From systems of records to systems of anticipation: You could not attend SAPPHIRE NOW 2016 in Orlando without hearing one term again and again: machine learning. No surprise, actually – in times when 2.5 quintillion bytes of structured and unstructured data are created each day, big data is everywhere, and for companies, it is all about not only having data, but about gaining insights and triggering actions faster, smarter and better. They don’t only need systems that say what happened but an intelligent decision support system that answers questions like “what might happen”, “what if”, and “what should I do”.
Gartner’s latest Magic Quadrant for Advanced Analytics predicts that by 2018, more than half of all large organizations worldwide will use advanced analytics and respective algorithms to compete in their markets. Advanced, predictive analytics are about calculating trends and future possibilities, simulating options, predicting potential outcomes, and making recommendations. This goes beyond queries. It reports to more sophisticated methods based on statistics, like descriptive and predictive data mining, simulations and optimizations that find trends and patterns in data, and finally machine learning.
Machine learning plays a significant role in the area of Artificial Intelligence (AI). AI stands for the ability of machines to execute tasks and solve challenges in ways humans would. Some tasks that human beings consider simple — recognizing something on a picture, communicating with each other, deciding on actions to achieve goals — are incredibly complex for systems. Systems need to learn all that what human just can do instinctively.
How machine learnings works
Consequently, the basic idea behind machine learning is to teach a computer to spot patterns and make connections by showing it large volumes of data. While in traditional programming, an engineer writes explicit instructions that computers follow, with machine learning, they don’t code instructions, but train computers. Data is used to “train” computers, enabling them to make accurate predictions and recommendations. Just think of recommendation engines or face recognition capabilities.
While from the consumer world, these examples can also be applied to the business environment. To stay ahead of the curve, companies need to be able to make intelligent predictions. They need a system that recommends quickly options what to do, based on both past and current data. Ultimately, they need a system that is proactively informing them, and not only about just what might happen, but also offer options and what they actually can do about it.
Imagine a system that tells you: “Chances are that next week’s forecast will be missed. There are three promotions lined up, but only for the week after. Do you want to bring forward them?” So the system not only knows that you will probably miss your forecast, but also gives intelligent advice on how to solve this challenge. You answering “Yes” results in a fully automated set up of the business process and the workflow, and your promotion indeed takes place and might save you the forecast. Still, the human being is the one who has control and finally decides.
Getting back to the customer experience, how would you like receiving personalized product recommendations in real-time, based on your unique user behavior? I am not talking about the standard recommendation tools like in online commerce, I am talking about really unique recommendations, exactly tailored to you, based not only on your shopping behavior in one particular store, but based on the device you use for surfing the Internet, based on your likes and posts in social media.
Getting faster, smarter, and better
This means nothing less to become more efficient. It not only takes customer experience up to a whole new level. Machine learning brings companies closer to their customers than ever before. Imagine you have a continuously evolving, dynamic picture of every single one of your customers. Combine this with machine learning and predictive maintenance – and you will be able to anticipate what your customers need before they feel the need.
At SAP, we follow an apps-led strategy, meaning that we are dedicated to ultimately making all our enterprise applications intelligent. We focus on use cases that have a direct impact for our customers, for example increasing sales, increasing customer satisfaction, decreasing cost, or increasing employee satisfaction and engagement.
During my SAPPHIRE 2016 keynote, we used the example of a marketing manager for an automotive company. The knowledge about the customer across every channel combined with predictive analytics and machine learning results in a seamless, yet individualized customer experience – and represents an unmatched opportunity for companies to boost their sales and increase customer loyalty.
And there is another example that impresses me a lot: Research has shown that companies spend up to 60% of their time identifying and matching job applicants to open positions. Thanks to machine learning, recruiters can automatically match CVs to identify the best candidates for a job, or the ideal job for a promising applicant. The result? Recruiters can spend more time on actually interviewing promising candidates rather than sifting manually through CVs.
It’s the human that matters
I predict that machine learning algorithms will act as a major differentiator in business in the near future and can transform enterprise data into business value. One thing must be clear: Machine learning by no means is something were data goes in and insights come out. No doubt that machine learning already is incredibly powerful when it comes to solving business challenges. Plenty of machines can do amazing things, often better than humans.
But this is not because these machines and systems are able to deal more intelligently with the world, but because behind every machine and system, there is a human being, making it successful. In every step, it is humans who make the world friendly to the machines and systems – by creating the data to train them, by building algorithms and make them faster, more intelligent, and more accurate. As such, I believe that transferring machine learning algorithms and linking them to business challenges will be crucial going forward when we are talking about skills – because no machine or system can be successful on its own and needs a human behind. My personal guess is that this is not changing any time soon.