Our comprehensive solutions and services guide you on your digitalization journey.
Finding the right business solution:
We speak your language! Our industry-specific solutions are perfectly tailored to your needs.
Gartner® Report: Transform Your CRM with 5 Essential Steps
Tips and updates for our solutions, informative blog posts, compelling case studies, and news about our company.
BE-terna achieves the 2023-2024 Microsoft Business Applications Inner Circle award.
7 min read • May 07, 2020
Machine learning is already disrupting our everyday lives and brings great promise to businesses. In this interview we asked Milan Listeš, Customer Landing Manager, about all the great benefits businesses can expect from leveraging AI, about the ongoing AI projects at BE-terna, and the importance of the data diagnostic project phase, which can help predict the outcome of AI projects.
One of the questions I often hear today about AI is “Why now?”. In my opinion, cloud popularisation is one of the key aspects which has moved things forward in this area. It has allowed companies to be quite cost effective while piloting various solutions and trying different approaches with their data. Furthermore, data gathering has led us into a situation where we have a lot of data, and those amounts can’t be properly analysed with standard reporting and analytics tools. However, it is not only data gathering that has led us to this situation, the issue is that many problems being solved by machine learning today are too complex to be handled (properly or optimally) manually. Even a process of medium complexity, such as ordering items in a distribution company, includes so many inputs and restrictions that it becomes difficult for a person to handle, if that person is responsible for making orders for 100 items, for example.
There are lots of buzzwords in this area but the important message to businesses is to put business issues in first place and then the technical buzzwords should be part of the solution to those problems. In general, there are quite a few data-intensive business problems in every industry. It’s hard to put them all under one umbrella, but they would include: prediction/estimation, optimisation, detection/recognition, and personalisation. These shouldn’t be viewed from a technical perspective, because not all of them are always an AI problem, some can be solved by simple reporting, analytics or improved processes, so they should be viewed as business issues which can be applied to various industries.
The future is already here – AI is disrupting every industry and every business process. AI has great potential not only to change our business, but also to change our personal lives – from the way we spend our holidays, order food, and drive cars. AI has great potential to revolutionise the economy, and I am very much looking forward to seeing this in the future.
If we look at industries which are likely to benefit most from various AI projects, there are some that are investigating the potential benefits more than others. Primarily we can mention all consumer-related industries, because of personalisation trends which are really data-intensive, along with financial services who will have to rely heavily on customer personalised targeting in the upcoming years, and also the whole supply-chain - meaning manufacturers, distribution and retail. If you take a look into the estimation of potential benefits done by McKinsey in various industries(1) you will see that depending on the industry, there are potential benefits of approximately 30%-40% in image recognition and some other deep learning algorithms of greater complexity.
A case we are currently dealing with most often is demand forecasting and stock optimization. This is a perfect fit in order to realise business benefits with AI and analytics. The core issue in this business problem is trying to estimate demand for your items and trying to do that as well as you can for different types of items; for example, seasonal items, those with steady sales, new products, promotional items etc. The demand forecasting problem is solved utilising machine learning, whereas the second part of the problem, the purchase recommendations, is solved using multi-objective optimisation techniques. The result is the near optimal order quantity for each analysed item. This process provides two major benefits - firstly preventing stockouts which would lead to lost sales, and and secondly minimising overstocking, which would lead to poor cashflow. Moreover, since the recommendations are automatic, it enables the purchasing department to focus on more critical items that are usually not a part of the forecast (e.g. expensive items). Obviously, this case can be applied to every business where warehouse and stock optimisation play key roles in success, manufacturing, distribution, retail etc.
Firstly, it’s important to understand how these projects are positioned from an IT investment perspective. Projects like ERP, CRM and BI in most (thought not all) cases are part of the lower quadrants(2) called “Key operational” and “Support”, meaning that they are solid bedrocks which your business is built upon, but having them means you are only avoiding disadvantages in the market. For example, if you are reselling plane tickets and don’t have a web portal, it’s almost as if you don’t exist in the retail market. And those solutions must be very well embedded and integrated into your company’s IT portfolio.
On the other hand, AI projects which manage to be well embedded in company processes can really create advantages in the market and give companies the opportunity to transform their business and create added value for customers. However, at moment zero, when companies are just starting to think about them, they are positioned like “High potential” projects which may lead to success in the future, but their outcome is still unclear. This means that companies often approach such projects in a “quick and dirty” manner. At BE-terna that “quick and dirty” approach is called Data diagnostics, which is a roughly 1-2 month phase in which we deeply analyse the potential of the data a customer has. If potential is shown, then the customer has a much clearer picture of the benefits they get from a real project and the likely ROI.
And finally, entering a full project is also a little different to those we may be used to. One McKinsey analysis(3) showed that one of the key differences between successful AI projects and those that were stuck in the pilot phase is in terms of the time and money spent in the “last mile” of the project, meaning investment in process redesign and education. We’re currently experiencing exactly the same situation, our project ‘Hyper Care’ has turned out to be perhaps the most important project phase in which we have worked closely with customers, fine-tuned the solution and helped them to build their confidence in the solution.
In general, there are two directions visible now that will play an important part in the future. One of them is additional (third-party) data which will be fused with business (transactional) data to extract contextual information for various aspects of business, which will in turn be used to predict/estimate and recognise important events more precisely. The second thing that will play an important role is to find a perfect fit between technology and people, because in many industries human knowledge (or wisdom) is still hugely important and will have to be combined with the results we get from AI systems.
Subscribe to our Newsletter and get relevant updates …