Sramana Mitra: Unintended bias has been very intentional in the past.
Mike Flannagan: That’s the thing. If you look at the way compensation is generally determined, it’s based on salary history. If you consider the fact that, historically, there was an intended bias, it’s based on a history. If we want to remove that inherent bias, it’s very difficult for people to do it because they’re biases they don’t even realize they have. That’s one of the things that something like machine learning gives us.
Sramana Mitra: I agree with that.
Mike Flannagan: In our machine learning portfolio, we have a resume matching application. We’re beginning to build workforce analytics that begins to help with the process of recruiting correctly. We’re moving down that path with our customers of helping them do a better job of selecting the right candidate for the right job while removing some of the inherent human bias that comes with candidate selection and interviewing.
Sramana Mitra: SAP has a long history in manufacturing and supply chain. Are there use cases that you could point to in those areas where interesting and fascinating things are happening?
Mike Flannagan: Clearly, predictive maintenance is the thing that’s closest to that area. It’s an area where SAP has a massive amount of domain experience. We’re working with customers using our machine learning and predictive analytics capabilities to get much smarter about the way they’re doing maintenance.
Those use cases, whether they’re predictive maintenance or retail scenarios, domain knowledge, i.e. having the data to train your models and having the process understanding about where to inject machine learning into a process is really one of the things that every company needs to consider.
Earlier, you asked me a question, “If I were an entrepreneur now, what would I think about?” From a technology perspective, machine learning, advanced analytics, and Internet of Things are very hot topics. All of those things are about data. IoT is about generating massive amounts of data. Machine learning is about being able to digest and understand from a massive amount of data. Artificial intelligence is being able to apply those learnings.
The question that I would be asking is, “What data do I have? What data does my unique history give me the ability to access a fairly unique volume of data and use that data to solve a meaningful business problem?”
Sramana Mitra: This is the answer that every single person who comes as an interviewee in this series gives. I truly agree with you. You must be tracking this trend as well. The engine providers are really trying to abstract out as much of the usability as possible. Microsoft, for example, is trying to pack in AI capabilities and models into Azure such that non-computer science people can just program those interfaces.
The battle then becomes the battle of domain knowledge and battle of algorithmizing domain knowledge such that machine learning algorithms can take that domain knowledge and build sophisticated AI underneath.
Mike Flannagan: This is something that our SAP predictive analytics product aims to deal with. How many people today actually use data-driven decision making in their core job functions? It’s also about how do you consumerize the ability to use advanced analytical techniques without requiring everyone to have a data science skillset.
I can’t have my HR manager trained in Python and R. I need them to be experts in HR. How can you make it easy? How can you integrate those advanced algorithms into the way that people work instead of making it some complex thing that they do outside of their core job. Netflix does a great job of now making you think of the fact that their recommendation engine is actually driven by some complicated algorithms in the background. It just gives you a recommendation and you either like the way the preview looks or you don’t.