There is a stupid little trap developing around AI and professional work.
If I tell somebody I do not use AI, the obvious question is why not. Am I behind the times? Am I taking three hours to do something that could take twenty minutes? Am I charging a client for inefficiency because I have decided the old way feels more authentic?
Those are reasonable questions.
But if I say that I do use AI — heavily, in some cases — another thought can arrive just as quickly: ah, so the machine is doing it. And tucked in behind that is the pricing question. If the machine is doing so much of the work, why should the work still cost anything significant?
Then there is the clearest version of all. The client has ChatGPT. They have Claude. They can make images, write copy, generate code, research ideas and ask a model to tell them what their website should contain. Why employ somebody like me at all?
My first answer is genuinely: crack on, then.
I do not mean that defensively. If you can solve the problem yourself, you understand the trade-offs, you are happy with the result and you want to spend your own time doing it, then you probably should. I have never believed that paying a professional is morally superior to doing something yourself.
The complication begins when the problem is not the task you first thought it was.
The tools were never really the job
I have worked across websites, development, marketing, production, SEO, video and related work for more than twenty years. For much of that time, a project involved coordinating people with different skills. A designer might be doing one part, a developer another, somebody else handling video, another person doing research or marketing, while I tried to keep the whole thing pointing in the same direction.
AI has changed the mechanics of that enormously. I can now use agents and models for coding, research, criticism, imagery, planning and production work that previously needed more people or considerably more time.
Sometimes the speed difference is absurd.
But the underlying job feels strangely familiar. Somebody still has to decide what the work is, break it into sensible pieces, know what good looks like, spot when an answer is plausible but wrong, and take responsibility for whatever gets handed over at the end.
I used to have a team where more of the people were people. Now some of the team are machines.
That is a huge operational change. It is not the same thing as replacing experience.
A perfectly sensible answer to the wrong problem
I tested this with a deliberately ordinary brief: a good-looking website with video, audio, YouTube, roughly five weeks and about £2,000.
An AI gave a perfectly reasonable answer. Keep it lean. Embed the media. Avoid infrastructure nobody has proved they need. Get a strong first version live.
The hidden requirement changed the answer. The person wanted to publish written posts, video, audio and podcasts themselves after launch. They did not merely need a website containing media. They needed a publishing operation with a website attached to it.
The AI had answered the noun in the first sentence. The useful question was what the person needed to keep doing after launch.
That difference is where experience earns its keep.
The consultation is not the bit before the work
I did not always value consultation highly enough.
For years, the initial conversation could feel like the preamble: the thing you did before the proper work started. I think almost the opposite now. On many projects, that conversation contains some of the most valuable work I do.
What are you actually trying to achieve? Who is it for? How often will you use it? Who has to maintain it? What happens when something changes? What does success look like six months later? Which parts matter because of the business, and which parts are in the brief because somebody saw them somewhere else?
By the time a client has properly answered those questions, the solution often looks very different from the thing they originally asked me to price.
That is not because I possess some mystical consultant power. It is because I have spent a long time watching projects succeed, fail, overrun, get overbuilt, get underbuilt and occasionally turn into things nobody expected at the start.
Experience changes the questions you ask.
It also changes when you are prepared to disagree.
AI can be extremely good at helping pursue a goal once the goal has been described. A model can challenge assumptions too, but the person asking still needs enough context and judgement to know which assumptions are worth challenging.
Sometimes the most useful thing I can say to a client is effectively: I am not going to give you exactly what you asked for, because after talking this through I do not think it is what you need.
That only works if I have listened properly first.
AI has made production cheaper. That is not the same as making expertise cheaper.
AI absolutely has reduced the cost of producing certain kinds of work. Pretending otherwise would be ridiculous. Tasks that once required hours can take minutes. Some research, drafting, development, image production and analysis are dramatically faster than they were even a few years ago.
That changes markets. It changes what is sensible to charge for routine execution. It changes which pieces of work are worth outsourcing at all.
What it does not do is make every part of the job equally cheap.
The difficult part of many projects was never the number of keystrokes involved. It was deciding what to make, understanding the consequences of that decision, recognising when the apparently correct output was solving the wrong problem, and being accountable when the finished thing had to work in somebody's actual business.
A client is not paying me for exclusive access to VS Code, Photoshop or an AI model. They have access to most of the same tools I do.
They are paying me to get from their messy problem to a finished result that makes sense.
If AI lets me do that better or faster, I am going to use it. I would find it odd to hire somebody who deliberately refused a tool that materially improved their work because they were worried efficiency might make them look less valuable.
The expertise is not the ceremony around producing the artefact.
It is the judgement applied to it.
Faster production can make judgement more important
When creating options becomes cheap, you get more options.
That sounds like an uncomplicated advantage until you have twenty plausible answers instead of three.
AI can produce multiple homepage structures, several technical architectures, dozens of names, design directions, research summaries and implementation routes almost instantly. The bottleneck moves. It is no longer always can we produce something? It becomes which of these things is actually right, what evidence supports it, and what are we prepared to stand behind?
I see that constantly in my own work now. The machines make it easier to generate. They have not removed the need to reject.
In some ways they have made rejection more important.
A bad idea used to have the small mercy of taking time to produce. Now it can arrive beautifully formatted in thirty seconds.
So what are you paying me for?
When I was asked to reduce the answer to one line, I initially came up with a list: humanity, experience, personality, critical judgement.
All true, but not much of a line.
The shorter answer is probably just this:
You are paying for me.
Not because I think I am irreplaceable, and not because I have some sacred human quality that a model can never touch. I am quite happy to let the tools do everything they are better at doing.
What the client gets from me is the combination of those tools with my particular history of decisions, mistakes, pattern recognition, scepticism, taste, experience and willingness to be responsible for the result.
That is also why I am perfectly comfortable telling somebody to do it themselves if they can.
If you have the interest, time and judgement to work through the problem, interrogate your own assumptions, choose the right approach, verify the output and live with the consequences, then genuinely: crack on.
But if you ask me to solve it, I am not charging you for how long it takes me to type a prompt.
I am charging you for knowing what should happen after you tell me what you think you want.