Todor Terziev on AI adoption in Bulgaria, the opportunities for e-commerce, and why bad business processes cannot be fixed by AI

AI adoption is moving quickly in conversations, but more slowly inside companies. Many managers know the tools, use chatbots themselves, and feel that they need to “do something with AI”. But turning AI into a real business capability is a different step.

For Todor Terziev, CEO and co-founder of Zenterra AI, this is one of the central challenges in Bulgaria today. Companies are interested in AI, but many are still unsure where it should fit, what problem it should solve, and what return they can realistically expect. During the interview, Todor referred to a cartoon he recently saw on LinkedIn that, in his view, perfectly captured the current state of the market:

“What do we want? AI. What do we want AI to do? We don’t know. When do we want it? Right now.”

For him, that summarizes much of the current AI wave. Interest is high, but many companies are still figuring out what concrete use cases actually create value.

“If you have 10,000 products that you have to optimize, follow, and manage, then our products shine.”

That is also why Zenterra focuses on sectors where volume matters. According to Todor, their most important sectors today are FMCG, logistics, and production. These are environments with many orders, many suppliers, many trucks, many warehouse movements, or many production steps. In such settings, small operatonal improvements can add up quickly.

What Zenterra actually offers

Zenterra works in two main ways. The first is custom digital and AI solutions. These start with a company’s problem, process, and data, and then move toward a tailored implementation.

“We find the problem, we see how we can help with process optimization and cost effectiveness, and then we look for a digital solution.”

The second part of the business consists of ready-made solutions that companies can use as a service. One example is a dynamic price matching system for e-commerce companies, which helps resellers compare suppliers, analyze previous sales, and decide where to buy products.

That is also where one of his strongest points comes in.

“No AI can make your old processes new again and better. If you have broken processes and bad data, no AI in the world can fix this.”

AI adoption in Bulgaria is still in its infancy

According to Todor, company-level AI adoption in Bulgaria is still relatively low. He estimates that only around 5 to 10 percent of companies have adopted AI in a way that is genuinely integrated into procedures and business processes.

At the personal level, usage is higher, but still limited.

“Most people use AI for their personal needs. But this does not mean the company uses it for its business needs.”

That distinction is important. Many employees already use AI to write faster, summarize information, or complete tasks more efficiently. But faster individual work does not automatically create organizational value.

“You do your job faster, but this does not mean you add more value to the company.”

According to Todor, the same issue exists at the management level. Many managers know ChatGPT, Claude, or similar tools and may even use paid versions themselves. But that does not automatically mean they understand where AI should be implemented inside the company or which business problems it should solve.

“Many CEOs and company managers have little to no experience with what the use case may be.”

For him, this is one of the main gaps in the market today. Employees may know how to use chatbots for personal productivity, while companies still struggle to connect AI to workflows, processes, and measurable business outcomes.

“The AI that the industry needs is AI that is integrated, well thought through in the company, and has a good return on investment.”

This is also why Zenterra often starts with a discovery and consulting phase before implementation. According to Todor, companies first need clarity on the actual problem, the operational bottlenecks, the available data, and where AI can realistically create value.

Only after that does it make sense to select or develop the right solution.

In that sense, the challenge is not simply AI adoption, but organizational adoption. The real value emerges when AI becomes embedded in how the company operates rather than remaining an individual productivity tool.

E-commerce is moving faster than many other sectors

When asked which sectors are moving fastest, Todor points clearly to e-commerce.

“E-commerce is currently at the forefront of adopting AI.”

That is not surprising. E-commerce companies already work in digital environments. They see new tools, platforms, and automation opportunities earlier than many other sectors. They also face constant pressure around visibility, pricing, customer service, fulfillment, and margins.

In Bulgaria, Todor sees e-commerce companies actively looking for ways to improve efficiency, visibility, and competitiveness.

“Everybody in e-commerce in Bulgaria is looking for some type of edge they can implement in their day-to-day service or operations.”

That competitive advantage can come from dynamic pricing, sourcing optimization, customer service automation, sales copilots, warehouse management, or fulfillment optimization. Many companies are small, often with teams of only two to five people, but still handle large numbers of orders and customer interactions.

This keeps his quote intact while clarifying the meaning in your own text.

This creates a specific opportunity for AI. Small teams can use automation to handle higher volumes without expanding headcount at the same pace.

Where AI adds value in e-commerce

For Zenterra, one important e-commerce use case is supplier-side price optimization. Many dynamic pricing tools focus on the customer-facing side: adjusting website prices based on competitor prices or demand signals. Zenterra’s approach, as Todor explains it, focuses more on the purchasing side.

