What wins

Published on
Shreshth Mehra-
3 min read
What to do, How to do it, Doing it

What to do, How to do it, Doing it

Contents

With AI progressing rapidly, the big question is: what skills will be valuable in the future? In any business, there are three key components—what to do, how to do it, and doing it. Of these, what to do pays the most, how to do it pays less, and doing it pays the least. As of May 2025, the doing part has already become easy with AI. The how still holds some value, but I believe its importance will also diminish over time. Ultimately, the only valuable skill in the future will be knowing what to do.

The Declining Value of Execution

Let’s consider a software engineering company as an example:

  • The CEO gets paid the most because they decide what to do.
  • A project lead or supervisor earns less by figuring out how to do it (deciding the architecture and tools).
  • A software engineer earns the least by actually doing it (writing the code).

What’s interesting is that the value of the doing part is dropping further rapidly due to AI coding agents. I rebuilt an old software project of mine, and it took me only 7hrs (vs 77hrs previously). This effect isn’t limited to software. Whether it’s video generation, writing, or research, the doing part has lost its value because AI can do it better, faster, and cheaper.

The "How" Is Still Valuable—For Now

That said, the how still holds some relevance today. For instance:

Let’s say you want to build a custom inventory management app for your small business. Even with no coding experience, you could—using tools like Replit and a few prompt iterations—build a functioning application.

However, without understanding how Replit did it (e.g., what database it used, where it's hosted, etc.), it’s hard to scale or move it to production. But if you know how it should be done and can guide an AI agent accordingly, your chances of success go up dramatically.

But the "How" Is Also Becoming Less Valuable

With AI agents improving so rapidly, I am willing to bet that the “How” will also become a lot less valuable very soon. Take photography for example:

There was a time when you needed to understand aperture, shutter speed, and ISO settings—the how of photography. Today, smartphones handle all of that, including post-processing.

Now, the most valuable part of photography is the what:

  • What story are you trying to tell?
  • What emotion are you trying to capture?
  • What creative choices are you making?

The difference between an average photo and a great one is no longer technical—it’s conceptual.

The Future Belongs to the Thinkers

With AI automating both execution and methodology, the people who will remain valuable are those who can decide what should be done. However, this is not to suggest that one should stop focusing on the doing or the how. After all, the best way to learn what should be done is by doing things yourself (See my article on Learning to Learn for more on this).