The bell rings, it is break time, and that full, chaotic sound of voices, footsteps and scraping chairs echoes down the corridor. We, however, are somewhere else: we have that laser focus that only comes when something truly matters to us.
We tear a clumsy scrap of paper from an exercise book, rest it on our planner and write: “Will you go out with me?”
Below, to make it almost “manageable”, we even add check boxes: “Yes ☐ No ☐”.
We read it over, fold it up, slip it into our pocket. Every now and then, we touch that exact spot to check it is still there. Even before handing it over, something happens: we feel just that little bit more capable. And let’s be honest: who hasn’t had a school crush and sweated bullets just to take that tiny step?
In psychology, that “okay, I can do this” feeling is called self-efficacy (Bandura, 1977): it is the tangible sense of being able to tackle a task. In a subtle way, that little note did exactly that: it lowered the initial friction and helped us take the plunge.
Today, when we have to write a sensitive message, an email we cannot afford to get wrong, or feedback that risks sounding harsh, we often look for the same thing. A foothold, a first draft, a way to get going without staring into the void.
In many ways, AI works like an “evolved” version of that little note: it does not necessarily replace us, but it lifts the weight of the first step off our shoulders.
This is good news, because so often we are not held back by a lack of skills, but by the mental and emotional effort of starting. Yet this is where the interesting question arises: does this support help us become more autonomous, or does it train us to feel capable only as long as it is there? In other words: are we building confidence that remains ours, or confidence that is merely “on loan”?
That note is also a textbook example of cognitive offloading: moving a piece of effort outside our heads so we can take action. It is a deeply human thing: reminders, lists, maps, sticky notes… they are all ways to lighten the mental load.
AI makes this offloading far more powerful because it doesn’t just outsource “things to remember”, but chunks of writing, synthesising, rephrasing, and reasoning. The result: less strain, more speed. It works, and precisely because it works, it reshapes our habits.
For instance: when the tool performs well, we feel stronger, faster—and that is when a classic cognitive trap kicks in: the illusion of control (Langer, 1975). With powerful tools and instant answers, our sense of mastery can outpace our actual mastery.
What deserves attention is that the response is often well written, sounds convincing, and makes it all too easy to say “fine, that’s sorted” and move on. In organisations, this produces two very concrete outcomes: on the one hand, decisions made with less scrutiny; on the other, expanding expectations (“if we are faster now, we can promise more”), because everything genuinely feels more manageable.
Then there is another effect that tricks us with utter gentleness: if a text “flows”, we trust it more. This is processing fluency (Schwarz et al., 1991): what is easy to read tends to feel truer and sounder to us. It is a completely normal automatic response.
So far we have talked about texts and decisions, but there is a delicate piece missing: relationships. Because organisations thrive on coordination, trust, negotiation, and feedback—things that are conveyed through tiny signals like pauses, rhythm, energy, hesitations, and irony, which are not just details, but crucial clues.
This is where the concept of Thin Slicing (Ambady & Rosenthal, 1992) becomes useful: the ability to gauge the atmosphere from “thin slices” of behaviour—a few signals in a matter of moments. If communication becomes cleaner and more standardised (always polished copy, flawless translations, ultra-smooth messages), those signals do not vanish, but they change form and can become harder to read. As a result, we risk missing subtle caution, light resistance, or silent misalignments along the way.
When the signals are ambiguous, the brain does something natural: it tries to fill in the blanks. A concept that helps us understand this dynamic is Theory of Mind (Premack & Woodruff, 1978)—the ability to imagine what someone else is thinking and feeling, while remembering that it does not match our own perspective. It is an essential skill for collaborating effectively, because it forces us to stay in “hypothesis” mode rather than “verdict” mode.
The problem is that when we are rushing, or when everything seems to be running smoothly already, this skill easily switches off and is replaced by a default reflex: projection (Freud, 1894). In practice, we fill in the other person using ourselves as the template. “If it’s obvious to me, it must be obvious to them.” “If it sounds like this to me, that’s what they must mean.” It is a quick way to resolve ambiguity and move forward.
And here, cognitive offloading can quietly amplify the issue: when everything becomes smoother and neatly packaged, we feel as though we understand the relationship better too. In reality, sometimes we are just jumping to conclusions faster.
So no, the issue is not “AI: yes or no”. The real question is: how do we use offloading without surrendering our critical thinking? Here, paradoxically, we don’t need heavy policies: we need light micro-rituals. After reviewing a draft or a summary, before sending or deciding: “What is the main assumption here?” “What context is missing?” “What consequences will this have on people, timelines, and quality?” In key conversations—those where a misunderstanding carries a high price—two simple questions can shift the tone: “How is this landing with us?” and “What are our key takeaways?”
In the end, it is much like that little note: incredibly useful for getting started, but real growth lies in remembering that, sooner or later, we want to be able to say the words ourselves.
Further reading: The brain at work
