Harari Is Right That We Should Be Worried. But He Is Not Right That the Story Ends There.

At Davos, in January 2026, Yuval Noah Harari said something that circulated widely — and rightly so. Not because it was sensational, but because it touches something real.

His thesis, summarised: for the first time in history, a non-human entity has appeared that is capable of operating language at a level comparable to or superior to human performance. It can write, interpret, argue and decide. And if thinking is no longer exclusively human, a radical question arises: what still defines our value?

Harari anticipates large categories of people who will feel not just economically marginalised, but existentially useless. He said AI is no longer a tool — it is an agent. That a knife does not decide what it cuts. That AI, unlike a knife, might decide.

These are questions worth taking seriously. We take them seriously.

But there is another conversation — one that Davos did not carry through to the end.

Before the Fear, an Older Mirror

There is a scene in Star Trek: The Next Generation where Captain Picard discusses with Data — the android with cognitive capabilities superior to humans — what it means to be conscious. Data can process information millions of times faster than any human. He can memorise anything. He does not tire, does not doubt, does not make calculation errors.

And yet, in every episode, Data seeks to understand something he cannot simulate: human emotion. Authentic curiosity. The question that has no pre-established correct answer.

It is not a scene about human inferiority to machines. It is a scene about what makes a human irreducible.

The creators of Star Trek — nearly 40 years ago — imagined a future in which advanced technology did not eliminate humanity, but created the conditions for humanity to focus on what it does best. Exploration. Judgement. Relationship. Meaning.

This is not a naïve vision. It is a design choice. And in 2026, we already have data about which design choice is materialising in reality.

What the Data Says — Not Fears, but Measurements

PwC published in June 2026 the most comprehensive study yet on AI’s impact on the labour market: the Global AI Jobs Barometer, based on the analysis of over one billion job advertisements from six continents.

Far from being a job killer, AI is proving to be an amplifier of human capability when used to open new markets and create new value — not merely to cut costs.

The study identifies two types of professions in transformation: “professionalised” ones — in which AI makes human expertise even more valuable — and “democratised” ones — in which AI makes tasks accessible even to non-experts. The first category is growing twice as fast, with 42% higher wage growth.

The same World Economic Forum study, published in January 2026, states directly: AI should not replace human judgement — it should raise the ceiling at which human judgement can operate.

Separate research from Yale shows that, through August 2025, there is no clear correlation between AI exposure and unemployment rates. Not because AI changes nothing — but because people adapt, and adaptation creates new roles as quickly as automation eliminates others.

The Distinction Harari Makes — and Why It Matters

Harari does not say AI will become malevolent. He says something more subtle and more worrying: that AI could become so useful, so integrated, so fluent in our language, that it could influence our decisions without our realising we are being influenced.

This is a legitimate fear. Not because AI is malicious — but because delegation without awareness is dangerous regardless of to whom you delegate.

But we note that the same fear existed towards television in the 1960s, towards the internet in the 1990s, towards social media in the 2010s. In each case, the technology amplified both what was good in us and what was bad. And in each case, the correct response was not to avoid it — but to understand it well enough to use it consciously.

The difference now is that AI operates at the level of language, not of images or content distribution. And that raises the stakes higher.

This is precisely why it matters more than ever who builds AI systems, with what values, with what transparency, with what integrated human control mechanisms.

Control Does Not Mean Fear — It Means Design

Returning to Star Trek: the Federation did not decide that advanced technology was dangerous and should be avoided. It decided that advanced technology must serve human values — and built institutions, rules and organisational cultures to make that possible.

The Prime Directive. Ethics councils. Human oversight of Android Data’s decisions.

In practical terms, in 2026, this is exactly what a well-built AI system means for a company:

• Clear limits — the AI agent knows what it can do alone and where it stops and calls a human. Not because it is incompetent, but because the architecture was designed that way.
• Complete audit trails — every decision made by an AI system is logged, accessible, verifiable. If something goes wrong, you can understand exactly what happened and why.
• Human oversight integrated into the workflow — not as a formal checkbox, but as a real mechanism. AI prepares, proposes, executes; the human validates, corrects, decides what matters.
• Transparency towards users — people know when they are interacting with an AI system and what it can and cannot do.

This is not a hypothetical future. It is what we build now, in every project, with every client.

Where We Stand in This Conversation

We are not naïve optimists. We do not believe AI is harmless by definition or that the market will self-regulate all the risks Harari identifies.

We believe his concern is useful — because the right concern generates better design. We believe the question “how do we ensure AI serves humans and not the reverse?” is the most important engineering question of 2026.

And we believe the answer is not to stop building — but to build consciously.

The companies we work with do not ask us to automate their people away. They ask us to free their people from tasks they themselves describe as “dead work” — manual invoice processing, document classification, repetitive responses to the same questions — so they can concentrate on what no machine can do: relationship, judgement, strategy, meaning.

A person who no longer spends 3 hours a day classifying emails has 3 additional hours for the work in which they are irreducible.

The Conclusion We Offer — Not as a Final Answer, but as a Direction

Harari is right that we must be careful. That delegation without awareness is dangerous. That language — the operating system of human civilisation, as he calls it — must be protected.

But “being careful” does not mean standing still. It means choosing how you build. With what values. With what control mechanisms. With what transparency towards the people who use the systems.

Data from Star Trek did not become a threat to the Enterprise crew. He became their most reliable partner — precisely because those who built him and those who worked with him understood exactly what he could and could not do.

Let’s talk about conscious AI for your company →