We‘re not mad scientists anymore.

Jan Záluský / The Art of Smart


Artificial intelligence can reduce costs for companies and help them retain their customers. Its deployment is now affordable even for smaller businesses, says Tervel Sopov, Deloitte‘s director of AI & Data Advisory Services.

What does the Art of Smart mean to you?

To me, the Art of Smart is the ability to change and adapt. I‘m sure we all feel like the older we get, the harder it is for us to learn and adopt new things. And the same goes for companies. But if they want to be successful in the long term, they need to learn how to adopt and leverage new technologies in AI and data analytics.

You‘ve been working with AI and data professionally for over a decade. What’s your perspective on its evolution?

The development of the AI field is exponential, as was predicted by computer visionary Gordon Moore in the 1960s. Such growth is hard to grasp, so we are constantly surprised when something new appears. When the ChatGPT chatbot appeared, people started to wonder if this was the mythical general artificial intelligence that would destroy us. Obviously, it wasn’t and eventually, the next new thing will come and people will be surprised again.


How are companies in the Czech Republic and Europe reacting to new developments in AI?

After the introduction of generative AI, many companies had the same first reaction, which was to ban the use of AI just in case. Some companies are sticking their heads in the sand and do not have a clear position. Others are taking a proactive approach and enabling their people to make the most of these technologies. But the most notable phenomenon is that firms are coming under enormous pressure from their own employees, who for the first time in history are knocking on boardroom doors in a big way, saying they want to use AI to make their work more efficient. Previously, working with data was more the domain of experts and enthusiasts, but now anyone can test AI on their computer or mobile and naturally want to start using it at work.

As a result, the topic of using AI is making its way onto boardroom agendas, and company executives need to figure out how to approach the topic. Hopefully, more and more companies will establish an “AI officer” or another form of responsibility for this topic within top management. It doesn‘t have to be an expert who knows how to deploy and implement AI. It should be a person who will be an evangelist and translator, setting the course and keeping an eye on how their company develops in this area.


Is the Czech Republic ready for the AI boom?

The Czech Republic has a huge advantage because we have a number of qualified experts who are dedicated to this issue, which is not commonplace everywhere in Europe. Our work is sometimes complicated by the smaller size of our market. For example, in Poland, which is five times larger, it is easier to create a functional business case and implement it. That‘s why we see many companies group their teams in Central and Eastern Europe into miniregions or clusters. What few people know is that a number of these regional data Analytics Hubs sit in the Czech Republic, providing analytics services to other companies in clusters across the Central European region.


How can a traditional company that approaches AI with suspicion become a company that is helped by AI?

A company like this needs a multi-disciplinary solution, which should start with a strategy, ideally a company-wide one. And this should be followed by an AI and data strategy that tells the company how to use data and how to use AI technology. Then it needs to start addressing its culture and changing the mindset of its people.

This is usually one of the biggest nuts to crack and often requires a strong leader to be the change agent. Only after these steps are taken does it makes sense to address the technology itself.

What’s the best way to influence company culture towards AI adoption?

I think honest education works best. Not saying everything will be great and perfect, but talking openly about the things that didn‘t work and highlighting the things that did.

Many people may see AI as a threat because they can‘t imagine how to realistically work with it. Once they understand how it works, they can better see how it can help them within their company. Organizing internal “TEDex” events, where employees from all over the company are invited to participate, works very well. Apart from the awareness itself, this also stimulates an internal demand for AI. As part of implementing AI into the corporate environment, you also need to ensure that the different internal teams are working together and have common goals. A simple example is to imagine a retail company in which category managers are traditionally in charge of pricing, and the data science department of this company suddenly deploys a solution that solves pricing algorithmically. However, the goals of a category manager are traditionally very different from the metrics that data scientists track. And so, this company will quickly find itself in a situation where the two parties are competing internally instead of effectively joining forces. It sounds like a given, but you‘d be surprised how often this is an issue.

How can a client move forward with their own data analysis?

First and foremost, at this point of time, it is no longer an option you can choose to ignore. AI tools have been so widely adopted in the past months that it now sits on a similar level as Word, or Excel. You don’t have a discussion whether or not to use Word or Excel in your organization and the same is valid for AI.

But to answer your question, AI and data technologies helps clients reduce costs and it can help them with increasing their company‘s sales and profitability.

