What every leader should know about AI techniques

Successful business leaders are often heralded for their ability to see connections that others did not. They challenge and inspire their teams to think bigger, better, and bolder through simple questions such as: What if we … ? Could we … ? Have we thought of … ?

But seeing new business opportunities often requires understanding what the possibilities might be.

AI has disrupted many industries and is expected to continue to do so. A recent McKinsey study estimates that the most advanced AI techniques may create between $3.5 trillion and $5.8 trillion in new value annually in 19 industries — from agriculture and automotive to banking and basic materials to travel and telecommunications.

To anticipate what’s next and chart their course wisely, leaders need to be versed in the essentials of AI techniques.

What do we mean by the essentials? Here are three places to start:

Know what your algorithms can do for you

Most leaders have a keen understanding of the market forces affecting their industry. However, today innovation is often found at the intersection of AI and industry.

Many companies use machine learning, one type of AI, to increase revenue or reduce costs. To achieve their goals, data scientists often choose from three types of machine learning: supervised learning, unsupervised learning, and reinforcement learning.

Supervised learning uses training data and feedback from humans to predict specific outcomes, such as potential customer churn. Conversely, unsupervised learning explores data without being given an explicit outcome to consider, such as in the case of customer segmentation. Less discussed, but growing in interest, is reinforcement learning, which takes a rewards-based approach to optimize outcomes (for example, maximizing points for increasing returns of an investment portfolio).

Clearly, leaders don’t need to be able to advise their teams on which algorithms to use or when. But those leaders who understand the basic differences at a strategic level can ask deeper questions and more accurately anticipate opportunities on the other side of the horizon.

Be prepared for deep learning

Deep learning, another type of machine learning, offers businesses the ability to process a wider range of data resources — images, audio recordings, and time-series data. It often provides more accurate results in applications such as image classification, facial recognition, or voice recognition. Deep learning is at the heart of some of today’s most forward-thinking AI-enabled services — from helping manufacturers detect product defects on the assembly line to chatbots that can respond to highly nuanced customer inquiries.

Data scientists frequently use two types of models to power deep learning applications. The first, convolutional neural networks, can uncover unique characters in images and classify those images. The second, recurrent neural networks, can learn time-series data or sequences from data such as audio recordings and text and provide an output.

As with other types of machine learning, leaders don’t need to be steeped in the intricacies of these models. But with electronic-device users generating quintillions of bytes of data per day, a basic understanding of these models — what they do, how they work, and the kinds of business cases they can support — can position leaders to better navigate where, how, and when to apply AI to their businesses.

Know the limitations of AI — and how to overcome them

AI is defined as the ability of a machine to perform cognitive functions we associate with human minds. And as human learning grows, so does AI. Scientists continually push the boundaries of what is possible with new techniques. As a result, it’s critical for leaders to understand not only the vision but also the reality of AI.

This includes knowing the organizational and technical constraints — and the possible solutions. Many AI initiatives have been sidetracked due to myriad roadblocks, including cultural barriers in adopting AI, the scarcity of data science professionals trained to build AI applications, the lack of resources to label data and train systems, the unexpected influence of algorithmic biases, and much more.

Leaders who understand what AI does well, what it doesn’t do well, and what new developments are emerging can more successfully direct their investments over the next few years.

In-line images by Digital McKinsey

Header image by Denys Nevozhai on Unsplash