Master a new subject your role demands
A subject that is not yours has landed on your desk, and in two weeks you will have to answer for it. This course shows what to do in those two weeks: what to ask the AI tool for, what you have to read yourself, and how to know you really understood.
- 01 · You are here
- 02 · Absorb documentation
- 03 · Prepare a decision
- 04 · Cut task time
Build a study routine with four steps, knowing which step the AI tool helps with and which step it gets in the way of.
Having the summary is not mastering the subject
Asking for a summary is the first thing almost everyone does, and it works: in a few minutes you have an organised view of a subject you did not know. The problem shows up later. In the meeting, someone asks why the conclusion is that one, and you find out you have the text in hand but not the reasoning that led to it.
At work, mastering a subject means three quite concrete things:
- explaining how the thing works in your own words;
- saying where you checked each piece of information;
- saying what you still do not know.
A summary gives you the appearance of all three and none of them in fact.
Rereading gives you confidence. Trying to remember is what sticks
This is not our opinion. Two memory researchers, Henry Roediger and Jeffrey Karpicke, ran the test. They gave the same texts to groups of students and asked for different things: one group reread the material; the other closed the text and tried to write down from memory what they remembered, with nobody correcting them.
Then they measured how much each group remembered. Five minutes later, the rereaders did better. Two days later and one week later, the ones who had tried to remember did much better.
And there is one detail that matters directly to anyone using AI: the students who reread were more confident in their own memory, while remembering less.
Reading an AI-written summary is the most comfortable form of rereading: the text arrives organised, you read it fast, and the feeling of having understood is excellent. Which means it is exactly the situation that raises your confidence the most and makes the content stick the least. That is not a reason to stop using the tool. It is a reason not to stop there.
Four steps, in this order
The routine below separates what the tool does well from what only you can do. Each step ends in something written down. If that thing does not exist, the step was not done.
| Step | What you ask AI for | What stays yours |
|---|---|---|
| 1. Define the question One sentence, written by hand | Nothing yet. This step is yours alone. | Deciding what the study is for: which decision it has to support. Without that, you try to study the whole subject and never finish. |
| 2. See the map Understand the terrain and the vocabulary | An overview, the technical terms you will run into, the points where people disagree, and where the original documents for each point are. | Refusing any number, rule or date that arrives without a link or document you can open. Here the tool is there to orient you, not to assert. |
| 3. Read the original document Only the part that decides | Finding the right passage, translating a hard sentence, explaining a technical term while you read. | Reading. This step has no shortcut, and it is shorter than it looks: usually a few pages carry the decision. |
| 4. Explain from memory Write with nothing open | After you have written it, ask the tool to disagree: where your text is weak, what a sceptical person would ask. | Writing the explanation from memory, with everything closed. This is where you actually learn — or find out that you did not. |
The study is over when you can explain the subject in one page, with nothing open, saying where you checked each piece of information and what is still unanswered. It is not over when you finish reading.
The step that usually gets skipped
In practice, step 4 is the first one people skip. It is the only one that takes effort and produces nothing to show anybody.
The report nobody could defend
An analyst has to assess a new government requirement that affects a process in her area. In two days she delivers a well-written document, well organised and with a firm conclusion, assembled from summaries and passages the tool quoted.
In the meeting, nobody questions the conclusion. The question is about an unusual case the document did not anticipate. To answer it, she would have to know why the rule exists — and that was in the original document, which nobody opened. The meeting ends with a request for another round of study.
Rereading won at five minutes and lost at one week
Here is the result organised by time. It is this table that decides how you should study.
Which condition wins, at each interval
Same texts, two ways of studying: reread it, or close the material and try to write it from memory. As time passes, the winner changes.
Why is rereading misleading?
What is the stopping criterion for the study?
At which step of the routine does AI get in the way?
This block applies to the course the method the course teaches: try to answer from memory before opening each answer. If you open without trying, you are just rereading.
Where the argument comes from
- Henry L. Roediger III and Jeffrey D. Karpicke, “Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention”, Psychological Science, vol. 17, no. 3, March 2006, pp. 249–255. How the test was run, according to the paper itself: two experiments with ordinary texts; one group tried to write from memory, once or three times, without correction, and the other reread the material the same number of times; the final measurement was taken five minutes, two days or one week later. What the paper found: at five minutes rereading was better; at the longer intervals the students who had tried to remember remembered considerably more, even though rereading left students more confident in their own memory. Verified 6 Aug 2026: authorship, journal, volume, pages, DOI and abstract wording confirmed on PubMed, Sage and the Washington University repository.
The next course
Step 3 asks you to read the original document. Course 02 covers the hard case: when that document is long, technical or full of rules, and you need to know where each thing the tool pulled out came from.