The Making of a LinkedIn Post: Where Human Judgement Shapes AI

A polished LinkedIn post can appear to have been created in a single moment: one photograph, one prompt and a few seconds of artificial intelligence.

The reality is more interesting.

Behind a recent post about my conversation with Ricardo Damas was a sequence of human decisions—what to emphasise, what to correct, what to preserve and which seemingly small detail changed the meaning of the story. This is a behind-the-scenes look at how AI helped shape the language, while human judgement shaped the outcome.A photograph. A meaningful conversation. A few notes about who was there and what we discussed.

A photograph. A meaningful conversation. A few notes about who was there and what we discussed.

That is enough for generative AI to produce a polished LinkedIn post in seconds.

But a polished draft is not necessarily the right post.

The making of a recent post about my conversation with Ricardo Damas, Vice President and General Manager of Alexion Spain and Portugal, reminded me that AI can accelerate writing, but human intervention gives the final text its truth, relevance and voice.

A strong first draft—and what it missed

I began with the essential facts.

Ricardo and I had spoken about the pharmaceutical sector, artificial intelligence, health data, rare diseases, organisational transformation and the growing Real Data, Better AI movement.

The first AI-generated version brought these ideas together effectively. It was clear, professional and suitable for LinkedIn. It highlighted responsible AI leadership, data quality, patient value, trust and collaboration.

It was a good post.

Yet it described the content of the conversation without capturing the full meaning of the encounter.

The detail that changed the story

Ricardo and I both live in Linhó, Quinta da Beloura. Our conversation took place at Fundação Albuquerque, also in Linhó.

That detail was not merely geographical. It introduced the power of place.

Two people with overlapping professional interests had met within the territory they both call home. It suggested that important collaborations do not always begin at international conferences, corporate headquarters or formal meetings.

Sometimes, they begin much closer to us.

I asked the AI to add this reflection. The new version was thoughtful, but it allowed the place narrative to dominate, pushing aside some of the original substance about AI, health data, rare diseases and industry leadership.

My next instruction was therefore precise: retain the first post and add the power of place as another layer, not as a replacement narrative.

That intervention changed the outcome.

The human role is deciding what matters

The process was not simply a matter of asking AI to write and then accepting what appeared.

The AI proposed a framing. I evaluated it.

It expanded one theme. I restored the balance.

It generated fluent language. I corrected the facts.

At one point, it even misunderstood the location, and I had to clarify that the conversation had taken place at Fundação Albuquerque in Linhó.

This is an important reminder: fluency is not the same as accuracy.

The human author remains responsible for checking names, titles, places, emphasis and meaning.

AI did not know why Linhó mattered. It did not experience the setting, the conversation or the unexpected discovery of shared local ground. It could only work with the significance I progressively revealed.

From prompt engineering to meaning engineering

Much is said about prompt engineering, but the more important capability may be meaning engineering.

A prompt can ask for a polished post.

Meaning engineering asks:

What is this really about?

What should remain central?

Which detail changes the story?

What sounds elegant but is not quite true?

Where is my own perspective in the text?

In this case, the post needed to hold several layers together: the quality of the conversation, Ricardo's leadership, responsible AI, trustworthy data, rare diseases, patient value, Real Data, Better AI and the significance of meeting in a place we both call home.

The final post became stronger because these elements were deliberately integrated rather than simply assembled.

Why share the secret?

There is a temptation to present AI-supported writing as effortless: one prompt, one answer, one finished post.

That is rarely how the best outcomes emerge.

I am sharing the making of this post because I do not believe the real value lies in hiding the tool or protecting a supposed formula. The value lies in judgement: noticing what is missing, correcting what is wrong, resisting generic language and continuing until the words reflect the experience.

Sharing the process does not diminish authorship. It makes authorship more visible.

It also helps us move beyond the unhelpful question of whether a text was written by a human or by AI. A better question is:

How did the human being shape the outcome?

My "secret" is therefore not a clever prompt.

It is the willingness to intervene.

To question the first answer.

To introduce lived context.

To protect the meaning.

To remain accountable for the final words.

AI may help us articulate an experience, but only the human being can determine what that experience means.

Paul Nunesdea, PhD, CPF