In the U.S., companies still rely on technology that is several decades old. Outdated accounting systems do not exchange data automatically, so finance departments manually reconcile figures across disparate programs for invoicing, payments, and taxes. Progress is slower than it seems. A significant portion of payments between companies is still made by check, and data from invoices is still entered by hand, line by line.
Oleg Morgoch, a Ukrainian software engineer with nearly 20 years of experience, calls himself an AI-Augmented Software Engineer and uses artificial intelligence (AI) every day to modernize legacy code in production systems on Microsoft’s .NET platform. At SoftServe, a global company with Ukrainian roots, Oleg has spent his career as a senior software engineer, mentoring colleagues and interviewing candidates up through the lead level. He cites the team’s trust in him as his greatest achievement. During one business trip, management told him personally that without him, it would have fallen apart.
A childhood passion for computers, along with his first games written in BASIC on a Soviet BK-0010 computer, led him to the profession. University, the engineer says, taught him the most important thing: how to learn on his own. That’s why, immediately after graduation, he mastered Oracle on his own, drawing on the database theory he had learned there.
Fear of Losing Control
“Managers are primarily driven by cost considerations, so any proposals for change run up against this wall,” explains Oleg. According to his observations, this resistance is largely psychological in nature.
It’s hardest to embrace change in places where people have worked for years without it and know their jobs well. A printing press operator with 25 years of experience keeps an Excel spreadsheet alongside a paper log. The proposed transition to a digital platform may be technically flawless, but the operator himself thinks quite differently: “If the system knows everything I know, then why do I need to be here?” What frightens people most in such moments is the loss of influence, status, control, job, and uniqueness.
The Role of a Navigator
The turning point came recently, with the emergence of tools like GitHub Copilot, integrated into the development environment. Writing and testing code that used to take hours now often takes just a few minutes, provided the task is clearly formulated in plain language. To explain this change to a non-technical person, he uses his own metaphor: “AI for an engineer is like a navigation system for a driver: it doesn’t know where you need to go, and it doesn’t replace the driver, but it helps you find your way faster, avoid traffic jams, and see alternative routes.”
The comparison holds because responsibility never leaves the driver’s seat. Finding the real cause of a bug buried in a large module used to take an engineer half a day of manual digging. With AI, the same search can turn up likely causes, possible fixes, explanations of the code, and test ideas within minutes. Whether a given fix is actually correct, safe, and free of new side effects remains a judgment only he can make. AI simply gives the driver better tools for reading the road ahead. He still has to drive.
The Cost of Delayed Decisions
Over the course of his career, Oleg has helped develop software for American companies in the real estate, oil and gas, and healthcare administration sectors, with results proven across a dozen projects. It is precisely there that outdated technology comes at the highest cost. He now combines this deep understanding of legacy systems with a measured approach to applying AI.
Companies often consider everything normal as long as it works technically. “Many people think that way, but over time they find themselves held hostage by this system,” he notes. Those who readily embrace change develop faster, while those who put off decisions for years are left behind. If he were given just one minute before a board meeting, the engineer concludes, he would highlight the lost time, the risk of errors, and the money that an outdated system quietly drains away.
When Experience Becomes Superfluous
Oleg describes one of the most dangerous new ideas in software development like this: “We no longer need senior developers.” AI can already produce code at the level of an experienced specialist, so some companies are hiring more junior developers instead of seasoned ones. For the first few months, everything looks great. The quality is indeed high. But quite soon, the architecture becomes chaotic, technical debt accumulates (cumulative code flaws that eventually slow down changes), and no one can fully understand the system anymore. The tool doesn’t know the hidden business rules, so the code compiles and passes tests while remaining dangerous in production.
The role of a mentor for junior developers is therefore more important than ever. Whereas before, they might have asked how to do something, now they need to learn to recognize whether it’s worth doing at all. Morgoch sums up his philosophy of collaboration with AI in a single phrase: “Think for yourself. Do it together with AI.” When AI writes code for an engineer, the person merely formulates the task and accepts the result. When AI merely enhances his work, however, the engineer remains the author of the idea and architecture, while the tool simply speeds up the search for solutions.
Technology or Problem?
According to the expert, the most common mistake companies make when first implementing AI lies in how they define the problem. They often start by asking, “How many programmers can we replace with AI?” instead of “Which development stages can we make faster and better with the help of AI?” The right approach is different: identify one specific process that takes the team 200 hours a month, then test whether AI can speed it up, reduce its cost, or improve its quality.
Morgoch can spot a company that’s implementing AI merely for show right away. They choose the technology before they’ve even defined the problem, and when asked about success criteria, they give vague answers like, “It’ll be more modern.” This, the engineer says, almost always means mimicking activity rather than achieving real results. Strong teams, on the other hand, set specific goals, such as reducing application processing time from 15 minutes to two minutes or increasing classification accuracy from 82 percent to 95 percent. Morgoch also draws attention to confidentiality: to prevent data leaks, his team has agreed not to include real customer information in AI queries, replacing it with test data.
In five years, Oleg believes, an engineer’s value will be determined by their ability to correctly formulate a problem, design the architecture, and verify the quality of the solution proposed by the AI, while the number of lines of code will take a back seat. “With the help of AI, a software engineer becomes a true architect,” Morgoch concludes. He considers AI a remarkably effective assistant for almost any task, always within reach. The engineer is convinced that, at the current stage of development, AI is not capable of completely replacing humans.
