Dobkin: Many talk about how many jobs AI will make redundant, but how many new tasks and jobs will it allow us to create?

In the world of AI, bold statements can very quickly become part of history.

Here's one of them: "By the end of this decade, I foresee mass unemployment among American programmers, system analysts, and software engineers."

Sounds very contemporary.

But in fact, this was written back in 1992 in the book "Decline and Fall of the American Programmer" by Ed Yourdon – one of the most influential thinkers in the field of software of his time.

He was a pioneer of structured analysis and object-oriented design, and one of the first to be inducted into the Computer Hall of Fame.

The logic of his reasoning seemed very convincing: globalization, the growth of human resources in India and other countries, as well as effective CASE tools for code generation, were supposed to make it possible to create cheap "software factories" that would cover most of the software needs of that time.

This book, by the way, became one of the impulses for the founding of EPAM in 1993. It was a signal for us that the global software development market was changing and opening up new opportunities.

A few years later, Ed wrote "Rise & Resurrection of the American Programmer", where he revisited his unfulfilled prediction.

Undoubtedly, globalization did indeed impact the industry. Automation and industrialization did too.

But the prediction proved wrong, as it was based on the assumption that job categories would remain the same and companies would perform the same work, only cheaper, faster, and with fewer people.

In practice, however, the job categories themselves changed. New tasks emerged, which were impossible to predict. And they appeared faster than automation managed to displace old tasks.

Thus, the "software factory" did not bury professions. It shifted value from writing code to creating platforms, integrating systems, understanding business context, unraveling complexities accumulated in old IT systems, and achieving concrete results.

Today we hear similar thoughts again. Now we're talking about "dark factories," where AI agents will create software with minimal human involvement. I agree, much of this will indeed come true with time and certain effort.

But most of what we today call "AI-first" is still focused on improving existing work: faster code writing, modernizing outdated applications, platform migration, automating support. And this is also very valuable.

However, this is not yet a new category of work. These are predominantly old categories, simply with much more advanced tools and the promise to perform the work significantly cheaper and faster.

Perhaps the real question today is not how much old work AI will automate or how many jobs it will eliminate. But rather, how many new ideas, solutions, platforms, and IT businesses will be created thanks to new categories of work, for which we haven't even coined names yet.

I don't think we know the answer yet or are capable of making accurate predictions.

But I suspect that value will once again shift to the "last mile" – transforming general AI capabilities into specific corporate solutions: reliable, secure, integrated, understandable for evaluation, and truly used by people.

This "last mile" will become much broader, more complex, and more influential. Which means great opportunities are opening up for us once again.

In 2008, I had dinner with Ed and shared with him how his book helped kickstart EPAM by showing us opportunity where others saw an end.

He signed my copy of the book. I still keep it today.

• 12
• 0
• 0
• 4
• 1
• 0
• 1
• 0