Article content
Technological developments often trigger long-lasting societal changes. The industrial revolutions of the past gained momentum from the introduction of new, ground-breaking innovations. In this historical phase, so it is happening with artificial intelligence and machine learning. There is an increasing dedication in Estonia to deciphering how to make these tools work across various areas of life. And as the consequences matter for productivity, work ethics, and living standards, a comprehensive vision for the future is what could make the difference.
Estonian company MindTitan is already a few steps ahead in this sense. Specializing in AI and machine learning development for businesses, they have detected a series of use cases for these solutions in a variety of sectors. From telecommunications to banking and finance, from pharmaceuticals to retail and e-commerce, the applications are limitless. Such an encompassing view denotes a deep knowledge of the topic at hand. But above all, this indicates that the present and future of business development lies in the competitive advantage that companies can gain from an adequate implementation of machine learning tools.
Markus Lippus is the Co-Founder and Data Science Lead at MindTitan. International experience with partners in Europe and the Middle East, as well as a successfully completed project with Estonian telco Elisa, figure already in the company’s portfolio. In this insightful interview, Lippus explains what AI is set to change and how companies can take advantage of it.
Markus Lippus, Co-Founder and Data Science Lead at MindTitan
From buzzword to effective application. Do you see a growing trend in the deployment of AI and machine learning tools in private businesses?
There is definitely a growing trend in the use of AI and machine learning in the private sector. It is driven by the fact that businesses feel the growing pressure of using AI, as the hype progresses. Also, tools and methods are becoming easier to use – and more effective at the same time. As more companies embrace machine learning, they generate more success stories further motivating others to come on board.
A lot of new research in AI originates from the research groups of tech giants such as Facebook and Google. It’s likely that current applications have a lot to do with ads optimization, and mining your pictures on social media for data profiling. Another set of companies working with this bleeding edge (author’s note – the use of latest technologies carrying high implementation risks) would be the ones dealing with self-driving vehicles – both hardware and software carry risks and advantages.
But as researching continues, such technologies find applications even more rapidly in more forward-looking companies. For example, it’s a standard procedure in our work to keep an eye on the outcomes to see if we could improve what we already have. It’s a fast-evolving field, we can only help people be competitive by staying competitive ourselves.
Beyond chatbots, on which segments of a company’s business process can AI have a significant impact? Where do you see more attention being focused on?
I’d say chatbots are a quite boring way to use machine learning. Their main benefits, as a first use case of machine learning for a company, are accessibility and ease to apply them to many different verticals. For most businesses, there are far more useful ways of using machine learning.
The biggest impact will probably unfold in industry-specific areas, like predictive maintenance in manufacturing and businesses employing heavy machinery. Solutions able to understand the patterns in cellular network data for telecoms fall into this category too. Anti-money laundering tools for banks are part of the picture, as image recognition and decision support in medicine.
But we see most of the attention on machine learning heading towards either well funded (private sector-driven) or very cool (academia-driven) areas. There’s generally a lot of work to do in reinforcement learning and generative models, which have limited practical applications. The same applies to recommendation engines and image recognition, offering wide and lucrative application areas in many verticals.
These days, machines can provide a high level of personalization using users’ habits, information profiling, and general background data. This kind of personalization is necessary for most customer-facing companies. Aims may vary, from holding users’ attention by providing relevant content, to recommending specific products and services to increase revenue.
As natural language processing rapidly develops, it will soon become easily applicable to many businesses with more low-risk results. Also, with the recent advancements on increased accuracy, decision makers – who are on the fence now – will be more prone to bite the bullet and start using NLP for advanced content analysis, text classification and better conversational interfaces.
Which sectors of the economy are you looking forward to work in?
We mostly aim at telcos and financial institutions, purely based on previous experience – we simply know how to provide value for these verticals the best. Use cases then are applicable across many industries, such as for personalization solutions, and customer services by text or voice.
I think that in some areas we’re actually reaching the point where, by using machine learning, you’re not a front-runner anymore. But you do find yourself way behind if you aren’t. Imagine, for example, launching an on-demand video service without content recommendation in a world where 75% of Netflix’s views are driven by the recommendations they make.
Let’s move to your own experience then. Where can the implementation of machine learning provide businesses with a competitive advantage over less innovative ones?
In most areas, machine learning is still very much an early adopter thing. Machine learning is developed at a high level in tech companies, which is what most businesses are not. If we’re talking about industry-specific solutions, like loan performance ratings or network performance estimation for telcos, it’s very likely you’ll still gain a competitive advantage over others. But these are custom work areas, entailing the existence of a data science section in your company. Not many have them, and those who do keep saying they’re overworked with BI and data science in marketing.
More generally, I do believe some areas present higher returns than others:
- Lead scoring – as telemarketers on average just call everybody. You mess with customer loyalty and brand image every time you pester someone, and the success rate is really low. We’ve seen machine learning increase of 2.5 times the effectiveness of this process, which means calling fewer numbers to get equal results. You also get better results and annoy fewer people, as more of them manifest interest in the products you sell.
- Personalization – it’s big in tech companies and some web stores (Amazon reportedly sells 35% of products based on recommendations). However, there’s a lot of interesting work going on in personalizing e-mails, on-site ads, and even webpages themselves, surfacing items the person is more likely to engage with.
- Text classification – whether it’s about routing incoming customer contacts, parsing internal knowledge bases, or surfacing relevant information at the right time. Working with text can dramatically increase the productivity of employees on many levels.
Technological developments in business processes can also trigger occupational change. In your opinion, how AI can impact the distribution of tasks and jobs within a private organization?
I think this depends a lot on where the company is located – in Estonia, for example, there is a structural lack of labour force. When you create a machine learning solution that makes your employees work more effectively, you gain more time for them to work on things you can’t or don’t want to automate. Who says that human interactions can’t be a part of your brand?
Machines are a lot less smart than people imagine. In the near future, I don’t think that machine learning is going to displace many workers. It’s more about automating the routine part of a job, so people can do what only they can do.
Of course, more disrupting changes will take place in the future, such as the widespread use of self-driving cars. However, implementation is getting more complicated than companies expected it to be. Add legal hurdles and huge capital costs around it – the rollout seems likely to be gradual than all at once. This will alleviate the issues that it would otherwise introduce.
