For decades, we organised human skill into two neat categories.
Hard skills were the serious ones. The technical ones. The ones you could put on a CV and point at.
- Code.
- Data analysis.
- Financial modelling.
- Legal frameworks.
- Medical procedures.
- Engineering.
These were called hard because they took time to learn, required credentials to prove, and came with a body of knowledge that not everyone could access. They were the currency of professional credibility. The thing that got you in the room.
Soft skills were everything else.
- Communication. Empathy.
- Persuasion. Leadership.
- The ability to read a room, motivate a team, or sell an idea to someone who was not yet convinced.
These were acknowledged as useful but never quite treated as serious. Most did not have certifications. They could not be easily measured. They lived in the grey area between talent and personality, and nobody could quite agree on how to teach them.
So we called them soft. And in doing so, we told an entire generation of professionals that the human stuff was the easy stuff.
We were wrong. As a software engineer myself who tried for years to convince people of this truth, I was met with deaf ears. AI just proved it in the most public way possible.
Here is what AI exposed about hard skills.
Most of them are formula-based. Not in a dismissive way. But in a precise and important way. They follow logic. They respond to rules. They can be broken down into steps that, if followed correctly, produce a predictable output. That is exactly why they were learnable. And it is exactly why they are automatable.
A machine does not get tired of following steps. It does not cut corners on a Friday afternoon. It does not make the small errors that accumulate when a human has been staring at a screen for six hours. Given the right training data and the right architecture, a machine will execute a formula better than a human every single time.
This is not an insult to the people who spent years mastering those skills. The mastery was real. The effort was real. But the nature of what they mastered, precise, logical, repeatable, was always going to be more vulnerable to automation than anyone admitted while the automation was still theoretical.
GitHub Copilot does not replace the developer who understands the problem. It replaces the developer who only knew how to write the solution once someone else understood it.
That distinction is everything.
Now look at what AI cannot do.
- It cannot sit across from a grieving family and know exactly when to speak and when to stay quiet. It cannot read the micro-expression that crosses a procurement manager’s face when the price is revealed and adjust the pitch in real time. It cannot sense the energy shift in a room when a presentation is losing the audience and change direction before the audience knows they have checked out.
- It cannot sell. At least not in the way a great salesperson sells. Because selling is not presenting features and benefits in the right order. Selling is understanding what a person needs to feel before they can say yes. And what they need to feel is different for every person, in every moment, shaped by everything that happened to them before they walked into that room.
- Marketing at its best is the same. Not the technical execution of a campaign. It is the human understanding underneath it. It is the ability to see a person clearly enough to know what they actually need to hear. To find the emotional truth inside a product or a service and connect it to the emotional truth inside the person you are talking to.
No model trained on human data fully replicates human feelings. It only approximates it. Sometimes impressively. But approximation is not the same as understanding. And in the space between approximation and understanding, the human professional still lives.
Soft skills are hard. They require a depth of human understanding that takes a lifetime to develop and cannot be reduced to a formula because the input, another human being in a specific emotional state, is never the same twice. They are hard because they require presence, judgment, intuition, and the kind of emotional intelligence that comes from having lived, having failed, having sat with someone in their difficulty and found the right words, not because a system recommended them but because you understood.
Hard skills are, in many cases, easier than we thought. Not easier to learn. But easier to replicate. Easier to automate. Easier to hand to a machine that will do them faster and with fewer errors than most humans ever could.
This does not mean technical skills are worthless. The developer who understands the problem deeply, who can translate a messy human need into a precise technical solution, is more valuable than ever. Because what they have is not just the technical skill. It is the judgment that sits above it. The human layer that no tool can replace.
That layer was always the hard part. We just spent decades calling it soft.
The professionals who will thrive in this era are not the ones who can do the most technical things. They are the ones who understand people well enough to do what the tools cannot.
AI did not create this truth. It just finally made it loud enough to hear.
In conclusion, the code was never the hard part. Understanding the person the code was built for was always the hard part.
It still is.
