Iurii Milovanov, SoftServe: How AI/ML is helping boost innovation and personalisation

Iurii Milovanov, SoftServe

Could you tell us a little bit about SoftServe and what the company does?

Sure. We’re a 30-year-old global IT services and professional services provider. We specialise in using emerging state-of-the-art technologies, such as artificial intelligence, big data and blockchain, to solve real business problems. We’re highly obsessed with our customers, about their problems – not about technologies – although we are technology experts. But we always try to find the best technology that will help our customers get to the point where they want to be. 

So we’ve been in the market for quite a while, having originated in Ukraine. But now we have offices all over the globe – US, Latin America, Singapore, Middle East, all over Europe – and we operate in multiple industries. We have some specialised leadership around specific industries, such as retail, financial services, healthcare, energy, oil and gas, and manufacturing. We also work with a lot of digital natives and independent software vendors, helping them adopt this technology in their products, so that they can better serve their customers.

What are the main trends you’ve noticed developing in AI and machine learning?

One of the biggest trends is that, while people used to question whether AI, machine learning and data science are the technologies of the future; that’s no longer the question. This technology is already everywhere. And the vast majority of the innovation that we see right now wouldn’t have been possible without these technologies. 

One of the main reasons is that this tech allows us to address and solve some of the problems that we used to consider intractable. Think of natural language, image recognition or code generation, which are not only hard to solve, they’re also hard to define. And approaching these types of problems with our traditional engineering mindset – where we essentially use programming languages – is just impossible. Instead, we leverage the knowledge stored in the vast amounts of data we collect, and use it to find solutions to the problems we care about. This approach is now called Machine Learning, and it is the most efficient way to address those types of problems nowadays.

But with the amount of data we can now collect, the compute power available in the cloud, the efficiency of training and the algorithms that we’ve developed, we are able to get to the stage where we can get superhuman performance with many tasks that we used to think only humans could perform. We must admit that human intelligence is limited in capacity and ability to process information. And machines can augment our intelligence and help us more efficiently solve problems that our brains were not designed for.

The overall trend that we see now is that machine learning and AI are essentially becoming the industry standard for solving complex problems that require knowledge, computation, perception, reasoning and decision-making. And we see that in many industries, including healthcare, finance and retail.

There are some more specific emerging trends. The topic of my TechEx North America keynote will be about generative AI, which many folk might think is something just recently invented, something new, or they may think of it as just ChatGPT. But these technologies have been evolving for a while. And we, as hands-on practitioners in the industry, have been working with this technology for quite a while. 

What has changed now is that, based on the knowledge and experience we’ve collected, we were able to get this tech to a stage where GenAI models are useful. We can use it to solve some real problems across different industries, from concise document summaries to advanced user experiences, logical reasoning and even the generation of unique knowledge. That said, there are still some challenges with reliability, and understanding the actual potential of these technologies.

How important are AI and machine learning with regards to product innovation?

AI and Machine Learning essentially allow us to address the set of problems that we can’t solve with traditional technology. If you want to innovate, if you want to get the most out of tech, you have to use them. There’s no other choice. It’s a powerful tool for product development, to introduce new features, for improving customer user experiences, for deriving some really deep actionable insights from the data. 

But, at the same time, it’s quite complex technology. There’s quite a lot of expertise involved in applying this tech, training these types of models, evaluating them, deciding what model architecture to use, etc. And, moreover, they’re highly experiment driven, meaning that in traditional software development we often know in advance what to achieve. So we set some specific requirements, and then we write a source code to meet those requirements. 

And that’s primarily because, in traditional engineering, it’s the source code that defines the behaviour of our system. With machine learning and artificial intelligence the behaviour is defined by the data, which means that we hardly ever know in advance what the quality of our data is. What’s the predictive power of our data? What kind of data do we need to use? Whether the data that we collected is enough, or whether we need to collect more data. That’s why we always need to experiment first. 

But I think, in some way, we got used to the uncertainty in the process and the outcomes of AI initiatives. The AI industry gave up on the idea that machine learning will be predictable at some point. Instead, we learned how to experiment efficiently, turning our ideas into hypotheses that we can quickly validate via experimentation and rapid prototyping, and evolving the most successful experiments into full-fledged products. That’s essentially what the modern lifecycle of AI/ML products looks like.

It also requires the product teams to adopt a different mindset of constant ideation and experimentation, though. It starts with selecting those ideas and use cases that have the highest potential, the most feasible ones that may have the biggest impact on the business and the product. From there, the team can ideate around potential solutions, quickly prototyping and selecting those that are most successful. That requires experience in identifying the problems that can benefit from AI/ML the most, and agile, iterative processes of validating and scaling the ideas.

How can businesses use that type of technology to improve personalisation?

That’s a good question because, again, there are some problems that are really hard to define. Personalisation is one of them. What makes me or you a person? What contributes to that? Whether it’s our preferences. How do we define our preferences? They might be stochastic, they might be contextual. It’s a highly multi dimensional problem. 

And, although you can try to approach it with a more traditional tech, you’ll still be limited in that capacity – depths of personalisation that you may get. The most efficient way is to learn those personal signals, preferences from the data, and use those insights to deliver personalised experiences, personalised marketing, and so on. 

