The Reality of Artificial Intelligence

Generated image of a brain as a hard drive

Open up any newspaper or social media feed these days and it’s easy to spot the tidal wave of hype that artificial intelligence (AI) has unleashed. Within this hype lies a lot of opportunity, but without the right knowledge and tools, it can be hard to identify the real breakthroughs. Spun by journalists driven by sensationalist agendas, AI news can make it easy to believe that super-intelligent robots are just around the corner. That reality is a long way off, but it doesn’t mean there isn’t a lot of value to be extracted from AI in its current iteration.

To date, real-world applications for AI have primarily emerged from today’s tech giants: Apple’s Siri, Google’s machine learning-powered search bar, and Netflix’s recommendation algorithms, to name a few examples. In today’s enterprises, the application of machine learning has been much more limited. It’s time to recognize that there are many more applications for AI in enterprises than just chatbots. With diverse applications already identified and many more to be discovered, today’s enterprise leaders have a rare opportunity to define what AI for industry will look like. But with all of the hype, it’s hard to know where to get started.

The key to successful enterprise engagement with AI lies in collaboration. Unlike a lot of enterprise software today, machine learning isn’t a plug-and-play technology. Instead, the technology must be customized for the complex networks of people, processes, and tools that define today’s enterprise environments. To ensure a comprehensive perspective on the enterprise’s distinct set of requirements, objectives, and aspirations, it’s critical to bring together expertise from both technical and business teams.

“With more data in the world than ever before, conventional ways of doing analytics have diminished in their value.”

Expertise is needed from both teams to exploit the way machine learning currently operates. For the technology to work effectively, teams that design machine learning must first clearly define both the problem or task the machine will solve and determine quantitative metrics that measure performance. With this in mind, it’s easy to see where the input of an enterprise’s business experts becomes critical. Being able to define existing problems, in addition to identifying new ones, that could be solved with machine learning is one of the key first steps to ensuring that the technology can create rapid and measurable value.

Adequately covering even the basics of machine learning demands at least an article in itself, but identifying the right use cases for machine learning within an enterprise is optimized when both technical and business teams can establish a shared language to discuss the technology and its capabilities. Education gets successful collaborations off the ground, making sure teams have a shared understanding and framework to discuss the technology and its applications.

With more data in the world than ever before, conventional ways of doing analytics have diminished in their value. Costly and taking a long time to complete from end to end, conventional analytics often make it difficult to provide project teams with timely and accurate insights. Machine learning presents a powerful alternative, powered by the new kinds of algorithms that make it possible. With exposure to huge amounts of data, machine learning algorithms can rapidly identify patterns and learn how to optimally perform specific tasks with a high level of accuracy. This marks a dramatic departure from conventional analytics, because project teams no longer have to define and execute all of the steps required for analytics from end to end. Time spent processing reams of data to draw insights can now be spent transforming those insights into action.

Right now, machine learning’s applications can be distilled into three primary categories: prediction, classification, and reinforcement. Prediction is the most straightforward, featuring algorithms that analyze historical data in order to predict the future. Prediction algorithms powered by machine learning can be used to augment insights surrounding a broad range of problems including security risks, customer churn, and climate patterns.

In business, classification algorithms are frequently used to gain a clearer perspective of customer behaviour. The algorithms are able to identify highly individualized patterns that elevate insights on customers’ motivations and satisfaction.

Reinforcement algorithms are where we’ve seen some of the most dramatic displays of machine learning’s capabilities. An example is DeepMind’s AlphaGo, which rapidly mastered the ancient game Go to beat Lee Seedol, the world’s (human) champion. In reinforcement learning, algorithms are set up to learn from a training environment in real-time, rapidly amassing insights used to optimize for a certain target.

For each of these categories, practical applications have only started to be realized. By learning to collaborate to translate business aspirations into structured tasks that machine learning can solve using available data assets, business teams can begin transforming the technology’s vast potential into action. Collaborating early across business teams on machine learning also helps ensure that the introduction of machine learning into the enterprise’s existing complexity will run smoothly.

As a deep technology, AI is significantly more complex to deploy than the majority of software as a service (SaaS)-distributed plug-and-play products dominating technology’s landscape in recent years. Starting early provides enterprise teams with a surefire way to test AI’s impact in low-risk areas of their business before moving on to larger-impact projects. These targeted rather than widespread applications also establish the groundwork for future AI projects, allowing teams to identify and smooth out bottlenecks early on. By making incremental rather than widespread changes to processes and infrastructure, enterprise teams can also optimize AI’s impact for long-lasting and sustainable changes to the business’s operations.

Clearing the path towards a pragmatic perspective of machine learning and its capabilities is the key first step in thinking about its current level of potential for your organization. While machine learning’s impact on business is still in its early days, enterprises that build their capabilities with a pragmatic understanding of both the technology and how it relates to their business will have a distinct competitive advantage.