The deck is ready. Are you?
Back when I used a paper diary, I knew my day almost by heart. I knew which appointments were coming up that afternoon, often without opening it.
Then I switched to Lotus Notes. My appointments were saved, and reminders arrived on time. One day, I realised I could no longer remember what I had planned a few hours later without checking my calendar.
My calendar remembered. I remembered less.
I hadn't stopped having appointments. I had stopped needing to remember them. For appointments, that was perfect.
With AI, the same question returns. But it is no longer about appointments. It is about what we write, present, and defend.
Finding isn't understanding
I don't miss my paper diary. Not having to remember everything is a real help.
The internet extended that help. Why memorise something you can find in seconds?
In 2011, Betsy Sparrow and her colleagues observed this in laboratory experiments: when people expect to be able to find information again, they remember where to find it better than what it says. This is the “Google effect.” [1]
It is sometimes called “digital amnesia.” The term goes too far. These experiments do not show our memory collapsing. They suggest a change in what we remember.
For an appointment, knowing where to look is enough. For an idea you need to defend in a meeting, it isn't.
Preparing the slides was preparing yourself
I saw that difference first-hand when I tried building a conference presentation with AI. The slides were there. But I couldn't present them as they were.
I still had to make the material my own.
I build my talks and training sessions myself: the messages, the narrative, the sequence. And I always start the same way: by hand, on a sheet of paper. The angles, the story, the overall picture. Only then do I develop the ideas. Choosing what to say, and how to say it, is part of my preparation. While I prepare the slides, I prepare myself too.
Finding a source, dropping an idea, changing the order: all of this takes time. Time we might think is entirely saved once AI takes it on. But it is also how we work through a subject. The document can then move faster than our understanding.
Research is beginning to document this gap. An exploratory MIT Media Lab study raises a related question through a writing task. Among 54 participants, those using an AI assistant had more difficulty quoting their own text a few minutes after writing it. [2] This is a preprint based on a small sample. The distinction is still useful: producing a text is not the same as remembering it.
What about the people using these tools? In two small student surveys conducted in 2026, 16 out of 22 respondents said they often or always felt that AI led them to put less effort into searching, thinking, or memorising on their own. [3] These are self-reported perceptions, not a memory test. But they raise a question about how much of the work we continue to do ourselves.
A good explanation isn't enough
AI can also help us understand. I use it for that: to rephrase an idea or make a complex document clearer.
But we still need time to pause, connect the ideas, and use them. That work has a cost. It is called cognitive load: the effort involved in processing information. And our working memory, which holds what we are currently processing, can retain only a few items at a time. [4], [5]
This is where time matters. In my first article, I wrote that time saved does not necessarily become time available. Here, AI can help us explore a subject more deeply. Or open up five more. We then have more to examine, without necessarily remembering it any better.
Try this: close the slides
How can you tell whether you have really made the content your own? If you have prepared a presentation with AI, try this: close the slides. Explain your three main ideas out loud.
Why those ideas? How do they connect? What supports your conclusion?
You will soon hear what holds together and what is still shaky. Then reopen the presentation to check and fill the gaps.
The idea draws on research into retrieval practice: making the effort to recall information helps us retain it. [6]
Three minutes for the slides. What about preparation?
One condition remains: having that time. In a team, that isn't entirely up to you. Imagine a manager saying: “AI can prepare the slides in three minutes. Can you present them to me in half an hour?”
The deadline was based on production time. The person still needs to prepare.
We can train a team to use AI and keep expecting everything faster. This is where individual habits meet management decisions.
A first change: make the request clear. “I need your recommendation, what supports it, and the point you're still unsure about.” The discussion then focuses on the substance, rather than the slides.
If the team keeps delivering one thing after another without ever having that time, the question moves up a level: do deadlines, priorities, and workload allow for it? Training people isn't enough. We also need to change those conditions. [7]
My calendar can remember my appointments for me. For a recommendation or a decision, I want to be able to rely on what I have understood.
For humans, with technology.
What about your organisation?
When AI speeds up production, who protects the time to understand, check, and decide?
This is what I examine with teams and their leaders: how tools are used, the expectations around them, and the working rules that need adjusting. The Digital Attention Check-up provides an initial assessment and helps identify three priorities for action, remotely, in three weeks.
Frequently asked questions
Does AI cause memory loss?
The research cited here does not support that conclusion. It mainly shows that what we remember depends on the task and how we use the tool. Producing something with assistance does not guarantee that we have learned its content.
What is the difference between working memory and cognitive load?
Working memory holds the information we need for the task at hand. Cognitive load is the processing effort that task requires. It varies with the difficulty of the subject, how it is presented, and what we already know.
What role do managers play?
Make clear what is expected from a deliverable and allow time to understand and check it. Asking for the reasons behind a recommendation opens a discussion about the substance.
How can you make an AI-prepared presentation your own?
Explain its main ideas without looking at the slides, then reopen them to check. This exercise does not replace checking sources and facts.
Gabriel Pitt is a speaker, author of Be Present, and founder of OFF/ON®. He advises organisations on digital work, attention, and AI.
Sources and method
- Sparrow, B., Liu, J., and Wegner, D. M. (2011). Google effects on memory: Cognitive consequences of having information at our fingertips. Science, 333, 776–778.
- Kosmyna, N., et al. (2025). Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task. Preprint, 54 participants.
- Gabriel Pitt / OFF/ON surveys, 2026. Two student surveys, 22 responses to the item. Nonrepresentative samples; self-reported perceptions.
- Cowan, N. (2001). The magical number 4 in short-term memory: A reconsideration of mental storage capacity. Behavioral and Brain Sciences, 24(1), 87–114.
- Sweller, J. (1988). Cognitive Load During Problem Solving: Effects on Learning. Cognitive Science, 12, 257–285.
- Roediger, H. L., and Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
- Nielsen, K., et al. (2017). Workplace resources to improve both employee well-being and performance: A systematic review and meta-analysis. Work & Stress, 31(2), 101–120.