What AI Cannot See: Intuition, Education, and the Human Work of Trust
- Jul 28
- 5 min read
Watch the full conversation below.
The best strategy may begin with salceson
Some stories stay with us because they explain more than a polished presentation ever could.
Around twenty years ago, Bartosz Jędrzejczak was helping build an innovative centre for people with disabilities in Mielnica, near Lake Gopło. The idea was practical and radical at the same time: adults should be able to live and work in their own community instead of being forced to leave it.
The project needed electrical installations, a new connection to the power grid, and help from an experienced engineering company. Bartosz met a businessman who lived in the United States and visited Poland from time to time. There were many obvious ways to start that conversation: explain the mission, present the budget, or walk through the technical requirements.
He chose a different opening.
He invited the man to try the best salceson in the area.
They ate together at the construction site. The businessman looked around, asked what was needed, and eventually offered to send engineers and a team to install the necessary systems.
The salceson was not a clever sales trick. It was a small human gesture that made room for curiosity. Bartosz did not try to force the conversation toward the thing he wanted to say. He looked for a point of genuine contact.
That is why the story matters. The decisive move was not hidden in the data. It was hidden in the context around the data.
The same pattern appeared in other stories from the conversation:
an unconventional welcome for a foundation leader;
an unexpected visual joke that helped unlock funding for training apartments;
rainwear distributed in towns preparing for a papal visit, an idea that looked irrational until the human situation became visible.
From a distance, these decisions looked strange. Up close, they made sense.
Intuition is not the enemy of strategy
We often describe intuition as the opposite of analysis. The conversation suggests a more useful distinction.
Analysis gives us a map. It helps us understand interests, strengths, experience, goals, risks, and constraints. That map is valuable. It can prepare us for most of the terrain.
But a map does not tell us everything about the person standing in front of us.
The detail that changes a conversation may be:
a joke that arrives at exactly the right moment;
a pause that says more than the sentence before it;
an unusual association;
an idea that sounds absurd until the surrounding context catches up with it.
These details do not replace preparation. They allow preparation to meet reality.
A technically sound proposal can still fail to create energy. Someone can have an impressive list of competencies and still move toward work that does not fit them. A conversation can follow a carefully designed script and miss the one sentence that reveals what really matters.
Intuition does not arrive with a spreadsheet label. It grows from experience, observation, and the willingness to notice what does not fit the current model. It also asks us to trust a quiet internal signal before we can fully explain it.
Education should not make everyone average
The conversation then turns to education, and to a familiar mistake.
We often spend enormous effort making people moderately competent at things they find genuinely difficult, while paying less attention to the abilities that come naturally to them. A learner struggles with a subject, receives more exercises in the same subject, and is measured by how closely they approach an average standard.
Meanwhile, their strongest abilities may remain almost invisible precisely because they look easy.
This is how education can confuse improvement with conformity.
A better starting point is to ask:
What does this person notice before others do?
What kind of work gives them energy rather than draining it?
Which strengths appear in informal situations but disappear in formal tests?
What unusual combination of interests could become meaningful?
Standards, fundamentals, and discipline still matter. They give people the tools to function. But they should not become the entire definition of development.
Education should help people become capable enough to choose their own direction, and distinctive enough to make that direction their own.
Where AI becomes genuinely useful
AI already has a valuable role in this process.
Bartosz describes an experiment in which he gave an AI system years of digital notes and meeting records, then asked it to identify his blind spots. An earlier model produced an ordinary answer. Newer models, working with much more context, offered a far more useful outside perspective.
The model did not discover a mystical new intuition. It surfaced patterns that were already present in the material. The information was visible in the notes, but the person who wrote them was too close to their own story to see it.
AI became a mirror.
It can help us:
organize scattered observations;
compare choices across many years;
notice repeated patterns;
formulate questions we would not think to ask ourselves;
identify strengths hidden behind formal performance.
That is not a small achievement. It can make reflection faster, broader, and more honest.
The limit is not intelligence. It is relationship.
The difficult boundary appears when we move from explicit information to lived experience.
Human work often depends on trust, vulnerability, grief, love, fear, confidence, and the tiny signals that appear in a live relationship. Those experiences are not simply missing fields in a database.
A coach or tutor does not only process what someone says. They notice how the person says it, what they avoid, when their voice changes, and when an apparently irrelevant sentence suddenly becomes central.
That does not make every human judgment correct. It does not make technology useless. It means that replacing the relationship with a system changes the nature of the work.
Trust is something we build, not a feature we add
Trust cannot be added to a product specification at the end of a project.
It is earned through repeated interactions and the experience of safety. When people share personal information with a coach, tutor, therapist, or mentor, they are not merely submitting data for analysis. They are entering a relationship with expectations about discretion, care, and responsibility.
Some people may feel safer starting a difficult conversation with AI than with another person. That possibility deserves to be taken seriously. It also raises concrete questions:
What is stored?
Who can access it?
How is it used?
What can the system honestly promise?
Where must human responsibility remain visible?
The useful question is not whether AI is good or bad in the abstract. The useful question is what kind of help someone needs, what risks are acceptable, and what should remain deliberately human.
Choosing what not to outsource
The conversation closes with an image from Blade Runner: a test designed to expose what is not human, and the possibility that the test may reveal something about the people administering it.
As machines become better at recognizing patterns, simulating empathy, and producing convincing advice, we will have to make more intentional choices. Someone may want a virtual companion for one kind of conversation. Someone else may prefer a live musician, craftsperson, tutor, or mentor precisely because the work contains presence, friction, and imperfection.
The important thing is not to choose one technology forever. It is to know what each option can do, what it cannot do, and what we value enough not to outsource.
The salceson story offers a practical conclusion. The most important move in a conversation is not always the most polished one. Sometimes it is the unexpected gesture that lets another person feel seen.
AI can help us study the record of our lives and notice patterns we missed. It can help us ask better questions. But the decision to listen, to trust an intuition, to recognize another person, and to take responsibility for what happens next remains a human choice.