Alexander Goldstein: “Vibe Coding Is a New Reality That Could Reshape Software Development”

The IT market has always been dynamic, but over the past year it has shifted so dramatically that, as the clock struck midnight, we collectively found ourselves in a new reality. In this landscape, the role of programmers is being rethought, and vendors are actively experimenting with artificial intelligence across their products and internal processes.

We sat down with Alexander Goldstein, CEO of the ARGUS Scientific and Technical Center, Doctor of Technical Sciences and Professor at the Department of Information and Communication Systems at SPbSUT, to discuss how 2026 is beginning and what the next 12 months may hold. We talked about how vibe coding could reshape approaches to software development, how neural networks and workforce management systems are influencing the labor market, how customers can protect themselves from unreliable contractors and when a product can truly be deployed off the shelf versus when deep customization is unavoidable.

– Let’s start with a question that concerns many developers in Russia: how have taxes and IT support measures changed? And more broadly, how would you characterize what happened in 2025 and what lies ahead this year?

– Let’s be completely candid. It is relatively easy to analyze and label 2025 now that it is behind us. Forecasting 2026, however, is far more complicated from a macroeconomic standpoint, because much will depend on geopolitics and other factors that we can influence only indirectly. What is immediately visible is the continued role of the state in shaping the industry.

We appreciate the government’s continued support measures. This year, the zero percent VAT rate on software licenses was maintained, even though social insurance contributions increased. The introduction of a profit tax also appears reasonable. If a company generates income, it is natural to contribute tax payments that serve the broader public interest.

As for overall business activity, it is logical to expect some slowdown during a period of adjustment. At the same time, the elevated rates of 2025 have already prepared the market for tougher conditions. I do not believe the digitalization and IT development trend will stop. It is also too early to dismiss import substitution. It is no longer just a trend but an operational reality. Yet many solutions remain unreplaced, which creates a substantial field of opportunity for both customers and vendors.

– Alongside AI, one of the most frequently mentioned buzzwords last year was vibe coding. Do you agree that it defined the moment, and what would you add from your own experience?

– AI and neural networks are already familiar terms. The real question is the context in which we apply them. Vibe coding, however, represents a genuinely new reality. It has the potential to reshape how software is developed, and we are already experimenting with this approach, at least for rapid prototyping and early-stage validation.

– Do you believe vibe coding can ease the shortage of programmers and developers? Could it accelerate the creation of new systems and the evolution of existing ones?

– As AI itself – and Wikipedia will quickly tell you – the term only emerged in February 2025, introduced by Andrey Karpathy, co-founder of OpenAI. So we are still in the process of understanding what it truly means in practice. On professional platforms such as Habr, opinions are sharply divided, and the same polarization appears in industry discussions.

At this stage, we see vibe coding as a strong tool for prototypes or localized, one-off solutions that do not require scaling or strict security compliance. We use it in that context. At the same time, we clearly understand that serious enterprise systems are not built this way. Anyone claiming otherwise is either overstating the case or referring to a completely different class of solutions.

If you need a small internal tool – say, a quick reporting service for a CEO within a particular organizational structure – you can assemble something fairly quickly using this approach. But when you are developing a business-critical application that supports complex processes and hundreds or thousands of users, vibe coding becomes a questionable strategy. The risk that the solution will lack scalability, expose security vulnerabilities or prove difficult to maintain and evolve is simply too high.

There is another important nuance that often goes unmentioned. Vibe coding does not eliminate the need for human intelligence, and it carries hidden costs. The idea that you can simply describe your task to an AI and instantly receive a production-ready result is an illusion. In practice, you need prompt engineers. You need to understand the specifics of each neural network model. You must monitor hallucinations and validate outputs carefully. The expertise of an experienced developer cannot be replaced, even by the most powerful large language model.

Tenders, Vendors and Customers

– You mentioned that you focus on enterprise solutions. In this era of acceleration, do you build fully custom systems for each client, or do you aim to sell standardized, off-the-shelf licensed products?

