Artificial intelligence gets discussed in two ways: either it will take every job, or it is nonsense producing weak output.
Both are far from reality. What I see in the companies I work with is that AI is a tool that saves time on specific things and fails completely at others — and the difference between success and failure is knowing those limits.
Where does it genuinely help?
- The blank page. The hardest part of any content is the empty screen. AI gives you a draft to start from.
- Repetitive work. Turning an article into posts, writing many product descriptions, or organising ideas.
- Summarising and reading. Summarising calls, customer comments, or long reports.
- Brainstorming. Twenty ideas in a minute, from which you pick two.
- Translation and rephrasing. Turning a message from colloquial to formal or into English quickly.
All of these take hours a week from small companies, and the saving is real.
And where does it fail?
- Knowledge specific to your business. It does not know your customers, your prices, or what has worked with you before.
- Judging quality. It produces convincing-looking text even when wrong, and you are the one who has to review.
- Numbers and facts. It can invent statistics and names. Any number from it must be verified.
- Your own voice. The output reads generically, which is the worst thing in a market where everyone says the same thing.
The problem is not the tool, it is lazy use
The reason people have come to dislike generated content is that it gets published as is, with no review and no addition.
The right use: you supply the context (who the customer is, what the problem is, what your style is), take the draft, edit it with your knowledge and experience, and add real examples from your work. The output then becomes yours, and the tool saved you the time of structuring and phrasing.
Will it take jobs?
It will change their shape. Jobs based on repetitive execution — writing descriptions, simple designs, ordinary translation — feel a large effect.
Jobs based on judgement, decisions and relationships — strategy, analysis, dealing with clients, negotiation — increase in value, because execution has become cheaper so the right decision matters more.
The practical advice for anyone in the field: learn to use these tools well, and concentrate your skills where context and judgement are required.
Start with one thing
Do not try to bring these tools into everything at once. Pick one task that takes you time every week and try it for a month.
Measure two things: how many hours you saved, and whether quality dropped. If time was saved and quality held, continue and extend to a second task. That practical approach prevents the enthusiasm that ends in weak content.
Watch your privacy
Do not put your customers' data or your company's sensitive numbers into open tools without knowing their policy.
That is not excessive caution, it is part of your responsibility to the people who deal with you. Use public or anonymised data in experiments, and keep sensitive things in tools you are sure of.
How do you write a good prompt?
The quality of what comes out depends on the quality of what you put in. A weak request produces generic text; a good one produces something close to what you wanted.
A good request contains four things: who you are speaking to (describe the customer in detail), the purpose of the text, your style (an example of your previous writing), and the constraints (length, and what must not be said).
Write that request well once and keep it, then use it as a base every time. That saves more time than anything else.
Reviewing is your job
Anything coming out of the tool has to pass three reviews: is the information correct? Does this sound like me? And is there something from my real work I could add?
That addition is what makes the difference. A real example from a project, a number from your experience, or a dissenting opinion — those are what make your content unlike everyone else using the same tool.
The value that cannot be copied
Once anyone can produce content quickly, the things that matter become clearer: real experience, examples from your own work, a clear opinion, a relationship with the customer, and delivery on time.
No tool does those. Which is why I am not worried about anyone doing the work properly — I am worried about anyone whose work was entirely repetitive execution with no added value.
In short
AI is neither an enemy nor a miracle. It is a tool that saves time on repetitive work, drafts and summaries, and fails at knowledge specific to your business and at judging quality.
Use it with context and review, start with one task, watch the numbers and the privacy, and concentrate your value in what cannot be copied.
This is part of a series on working in digital marketing — which also covers the jobs in the field, and the difference between digital and traditional marketing.