Mention AI in connection with programming, design, writing, or almost any kind of digital work, and the same reaction still comes up surprisingly often.
“So AI did it, then.”
As if that automatically made the result less valuable.
I understand where the feeling comes from. Generative AI can produce a piece of copy, an image, working code, or the outline of a solution in seconds. Only a few years ago, the same first draft might have taken considerably longer.
But I think this confuses two very different things.
Producing an output is not the same as producing a good result.
The longer I work with AI, the more convinced I become that this distinction will matter more, not less.
AI can produce average work at remarkable speed
Open almost any current AI tool and type:
“Design a modern website for a manufacturing company.”
The result probably will not be terrible. It will have a large photograph, a confident headline, several services, a few testimonials, a button, and a contact form. It may even look quite polished.
The problem is that exactly the same concept could belong to almost any manufacturer.
Without meaningful context, AI cannot know what actually matters about this particular business. It does not know that the company makes something its competitors cannot. It does not know that the most commercially valuable service is buried on the fourth page. It does not know that the owner no longer wants a certain kind of project. It does not know that the service presented as the main offer on paper is, in practice, only a minor part of the business.
Most importantly, it cannot know which decision is right for this situation unless someone helps it understand the situation first.
Without a person who understands the problem, AI can be an extraordinarily efficient generator of average work.
AI has not taken my work away
It has removed a great deal of the time between a question and the first useful direction.
That is a very different thing.
When I work on a new website, I do not ask AI to “make me a website”. I use it across dozens of smaller steps: research, sorting large amounts of information, finding connections, reviewing copy, identifying what a page may be missing, breaking a technical problem into manageable parts, checking code, looking for edge cases I may not have noticed, and documenting decisions.
I also use it to analyse client communication, or simply when I have an idea and need something that will do more than agree with me. I want it to test the idea and look for the weak point.
That has probably changed the way I work more than anything else in recent years.
AI did not become the person making decisions for me. It became another layer in my working process.
The greatest value is not generation
It is dialogue.
When I think through a problem alone, I naturally operate within my own experience, habits, and assumptions. That is useful, but it is also limiting.
Used well, AI gives me a quick way to put my view under pressure from another angle.
“Why might this be the wrong solution?”
“What have I missed?”
“Which part could confuse the client?”
“Which claim have I not supported properly?”
“If this system fails, where is it most likely to happen?”
“Is this a real problem, or am I polishing a detail that does not matter?”
Those questions are far more valuable to me than: “Write this copy.”
AI often does not tell me what to do. It helps me see the problem more clearly, so I can make the decision myself.
I had to learn to think more clearly too
This may be the most counter-intuitive part of the whole experience.
You might expect a tool that accelerates part of the thinking process to make its user think less. My experience has been almost the opposite.
The more I use AI, the more precisely I have had to define:
- the problem I am actually solving,
- what I already know,
- what I am merely assuming,
- the outcome I want,
- the constraints around it,
- what must not happen,
- and how I will recognise a good result.
An unclear brief produces an unclear result very quickly.
Over time, I began externalising much more of my own thinking. I write rules, define processes, record decisions, and separate facts from hypotheses. When a task repeats, I try not to reinvent the same method from scratch. I turn it into a system instead.
That is one of the most interesting effects AI has had on my work.
A good tool did not make me more passive. It forced me to become more precise.
Working faster does not make the work less valuable
I think we will wrestle with this idea for some time.
If a task used to take eight hours and better tools reduce it to two, we are tempted to conclude that the work must now be worth less.
But a client does not need to buy eight hours of my life. The client needs the result.
If I can understand the problem, design the right solution, test it, implement it, and remain accountable for it in less time, I see no reason to treat that as a disadvantage.
A developer does not write every byte by hand. A designer does not mix their own ink. An accountant does not calculate everything on paper. A photographer does not develop every image in a darkroom.
Tools change.
The value of human judgement does not disappear with them.
AI can create a convincing illusion of competence
This is where the real problem begins.
AI can sound confident while being wrong. It can produce code that looks correct but contains a subtle fault. It can write professional-sounding company copy while inventing a detail that is not true. It can propose a technically elegant solution without understanding the environment in which that solution has to work.
If the person using it cannot evaluate the result, a confidently written answer can easily be mistaken for a correct one.
That is why one part of my work remains non-negotiable.
I check.
If the work concerns a specific company, I verify the claims. If it concerns a website, I inspect the real website. If it concerns communication, I need the context of the full conversation. If it is a technical solution, it has to work outside a tidy demonstration.
And if AI suggests copy that sounds professional but does not sound like me—or simply makes no sense in the actual situation—I do not use it.
An AI output is not finished work. It is material I continue to work with.
The human element did not disappear
If anything, it became more important.
When almost anyone can produce a reasonably good first draft, the difference moves elsewhere: into recognising which direction is right, having taste, applying experience, understanding the context, knowing when to say no, checking the details, and understanding the person on the other side.
Ultimately, it moves into responsibility for the result.
AI does not live with the consequences of a website I launch for a client. It does not build a long-term relationship with them. It does not pick up the phone when something stops working. It does not explain why a decision was made. And it does not suffer the reputational damage if the work is delivered badly.
That responsibility still belongs to a person.
Refusing to use AI will increasingly feel like the unusual choice
I do not know where these technologies will be in five or ten years.
I am fairly certain of one thing: they are not going away.
They will become an ordinary part of the working environment, much like the internet, search engines, cloud services, modern development tools, and countless other technologies we no longer think of as a competitive advantage.
For that reason, dividing people into those who use AI and those who do not seems less and less useful to me.
Another distinction will matter more.
There will be people who use it without thinking, and people who learn to use it as a tool for their own thinking.
Conclusion
AI has allowed me to do certain things much faster over the past few years.
But speed is not the most important thing it has given me.
It has pushed me to define problems more precisely, document processes more carefully, challenge my assumptions more often, and distinguish between a first idea and a sound decision.
In many situations, it helps me reach the point where the work is no longer about generating more options. It is about exercising judgement.
That is why I do not see AI as a replacement for a person, and certainly not as a button that allows us to bypass the work.
It is a tool.
A very powerful one.
And as with any tool, what matters in the end is who is holding it—and what they are capable of doing with it.