The question is not only: what price should we sell at?

It is also: where should we buy from, at what price, and with what expected demand?

The system compares suppliers, brands, product portfolios, historical sales, and market signals. It then helps companies decide which products to order from which supplier.

“You save time and money. The system gives you the best price for the trending product from the best supplier.”

This is especially relevant when product and supplier complexity becomes too large for manual comparison.

“If you have ten products from two suppliers, why do you need us? But if you have 14,000 products, 186 brands, and seven suppliers, then you need some kind of system.”

That point is important. AI does not add equal value everywhere. It becomes more relevant when volume, complexity, and repetition make manual decision-making inefficient.

Not every AI solution is generative AI

Another useful clarification is that Zenterra’s dynamic price matching system is not simply a generative AI tool.

“It is built on a combination of structured data processing, algorithms, and AI-driven prediction models.”

According to Todor, the system combines traditional software logic with AI capabilities. Historical sales data, supplier information, and live market signals are analyzed to generate predictions, comparisons, and recommendations.

This reflects a broader point that often gets lost in current AI discussions. Many valuable business applications are not purely generative AI tools. Instead, they combine automation, analytics, prediction models, and AI-assisted interfaces to improve operational decision-making.

For e-commerce companies, that distinction matters. The value often comes less from the fact that something is “AI-powered”, and more from whether it improves buying, selling, stock management, customer service, or operational decisions.

Many e-commerce companies are still experimenting

Despite strong interest, Todor does not think most e-commerce companies in the region are already approaching AI strategically.

“I think they are mainly in the experimentation phase not to be left behind.”

That is a familiar pattern. Companies see competitors discussing AI, launching tools, or adding automation. This creates pressure to act. But acting without a clear business problem can easily lead to fragmented experiments.

“If my competitors have AI, I should definitely have AI as some kind of tool.”

For Todor, that is not enough. Companies still need strong fundamentals: capable people, efficient processes, competitive pricing, and reliable service. AI can strengthen those foundations, but it cannot replace them.

“If you have good people, good processes, and good prices, you are already in a strong position.”

The main barrier is not technology, but understanding

When asked about the biggest barrier to AI adoption in Bulgaria and the Balkans, Todor does not start with model quality or software availability. He starts with understanding.

“People need to understand how AI can help them. It is not a question of whether I have AI or not. It is how I can use it for my benefit.”

For Todor, the discussion should start with concrete operational problems rather than with the technology itself. The real question is where AI can improve workflows, reduce manual work, or help companies make better decisions faster than their current way of working.

This is also where government, industry associations, and ecosystem initiatives could play a role. Todor would like to see more practical workshops with companies in sectors such as production, logistics, and commerce.

“What AI or digital solutions can you implement in your company today and have visible results tomorrow? That would be a good workshop.”

For him, the discovery phase is crucial. Many companies need help identifying their own gaps before they can make sensible investment decisions.

The Bulgarian AI ecosystem is still taking shape

Todor describes the Bulgarian AI ecosystem as moving in the right direction. He points to local model development, data center infrastructure, strong software talent, and the broader digital infrastructure of the country.

One development he mentions is BgGPT, a Bulgarian language model developed by the Institute for Computer Science, Artificial Intelligence and Technology. In his view, such models can support local integration, language-specific applications, and use cases in education, business, healthcare, and public services.

“BgGPT is trained specifically for the Bulgarian language and allows the creation of applications for Bulgarian education, businesses, and healthcare.”

International models such as ChatGPT, Claude, and Gemini also work well in Bulgaria, according to Todor. But local models may better understand expressions, local language patterns, and specific institutional needs. They may also be cheaper and easier to integrate into national systems.

This is not only a technical issue. It is also part of a broader question: how much of a country’s digital infrastructure should depend on global private AI platforms, and how much should be developed locally or regionally?

Bulgaria has strong technical talent, but AI education is still developing

Bulgaria has a strong reputation in software development, engineering, and technical problem-solving. Todor sees that as an important advantage.

“Bulgaria is well known for producing IT specialists, coders, problem solvers, and engineers.”

He mentions technical universities and institutions in Sofia and Plovdiv as part of that talent base. For software development, machine learning, and engineering, he sees good foundations.

At the same time, he is more cautious about dedicated AI education.

“I am not quite sure if we have university programmes directly into AI programming and training. Maybe some courses, yes, but it is not a well-known programme.”

