: a root folder

Create a single folder for the AI, with your work folders laid out flat and three technical folders. → Organising your files

Day 3: three projects

One Claude project per mission: Care & department, Teaching, Research. In each one, add the reference documents (slide template, department rules, recommendations of your CNU subsection).

Day 4: a first high-yield use

Pick the one that saves you the most time this week:

Day 5: anticipation

List your deadlines for the next three months and build a first anticipation dashboard. → The Reactor

Next

When an instruction comes up three times, turn it into a skill (Skills and hats). When a check comes up every day, turn it into a routine (Scheduled routines).

The tools used

Tool What it is used for here
Claude (claude.ai, app) Conversations, one project per mission, artifacts
Projects Keeping a mission’s context (reference documents, instructions)
Artifacts Interactive pages: anticipation dashboard, email triage, validation pages
Skills Reusable know-how (“review this thesis”, “build this case”)
Connectors Email, calendar, Drive, PubMed, read-only by default
Cowork / Claude Code Working on the files on your computer (filing, imaging, databases)
Scheduled tasks The morning, evening, weekly and monthly runs

Tool names and features change: refer to Anthropic’s documentation.