Creating the first draft of a screen, the start of a function or alternatives for a piece of text is very fast today. Speed can ease the most tiring part of making things. But the fact that something has been produced does not mean it is the right solution to the right problem.
I do not think the most valuable skill with AI is only writing good prompts. What matters more is being able to look at the result and ask the right questions: Where does this information come from? Is this flow understandable for a real user? How does the code behave in an unexpected case? Does the product support a person’s decision, or does it create a false sense of confidence?
Accepting a suggestion is easy. Testing it takes time. Especially when someone else’s data, money or an important decision is involved, that time is part of the work. Checking the source, thinking through edge cases, saying clearly what you do not know and leaving room for human review when needed are not details that slow the work down; they are the decisions that make a product trustworthy.
I see something similar in design. Tools can produce dozens of look-alike interfaces in seconds. But a tool alone cannot know why a screen is arranged the way it is, which word will reassure the user, or where it is better to stay quiet. Those take context and judgement.
So I see AI neither as a magical partner nor as mere autocomplete. It is a powerful tool that widens the space for trying things. It lets me see more paths. Which path to take, and how to carry responsibility for the resulting work, is still my decision.
Perhaps this is the core of the new way of working: producing faster is possible. But producing better still takes looking carefully, testing, and being able to say “this didn’t work” when needed.
◆