That is a familiar issue across Europe. Many countries have strong software and engineering talent, but the educational system is still catching up with the speed of AI development. Much of the current skill development happens through self-study, private courses, company learning, and hands-on experimentation.

Mapping the AI ecosystem remains difficult

One challenge in the Bulgarian AI ecosystem is simply knowing how many AI companies there are. Todor is skeptical that a clear number currently exists.

“There is no formal certification for being an AI company. Basically, any two people can register a company and call it AI.”

That creates a measurement problem. Some companies build serious enterprise-grade systems. Others provide simple automations, resell services, or use AI mainly as a marketing label. In a market that is still developing quickly, the boundaries of the category become blurred. Some companies build serious AI infrastructure and products, while others mainly position themselves around the AI hype.

“It is fashionable to be an AI company.”

For RankmyAI, this is exactly why ecosystem mapping matters. The challenge is not only to count companies that use the term AI, but to distinguish between different types of companies, solutions, business models, and levels of technological depth.

Todor expects the market to become clearer over time.

“In the next couple of years, it will become clearer which companies actually provide real value and therefore remain on the market.”

Platforms such as RankmyAI can help bring more transparency to this fast-growing ecosystem by distinguishing between hype and companies that are building sustainable AI solutions. Check out the Bulgaria AI Tools Ranking, where we track the leading AI companies and tools shaping the country’s rapidly evolving AI landscape.

Government support could focus more on discovery and adoption

Todor does not see strong direct government support for companies like Zenterra. There are AI and digital innovation hubs in Bulgaria that help smaller companies experiment with digital solutions and develop MVPs. However, according to Todor, these initiatives mainly support companies looking to adopt AI rather than companies building AI solutions themselves.

He would like to see more initiatives that bring AI specialists and companies together in practical formats.

“If the government says this is our programme and we invite these specialists as workshop leaders, that would be good.”

In his view, government support could help companies finance discovery phases, understand where AI can add value, and then co-finance implementation where there is a clear business case.

“If the discovery stage is financed, more companies would want to understand where their gaps are and which AI solutions they can realistically implement.”

That reflects a broader challenge in the market. Many companies are not yet ready to invest heavily in AI, but they are interested in understanding where AI could improve their processes and operations.

Fully autonomous AI agents remain far away

Looking ahead, Todor expects more development toward AI operating systems and personal assistants for business. He uses the familiar image of a Jarvis-like system that gives companies business analytics, automation, and support through a single interface.

“I think we are going more into the area of AI operating systems.”

However, he is skeptical about fully autonomous agentic AI in commerce, at least in the near future. He expects companies to keep humans in the loop.

“The agentic AI will almost match human interaction, but I think most people will try to keep the human interaction in the loop.”

His concern is not only technical reliability. It is also strategic. If AI systems make all decisions, then differentiation may depend mainly on who can afford or train the strongest AI system.

“If AI does all the things for you, how do you position yourself in the market?”

That question goes beyond e-commerce. If AI systems increasingly select, buy, sell, and optimize automatically, companies will need to rethink where human judgment, brand, process knowledge, and strategic differentiation remain.

Zenterra’s own future plans

For the coming years, Zenterra wants to strengthen its position in Bulgaria, especially among middle-sized and larger companies.

“We would like to establish our market presence in Bulgaria.”

The company currently has a core team of seven people and expects to grow when new markets and solutions require additional expertise. Todor says Zenterra will focus more on enterprise clients and serious implementation projects, while still offering subscription-based solutions where relevant.

The company also has international ambitions. In 2027, Zenterra aims to expand into the Balkan region through partners.

“We will try to expand to the Balkan countries and find partners there for implementation of some of our services and projects.”

That makes the company part of a broader pattern we also see in the AI landscape: AI companies trying to move beyond their domestic market by combining local implementation knowledge with scalable digital solutions.

Start with the why

At the end of the interview, Todor gives simple advice to anyone wanting to start an AI company in Bulgaria.

“Start with the why.”

For him, starting an AI company should not begin with the technology. It should begin with a reason, a niche, and a concrete problem.

“Find one specific problem that you can solve with your solutions.”

That advice reflects the central theme of the entire interview. AI adoption is not about adding AI for the sake of AI. It is about understanding processes, identifying bottlenecks, and using digital systems where they create measurable value.

“People would like to pay you money if you can solve their problems.”

In a market where many companies are still experimenting, that may be the most important point. The companies that survive will not be the ones that simply call themselves AI companies. They will be the ones that solve real problems, with clear value, in contexts where AI is actually needed.

About Todor