Essentially, AI/ML acts as a sort of black box between the signal and the user and specific preferences, specific content that would resonate with that specific user. As of right now, that’s the most efficient way to achieve personalisation. 

One other benefit of modern AI/ML is that you can use various different types of data. You can combine clickstream data from your website, collecting information about how users behave on your website. You can collect text data from Twitter or any other sources. You can collect imagery data, and you can use all that information to derive the insights you care about. So the ability to analyse that heterogeneous set of data is another benefit that AI/ML brings into this game.

How do you think machine learning is impacting the metaverse and how are businesses benefiting from that?

There are two different aspects. ‘Metaverse’ is quite an abstract term, and we used to think of it from two different perspectives. One of them is that you want to replicate your physical assets – part of our physical world in the metaverse. And, of course, you can try to approach it from a traditional engineering standpoint, but many of the processes that we have are just too complex. It’s really hard to replicate them in a digital world. So think of a modern production line in manufacturing. In order for you to have a really precise, let’s call it a digital twin, of some physical assets, you have to be smart and use something that will allow you to get as close as possible in your metaverse to the physical world. And AI/ML is the way to go. It’s one of the most efficient ways to achieve that.

Another aspect of the metaverse is that since it’s digital, it’s unlimited. Thus, we may also want to have some specific types of assets that are purely digital, that don’t have any representation in the real world. And those assets should have similar qualities and behaviour as the real ones, handling a similar level of complexity. In order to program these smart, purely digital processes or assets, you need AI and ML to make them really intelligent.

Are there any examples of companies that you think have been utlising AI and machine learning well?

There are the three giants – Facebook, Google, Amazon. All of them are essentially a key driver behind the industry. And the vast majority of their products are, in some way, powered by AI/ML. Quite a lot has changed since I started my career but, even when I joined SoftServe around 10 years ago, there was a lot of research going on into AI/ML. 

There were some big players using the technology, but the vast majority of the market were just exploring this space. Most of our customers didn’t know anything about it. Some of the first questions they had were ‘can you educate us on this? What is AI/ML? How can we use it?’ 

What has changed now is that almost any company we interact with has already done some AI/ML work, whether they build something internally or they use some AI/ML products. So the perception has changed.

The overall adoption of this technology now is at the scale where you can find some aspects of AI/ML in almost any company.

You may see a company that does a lot of AI/ML on their, let’s say, marketing or distribution, but they have some old school legacy technologies in their production site or in their supply chain. The level of AI/ML adoption may differ across different lines of business. But I think almost everyone is using it now. Even your phone, it’s backed with AI/ML features. So it’s hard to think of a company that doesn’t use any AI/ML right now.

Do you think, in general, companies are using AI and machine learning well? What kind of challenges do they have when they implement it?

That’s a good question. The main challenge of applying these technologies today is not how to be successful with this tech, but rather how to be efficient. With the amount of data that we have now, and data that the companies are collecting, plus the amount of tech that is open source or publicly available – or available as managed services from AWS, from GCP – it’s easy to get some good results.

The question is, how do you decide where to apply this technology? How efficiently can you identify those opportunities, and find the ones that will bring the biggest impact, and can be implemented in the most time-efficient and cost-effective manner? 

Another aspect is how do you quickly turn those ideas into production-grade products? It’s a highly experiment-driven area, and there is a lot of science, but you still need to build reliable software on the research results. 

The key drivers for successful AI adoption are finding the right use cases where you can actually get the desired outcomes in the most efficient way, and turn ideas into full-fledged products. We’ve seen some really innovative companies that had brilliant ideas. They may have built some proof of concepts around their ideas, but they didn’t know how to evolve or how to build reliable products out of it. At the same time, there are some technically savvy and digitally native companies. They have tonnes of smart engineers, but they don’t have the right expertise and experience in AI/ML technologies. They don’t know how to apply this tech to real business problems, or what low-hanging fruits are available to them. They just struggle with finding the best way to leverage this tech.

What do you think the future holds for AI and machine learning?

I generally try to be more optimistic about the future because there are obviously a lot of fears around AI/ML. And I think that’s quite natural. If you look back in history, it was the same with electricity and any other innovative technologies.

One of the fears that I think does have some merit is that this technology may replace some real jobs. I think that’s a bit of a pessimistic view because history also teaches us that whatever technology we get, we still need that human aspect to it. 

Almost all the technology that we use right now augments our intelligence. It does not replace it. And I think that the future of AI will be used in a cooperative way. If you’ve seen products like GitHub Copilot, the purpose of this product is essentially to assist the developer in writing code. We still can’t use AI to write entire programs. We need a human to guide that AI to our desired outcome. What exactly do we want to achieve? What is our objective? What is our user expectation?

Similarly, maybe this technology will be applied to a broader set of use cases where AI will be assisting us, not replacing us. There is a quote that I wish was mine but I still think it’s a very good way of thinking about the role of AI: if you think that AI will replace you or your job, most likely you’re wrong. It’s the people who will be using AI who will replace you at your job. 

So I think one of the most important skills to learn right now is how to leverage this tech to make your work more efficient. And that should help many people get that competitive advantage in the future.

  • Iurii Milovanov is the director of AI and data science at SoftServe, a technology company specialising in consultancy services and software development. 

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The event is co-located with Digital Transformation Week.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

Leave a Reply

Your email address will not be published. Required fields are marked *