After ChatGPT entered mass use, we took the next step toward widespread AI implementation, and it created a massive buzz around the technology. There have been other events, of course, but I see this one as the most significant. Today, it’s hard to find a tech (or even non-tech) company that doesn’t have the abbreviation "AI" in its name, description, or public messaging. For the average user, all of this feels like a sudden miracle. Just yesterday, we were using Google or Wikipedia to look up information, and now everyone simply asks the chat what to do.

On the day I was writing this text, my daughter was out playing with a friend, and they found spider eggs. In the morning, she told me about it. When I asked how she knew they were spider eggs, she said they took a picture and asked ChatGPT what it was. User behaviour patterns have changed forever, and it happened instantly. Perhaps in the future, there will be a different way of interacting with AI, but the fact that AI will be part of our lives and that we’ll rely on models—there’s no doubt about that.

As with any other revolutionary technology (and AI is certainly that), we’re witnessing how it changes the world around us. Technological breakthroughs usually change the world for the better, but there’s always a flip side. One of the downsides is that some industries—or parts of them—are heavily transformed or even cease to exist. When machines were introduced, working conditions and factory jobs changed. When automatic switchboards came along, the profession of the telephone operator disappeared. So, what will change with AI? I’m not writing this article to answer that question—it’s a massive topic, and besides, you can ask ChatGPT 😊. Content is being devalued; many things are changing. We’re seeing a big impact on intellectual professions, and I believe that the effect on manual labour isn’t far off either.

As an owner and leader of several IT businesses, I’m primarily interested in how AI will affect my companies. I personally divide AI’s impact on the IT business into two areas: product-based business and service-based business. In the product world, the impact is huge—especially in the context of generative AI. One day I’ll write more about that. But in this article, I’m focusing on the impact of AI on the service business, software development, and the profession of software engineers in general.

Every day, we see numerous posts, articles, and interviews with IT company leaders claiming that AI will partially or fully replace developers in their companies. As of writing, it’s been just 2–3 days since the CEO of an AI startup—one that develops generative AI for code—stated that AI would write 90% of code in the next 3–6 months, and 100% within a year. That statement caused a stir, and several friends and colleagues sent it to me. The IT community is split in its views. There are AI optimists who believe that tomorrow, AI agents will replace humans in team meetings. And there are skeptics who think otherwise. I count myself in the second group, and I’ll explain why.

While for the world at large AI appeared rather suddenly (especially in historical terms), for the IT industry it’s not new. We’ve been using various AI models and training them to solve different types of problems for a long time—mostly things like recognition, prediction, classification, etc. With the rise of generative models, the range of applications has expanded dramatically, and we’re seeing many new tools emerge. Tools—that’s the key word. I’ve always viewed AI as a tool in the hands of the user. In our context, the user is the programmer, and now they have more powerful tools in the form of generative models. But as we used to say at my university, "Technology in the hands of a savage is just a piece of metal."

We’ve had powerful tools for a while—ones that let us write code quickly and efficiently. These include various assistants built into our development environments or installed as plugins. Development speed has been increasing steadily, especially in the early stages when we’re working in a so-called "green field"—when nothing has been written yet. With AI, this process speeds up even more. A model can generate lots of code, even create the first version of a product, deploy it to a server, and do many other tasks.

The path from idea to working prototype has truly shortened—and will continue to do so. This first generative phase is what’s currently experiencing the most disruption, and it’s capturing people’s attention both inside and outside the industry. But as the VP of R&D at one of my companies says, building the first working version of a product (a real working version, not just a prototype!) is the easier first 90% of the project. The harder 90% comes next. Only those who’ve brought products from inception to market will truly understand this.

So yes, in the beginning, AI can dramatically boost speed. But over time, that speed may decline. Code has complexity, and as a codebase grows, so does its complexity. If left unchecked, this complexity can spiral out of control and kill the project with technical debt. We may reach a point where making a simple change is so hard that it’s easier to rewrite a module—or the entire project. A decently functioning piece of software used by many people is made up of tens of thousands, hundreds of thousands, or even millions of lines of code. If we hand this over to AI entirely and just keep feeding it feature requests, we’ll quickly find ourselves dealing with an uncontrollable mess.

Beyond the technical challenges I see in using AI, there are also social ones.