The interesting opportunity, however, lies in changing the company’s business model because of data and AI and coming to market with a product or service that is completely data driven. That, to me, is the most exciting moment. In any case, data analytics and AI can do a lot, but a company must be able to correctly react to the results. For example, you can quite accurately identify which of your customers are most likely to leave for a competitor. But part two is how to stop them from leaving. A nice example of this is the UK‘s Sorry as a Service brand, which will come to the rescue if you somehow let your customers down. It can deliver a personalized apology or a gift, which paradoxically increases their loyalty to your brand. For example, the Czech Rohlik brand works very well for us in this area, as it absolutely trusts its customers and compensates them automatically in the event of a complaint.

What are some of the myths you encounter when discussing the introduction of AI?

There are quite a number of myths in AI, and they come and go at the same speed as the letters A and I appear on LinkedIn profiles. I recommend, therefore, taking a healthy, critical approach to everything you hear around AI. An example of a recent myth is the fear that when using tools built on generative AI, companies are handing over data to the provider of the tool. While this is true in some circumstances, it doesn‘t mean that the tools in question can‘t be deployed in a way that keeps the data isolated at the client. Another example of such a long-standing myth is the highly exaggerated expectations, but this is often caused by the AI solution providers themselves, who promise amazing results but don‘t always deliver. You need to set realistic expectations. Yes, for example, personalizing an online store can lift sales by a few percents. But when a customer expects a six percent increase and, it’s “only” a percent and a half, that‘s a problem. But for a company with billions in turnover, even one percent will mean a huge amount of extra revenue, and it‘s a massive success.

Have you noticed a shift in the type of AI questions clients are asking?

Just a few years ago, it was not uncommon for us to have to introduce the whole topic of AI from the ground up so that the client understood what AI could do. Today, clients are much more educated, but it‘s still true that the devil is in the details, so we work best when we show them what‘s possible with specific examples from given market segments.

At Deloitte, you have been involved in the AI & data strategy, data science and machine learning domains for a long time. Has the perception of these domains shifted within your company?

Definitely. First of all, we are no longer a group of mad scientists who are locked in a room that everyone else is afraid to enter. The main change is how intersectional it has become. Historically, data analytics has been treated as a very isolated topic. A client would order a specific model, or niche AI solution. Today, on the other hand, AI is integrated into the vast majority of areas in which clients seek advisory services. Want to implement a new contact centre solution? You automatically want to have chatbots and voice-bots there already, and you also need to know how to prioritize and segment your customers.


How is chatbot technology changing?

Not so long ago, a conversation with a chatbot had to be programmed into a conversation tree, and once you started asking a question that the chatbot didn‘t have pre-written, you didn‘t get an answer. With the advent of generative AI, interacting with a chatbot has become much more natural and opened a lot of possibilities for using the tool.

A nice example, although before GenAI, is the system we put into operation a few days after the Russian military aggression against Ukraine. It was an intelligent contact centre that connected refugees seeking accommodation in Slovakia with accommodation providers.

A voice-bot was able to obtain information about the specific capacities and locations of accommodation facilities, find out the conditions the accommodation offered, and enter everything into a database. And all this without the need for human operator intervention. Such a thing would not have been possible a few years ago.

When will AI be indistinguishable from a living human?

That moment is already here. In fact, the biggest shift has come in the field of AI that is meant to mimic real humans.


This is especially evident in voice and text work. Although voice transcription technologies have been around for a while, the usability in the Czech Republic has not been optimal. Recently this changed fundamentally, and today it is very easy to provide transcription of the spoken word into text with great success. We’re seeing an even more fundamental difference in how AI manages to mimic the human voice. Nowadays, if you get a call from AI, you‘ll find it very hard to detect it. Because the way AI says things and the speech it gives is now very similar indeed to a real human. By the way, an unnamed radio station is already testing the introduction of synthetic presenters in the Czech Republic.

So, can we tell if it’s an AI voice on the line?

I‘m afraid it won‘t be possible purely from the AI‘s speech. Recently in the Czech Republic, there was the first case of an attempted attack that consisted of a company‘s CFO receiving a call from his CEO and being asked for details of the company‘s bank account. By making the cost of generating such traffic absolutely marginal there will clearly be more and more scams where criminals will try to use an AI voice to get your data. The solution, in my opinion, is to change the way we think. Until about 2000, it was automatically assumed that everything you see on TV is true. Gradually, society has come to accept that it doesn‘t have to be. A few years ago, the same thing happened with photography, and today most people understand that what they see in a photograph is not necessarily true.

Now we will have to learn that when we are on the phone with someone, it may not really be that person at all. The appropriate measure, in my opinion, is to get used to the idea that if the other party asks me for some sensitive information, I will do the equivalent of multi-factor authentication and call the person back through another channel. On the other hand, AI has the capability to analyse the human voice and detect if the person on the other end of the receiver is telling the truth.