– Any vendor would ideally like to create a highly universal product and simply distribute licenses at scale. In practice, however, there are many nuances. Our portfolio is divided into two broad categories: highly standardized products – for example, ARGUS WFM CC for contact centers, where customer profiles are relatively typical, processes across the market are similar and best practices are well established – and more customized products, such as ARGUS NRI or ARGUS FSM for mobile workforce management, where industry specifics and the client’s internal business processes must be carefully considered during implementation.

A complex and sometimes sensitive question inevitably arises: what should be standardized, and what should be customized? Take system entities or roles as an example. They can be hardcoded into the architecture, or they can be made configurable through reference directories and flexible settings. The logical conclusion is straightforward. If a product is expected to support numerous variations and frequent changes, it makes sense to build configurators and design-like subsystems. If variation is limited, hardcoding can be justified. These are the kinds of nuances we often have to explain to customers who may assume a system can simply be installed “off the shelf” and immediately put into operation.

In today’s environment, it is not enough for a product to be deeply embedded in a company’s processes. It must also adapt quickly to evolving requirements. If an IT solution’s development pace lags behind market trends, the organization faces a strategic choice: continue investing in outdated software, or allocate resources to upgrades that position the company ahead of market expectations.

– In your view, what role does the customer play in shaping the product? How much influence can a customer exert over a vendor?

– As software developers, we constantly balance two opposing forces. On one side, there is the principle that “the customer is always right” and defines the technical specification. On the other side, we understand that certain custom requests or attempts to replicate legacy habits can complicate and weaken the product. Custom functionality may affect core architectural components and, as a result, require significant resources for maintenance and further development. In other cases, customers push for highly specific features aligned with current management preferences. Leadership changes, those “must-have” features become unused, yet they remain embedded in the system, again increasing complexity and slowing future evolution.

Therefore, applying pressure to a vendor is reasonable when discussing timelines and realistic cost estimates, ensuring the software provider does not drift into academic perfectionism, unnecessary complexity or aesthetic priorities over functionality. At the same time, one of the key advantages of working with a vendor – especially compared to in-house development – is access to best practices and the accumulated experience of other users who have implemented similar solutions.

Trust and partnership between developer and customer are fundamental. If you demonstrate expertise in business processes, I will trust your understanding of how operations should function and will tailor global and domestic best practices to your needs, automating processes in a systematic way. In return, you should trust the implementation and avoid claims that “someone around the corner can build it with vibe coding five times faster, exactly according to my specification and ten times cheaper.” Every developer has heard that argument many times. What no one has seen – first and foremost customers themselves – is a real-world example where that promise actually holds up.

– In procurement tenders, companies often appear claiming they can quickly build a system of virtually any class or level of complexity. How can customers protect themselves from unreliable vendors, and what should they pay attention to?

– That is a fair question and a very familiar situation. In fact, it is really two questions. At the vendor selection stage, customers should avoid focusing solely on price and instead conduct pilots, carefully examining a company’s experience and actual capabilities. Beyond simply checking boxes in a technical specification, it is essential to assess real functionality in action.

As for protective measures, the vendor must properly document the code and transfer all necessary artifacts to the customer so that, if needed, the organization can involve its own IT team or assign the project to external contractors. It is also wise to secure in the contract the right to modify the software in the event of force majeure involving either the product or the vendor. And, of course, the standard recommendation still applies – evaluate the vendor’s stability, ensure they operate within a modern development environment and review their portfolio and reputation before making a commitment.

AI and WFM Systems in the Labor Market

– You mentioned that you work on workforce-related systems: one focused on contact centers and another on mobile field teams. How do your solutions help companies cope with labor shortages, and should businesses rely on AI?

– Unfortunately, our solutions are not a magic cure for workforce shortages. Let us take a closer look at where the most acute gaps actually exist. The situation is far from simple. Among the top five most understaffed sectors are manufacturing, transport and logistics, healthcare – which is particularly sensitive from a social perspective – as well as construction, retail and services. Each of these industries has its own workforce profile and operational specifics.

In some cases, workforce management systems can at least partially alleviate the pressure by distributing workloads more evenly among the employees who remain, reducing downtime and minimizing inefficient labor. In other cases, companies need AI-enabled tools that act as a copilot – helping employees reach required performance levels, handling routine tasks and identifying potential errors before they escalate.

