CONTRIBUTOR
General Manager and Editorial Director,
Techstrong Group

The trouble with data scientists is that far too many organizations expect them to somehow intuitively apply artificial intelligence (AI) to advance business goals even though most of them have limited exposure to the industry segment they now find themselves employed. Very few data scientists before that are hired by an organization have any kind of vertical industry expertise.

The gap between data science teams and the rest of the business as highlighted by Alejandro Muller, founder of Savvy AI, a digital lending platform that makes use of AI to reduce loan processing costs, during a virtual AI at Scale Summit hosted by the AI Infrastructure Alliance is wider than many business executives fully realize.

The core issue is that data science teams often lack context when it comes to creating AI model. Many businesses hire data scientists without providing them with a clear charter, noted Muller. As a result, data scientists will spend a lot of time creating an AI model that identifies, for example, what combination of offers are the most profitable.

The issue is business executives assumed the data science team would in addition also determine when the optimal time to make that offer is. No one in the organization, however, actually specified that goal, so the data science team focused on surfacing insights that most of the people working in the business already know.

Most digital business transformation leaders recognize the potential of AI but the cultural divide between the data science teams that create AI models and the rest of the organization remains wide. Data scientists generally don’t have degree in business administration. They also have their own unique lexicon for describing processes that few business executives comprehend. Business executives tend to assume that data science teams are going to somehow intuitively grasp all the nuances that enable businesses to maximize profitability.

Not surprisingly, there’s a lot of frustration. The hiring of a data science team is an expensive position. The pressure to show a return on that type of investment is always going to be high. The trouble is that as AI initiatives start to go sideways many executives are simply not sure how to recalibrate these initiatives. It’s simpler to call it another failed experimental initiative gone awry. Those that persevere will, of course, reap the rewards. The issue is that in the absence of any clear understanding of how to move forward there are not many business executives willing to stake their careers on data science teams that admittedly are often far removed from the goals and metrics that shareholders are using to determine how successful the business may be.

Data scientists are, of course, not obtuse. They just need someone to spend some time explaining what metrics really matter most to the business. Thus far, unfortunately, that’s involved a lot of trial and error. In an ideal world, business executives would be spending a lot more quality time with data science teams describing the challenge an hand. Every business has key performance indicators (KPIs) the organization tracks to ensure goals are met. It shouldn’t require an AI model, however, to mathematically deduce that the probability any data science team will, of their own accord, identify which KPIs matter most is nearly zero. It’s, as always, up to business leaders to bridge the AI gap because there’s no scientific method that will automagically do it for them.