• h

    We produce insightful, relevant and custom fitted counsel that balances immediate success with sustained growth, on both a corporate and individual level

Top management must be data smart to get AI right

There is plenty of reason for top management to get excited about AI. The potential is enormous, progress seems spectacular, and quite frankly, artificial intelligence sounds much more exciting than good old analytics or statistics. A wide range of path-breaking applications from autonomous vehicles and chatbots to image recognition and diagnostic algorithms all fall under the broad umbrella of AI. Undoubtedly, AI will play a key role in driving innovation, transforming industries, reshaping the way we work, and putting data and analytics at the heart of several business enterprises. Simply calling something AI, however, does not turn garbage into gold.

Amid all the hype, the AI magic dust has become an easy sell and smart marketers are taking advantage. Venture capitalists, for example, are facing countless startups that are all about AI when raising money and talk the talk about cutting-edge machine learning when recruiting. When actually building their solutions, however, many of these startups still resort to old school statistics and traditional rule-based automation logics. Similarly, corporations must be alert when building or buying their new AI solutions. Most long-standing analytics challenges and pitfalls carry over to the brave new world of machine learning and AI. And with the focus increasingly on integrating AI into business processes and automating important decisions, the quality of the underlying analytics is becoming more important than ever.

One of the most common mistakes in applying AI to business problems is misinterpreting correlations as causal effects. While powerful prediction algorithms have numerous value-adding business applications ranging from content recommendation and promotion targeting to risk scoring and demand forecasting, many fundamental business problems center around causal questions. How does customer demand respond to price changes? How much additional revenue does an extra euro of marketing spend generate? Does early engagement with a digital service cause or simply predict customer loyalty later on? Questions like these cannot be answered by simply applying predictive AI to historical data. For example, higher housing prices might be a great predictor of larger transaction volumes because of their correlation, yet construction companies would clearly be foolish to let algorithms hike up their prices in the hopes of selling more apartments.

Other typical analytics pitfalls include misinterpreting patterns that have been imposed on the data as actual business insights and defining the analytics problems too narrowly to see the forest from the trees. The former pitfall often occurs when granular profitability data is used carelessly in AI applications. Steering sales and marketing decisions based on profitability measures that depend on arbitrary cost drivers and allocation rules is bound to lead you astray, no matter how sophisticated your optimization algorithms are. The latter pitfall, on the other hand, often manifests itself in simplistic optimization exercises. For example, optimizing price quotes for individual contracts without considering their ramifications on the capacity situation and competitor behavior in future negotiations can really hurt profitability.

Understanding what different data can and cannot tell us is a great asset for business executives as they look to harness the power of AI to solve business problems and create new growth opportunities. With cutting-edge analytics becoming a key source of competitive advantage and data assets so strategic that they motivate major transactions such as Microsoft’s $26 billion acquisition of LinkedIn, top executives must get smart about data. This does not mean that CEOs must suddenly become data scientists and learn to code in R or Python. What is needed instead is that they realize how data and analytics are transforming their industries, understand the value and blind spots of their data assets, and recognize business problems and opportunities that can be tackled with AI. Data literate top management is becoming a key advantage for companies as we move towards an increasingly data and AI-powered future.

Iiro Mäkinen

By: Iiro Mäkinen

Chief Data Scientist