As for AI itself, as I mentioned earlier, we actively explore and support modern approaches. We have run multiple experiments integrating AI into our platforms. For example, AI seemed like an obvious fit for workload forecasting modules. However, the results were sobering. Neural network–based forecasting turned out to be more hype than practical advantage. A classical mathematical model delivered better performance. Moreover, neural networks currently demand significant hardware resources, and research suggests that, at least for now, they function more effectively as supporting tools rather than as fully autonomous core models. That said, the landscape continues to evolve.

– So does that mean you remain skeptical about the narrative that AI will come in and take over everyone’s jobs?

– On the contrary, we are paying very close attention and asking ourselves where AI can be genuinely effective. For example, we are now seeing full-fledged voice experiences that move beyond simple speech-to-text and operate in a speech-to-speech format. These systems can capture intonation, emotional tone and semantic nuance. Together with colleagues from a partner company, we are developing a voice training simulator for the contact center market. It enables specialists across different industries to train for interactions with a wide range of customer types – aggressive, dissatisfied, offended, frustrated or even distressed callers.

We are also seeing solutions capable of parsing complex technical documentation. For us, this is particularly important. In our ARGUS NRI product, we long struggled with customers who stored large volumes of documentation on paper, which we had to enter manually. Today, that process can be automated: documents are scanned, uploaded and the model populates the database independently. So yes, we are open to innovation. At the same time, we do not adopt new technologies blindly. We develop serious enterprise solutions that handle large volumes of critical data. Experimenting recklessly is not an option, because once errors are embedded at scale, correcting them becomes extremely costly.

– Staying with generational theory and the challenge of employee retention: hiring and onboarding are expensive, and when people leave, companies lose competencies and institutional knowledge. How can WFM systems help in this context?

– In our systems, we incorporate elements of gamification to make work more engaging: competitions, game-like mechanics and KPI-driven incentives. Another emerging trend is short-term engagement. For example, in WFM for contact centers, it is possible to assign micro-shifts. A student might log in for an hour between classes and then log off. Similarly, in WFM for mobile workforce teams, workers can connect to one-off assignments. At one point, we even explored an “uberization” model, where management is structured not around fixed shifts but around pools of tasks that users select and complete for compensation.

When a company is already facing a labor shortage, implementing a WFM system provides clear visibility into who is doing what and how effectively. It helps assess skills, identify top performers and introduce performance-based bonuses through gamified mechanics. People tend to respond positively to transparent systems where they can see their progress, track their “five-star” ratings and understand how they compare. Consumer apps have conditioned us to value scores and rankings, and that psychology translates into the workplace.

And speaking frankly, deploying these systems reduces the amount of manual effort required for routine operations. That can sound unsettling, but in practice it often strengthens stability. It does not automatically mean layoffs. In many cases, roughly 20% of time is freed up and then reallocated to new responsibilities. Instead of merely keeping up with essential tasks, employees can take on additional initiatives, support new business lines or focus on higher-value activities. For prepared and loyal staff, that shift can be motivating. They see the impact of their work, and the company gains the confidence to invest in future growth rather than simply maintaining the status quo.

– Right now we see two opposing trends: companies bringing employees back to the office en masse, which often triggers dissatisfaction and even resignations, and on the other hand, preserving remote or hybrid work as a competitive advantage in the labor market. Can WFM systems help here?

– For contact centers, the ability for operators to work from home – or even from a car or the subway – remains highly relevant. That is why WFM CC includes a mobile personal dashboard and the ability to log in at different times, pick up tasks, monitor workloads and manage schedules on the go.

If I speak about our own company as a developer, we operate in a hybrid format. We keep the office, and many people prefer coming in because it is more comfortable for them than working from home. Others work fully remote and have actually increased their productivity. At the same time, we see that even when efficiency remains high, communication can suffer. That is why we encourage even the most committed remote workers to come into the office from time to time. It helps maintain psychological connection and team cohesion.

For companies to function effectively in a hybrid model, it is critical to remember that success depends heavily on employee self-organization and on monitoring tools that provide transparency into workloads, assigned tasks and overall performance. In that sense, WFM solutions are not just scheduling tools. They create visibility, structure and accountability across distributed teams. And in our experience, they perform very well in that role.