That sounds really futuristic. What exactly can AI detect in speech?

It compares the behaviour of already detected fraudsters with the current call and watches for the same patterns to emerge. Thanks to advanced technology, you can evaluate many more parameters today than in the past. You don‘t just consider what the person has to say, you also look at how they say it, what their voice level is, what kind of pauses they make, how often they breathe. When you analyse all these parameters together, you can estimate how likely it is that the other party is telling the truth or lying.


You work with customers across the Central European region. Do your customers ever take AI inspiration from beyond its borders?

I like to ask clients if they are interested in inspiration from abroad. Personally, I find Asia inspiring. They have completely skipped the desktop era because of the mobile-first approach and are ahead of us in many ways. A beautiful example is their integration of social networks with sales channels. A customer there might watch a live stream of their favourite influencer using some cosmetics, and at the touch of a button on their phone, the customer can instantly buy that cosmetic. Research shows that two-thirds of Chinese consumers have purchased products via livestream. Now, let‘s put this in the context of generative AI and imagine that such livestreams wouldn’t even depend on a live influencer but would be created by GenAI.

Is there any danger of companies falling asleep at the wheel of their own digitalisation?

We recently looked at the benefits of companies adopting AI tools when Deloitte conducted an analysis among Fortune 500 companies. We analysed millions of documents to see how these companies‘ digitalisation initiatives correlate with their valuations. It was confirmed that investing in the adoption of modern technologies has a positive impact on the financial performance of companies and their valuation in the market in general. What was surprising about the survey’s results, however, is that companies that do not work well with technology feel it negatively on their bottom line. And this negative impact outweighs the positive potential, so not investing in AI is a bigger risk than a potential bad investment.

When implementing AI tools, are there specific differences among market segments, like banks, e-commerce and FMCG?

The difference is mainly in the level of data availability in each sector. While banks know almost everything about their clients, e-commerce knows significantly less, and FMCG manufacturers know almost nothing about their end customers and must rely on third- party analytics. This is why we are now seeing a trend on the market of increased collaboration between retailers and manufacturers across sectors. One gets the data on future demand and can adjust production accordingly. Others will increase the stability of their supply chains while monetising their data.

In the past, leveraging the power of AI has typically only been available to large companies with hefty budgets. What can AI offer small- and medium-sized businesses today?

Our industry has clearly opened up to small and medium-sized businesses, and smaller clients can now take advantage of pre-packaged, commoditised solutions. What years ago meant a long development process for a single customer and a large implementation project, today a client can literally get in a few clicks. And they can simply integrate the solution themselves. Furthermore, commoditised AI solutions mean that the client pays according to how much AI services they need, thus AI has become more affordable. The factor bigger than budget is now whether the client is “culturally” ready to start using AI.

What advice would you give to someone wanting to introduce AI tools into their business?

A client should make a strategic decision about which areas they will build their own competency in, and where they will rely on commoditised solutions. I‘ll give an example from the banking environment. Imagine a new bank coming to the market as a disruptor. You would think that this digital disruptor would handle everything about AI in-house. Paradoxically, it will be better off relying on commoditised solutions that it can buy cheaply to reduce the cost of market entry. In contrast, a traditional large bank is better positioned to experiment and be a true innovator in AI because it has the capacity and resources to do so. Certainly not everyone can afford to be in the position of blazing the trail for others. Fortunately, the availability of proven solutions is already abundant.

Any company that wants to venture into the world of AI should evaluate where the biggest opportunities are with AI and how much money can be made or saved using AI tools.

Are there any blind spots companies face when implementing AI?

The most common impasse is that businesses are tackling AI without strategic consideration. Things are being created in companies from below instead of the leadership setting a clear direction. Any direction is better than no direction in my opinion. In practice, just having the data team sitting close to the business teams they should be working with can advance the success of bringing AI tools into the business.

How do you think the world of AI will evolve in the coming years?

First of all, I don‘t think we will ever be done when it comes to AI. It‘s a never-ending process of finding new opportunities, and we can expect this process to continue to accelerate. So, if we move five years into the future, I expect AI and its application in companies will be further along than we can even imagine right now. But we‘ll still be dealing with essentially the same thing, how to make the most of the capabilities that are available.

About the Tervel Sopov:

  • Director, Deloitte Central Europe

  • As part of a regional AI & data team of 150 people, he is responsible for the AI & data strategy, data science and machine learning domains.

  • Tervel has been involved in the analytics, data, and artificial intelligence domain for over a decade and specializes in topics related to strategy and AI transformation.

  • He has spearheaded transformation and implementation projects for clients in the retail industry, e-commerce, FMCG, insurance and banking.