Anyone who has worked in teams knows that software development isn’t just about coding. Coding is a small part of the process—especially once the product is on the market. Then, other processes kick in: data collection and analysis, planning, design, prioritization, and so on. The Software Development Life Cycle (SDLC) is a complex and ongoing process with a cyclical structure. Many people are involved, and their initiative drives the next steps. AI doesn’t have initiative.

So, I believe that generative AI—both now and in the future—is a powerful tool that will enhance our productivity and change some of our approaches, but it won’t fundamentally transform the software development process.

Tomorrow, we’ll write code faster, but it will still need to be reviewed, managed, integrated, and tested—and that work must be done by humans.

I currently don’t see any way to fully remove people from this process. What I do see are changes to the programmer’s role.

These changes will affect their daily tasks and responsibilities. In my view, the role of a developer will expand, and coding will no longer be as central as it once was. Yes, you’ll still need to know how to write code—but knowing how to read it will become far more important.

By this, I mean reviewing AI-generated code for:

• quality,
• complexity,
• task alignment,
• integration with the project.

Tasks that were once the domain of team tech leads and architects must become part of every developer’s routine.

Let me remind you: a tech lead or architect is the highest rank and greatest responsibility among hands-on engineers. And this, in my view, is the most challenging shift our industry will face. It takes years to grow into a tech lead or architect role—and many people never make it. To rise to that level, you need to understand far more than an engineer who’s focused only on backlog tasks. You also need initiative and a curiosity about your field that goes beyond day-to-day work.

Now, we—the leaders of software development companies—face an ambitious challenge:

To turn all our engineers into architects who, armed with modern tools, become many times more productive.

How’s that for a goal? 😊

I’ve been thinking about this challenge for several years now. In the early stages, during all-hands meetings with my engineers, I urged them to follow AI developments and adapt to changes in the industry. At the same time, I realized I didn’t want to give them abstract advice like "Just keep learning and everything will be fine." I wanted to share concrete steps.

Over time, I developed an idea that I called the Engineering Backbone. I’ve gone from Junior Software Engineer to System Architect, CTO, Head of Sales, and CEO. That’s a full cycle—both an engineering and business career. At every stage, I learned a lot, and I always started by turning to the classics of the field I was working in. That always paid off. As a result, I now have a large library of books and authors across various areas of IT and business.

That inspired me to turn Engineering Backbone into a course—a kind of modern education for IT professionals. The course is primarily intended for developers, but I recommend parts of it to anyone working in tech, including non-technical roles like HR, Sales, Marketing, etc.

I made the decision to invest in the additional education of my people, so they would be:

• ready to use the latest tools,
• well-rounded,
• prepared for the challenges and changes AI is bringing to our industry,
• and competitive in the job market.

As mentioned earlier, Engineering Backbone is a course I’m creating (still in progress) for employees of my companies. It consists of ten directions, listed here in English:

In each direction, I select 3–5 books that make up that section of the course. My team and I are also developing self-assessment tests to help people gauge how well they’ve absorbed the material.

Now let’s do the math:

• We’ve built a course that is primarily based on technical literature
• It has 10 sections
• Each section contains, on average, 4 books.
• As a result, engineers need to read 40 books to be prepared for the challenges of today.

And this is just the basic level that forms the core professional knowledge. This is the approach we’ve chosen to invest in the future of our business. I don’t yet know what results it will bring, but I wanted to share my thoughts and approach here. We’ve already finalized the materials for several sections of the course, and we’re actively working on the rest. The selection of books and materials is very deliberate: we test 10 to 20 books before choosing the ones to include. It’s a large and painstaking process. When the entire course is complete, I plan to write another article that describes in detail all the materials we’ve selected.

AI is changing our industry—and fast. I believe that people will remain at the center of software development, but their roles will evolve. The entry threshold into the profession will rise and become more complex.

To me, that’s good news, and we’ll soon see a new IT industry where human–AI interaction is more integrated than ever. But human responsibility will only grow.

I’m glad to share these thoughts in this article.

I hope it sparks reflection in someone—whether your reaction mirrors mine or is completely opposite 😊.

I would be very happy if someone adds more important topics to my 10 core directions—or shares suggestions for materials that might strengthen the course.