It was only four years ago that AlphaGo began beating the world’s top professional Go players and Google released TensorFlow, an open-source software library for machine learning applications such as neural networks.
AIs are becoming capable of making corporate strategy decisions.© 2018 BLOOMBERG FINANCE LP
Since then, along with the expansion of IBM’s Watson, Amazon’s Alexa and Microsoft’s Cortana, the buzz associated with artificial intelligence (AI) has grown steadily as leaders around the world begin to examination how automation can help their organizations and customers. However, recent research from PWC shows that despite the fact that leaders perceive that AI is important—more important than the advent of the Internet—very few are actively investing in this new technology.
This is a problem. AI and automation are simply the latest in a long line of technological innovation—and will soon become essential in every part of the organization. Automation, for example, will not be contained to manufacturing plants and their shop floors. Automation is already affecting marketing, operations, human resources, and it will soon assist the strategy decisions that happen at the executive team and board level. After all, why should anyone believe that the same principles that allow robots to enable marketers cannot simultaneously enable business executives.
The only thing slowing this front office transformation is a lack of investment dollars by the C-Suite. What we have heard time and time again is that leaders don’t understand how strategy can be automated. Part of this hesitancy is a sense of hubris that machine intelligence will never trump our own human intelligence. We agree that human intelligence will always be a part of strategy creation, but with the enormous amount of relevant data available today – from financial to customer to employee to competitor to market and beyond – human beings are simply unable to keep up with the data crunching. Machines are becoming essential for their ability to gather and process large amounts of data.
The core advantage humans have over AI is in “task flexibility”. The same machine (the human brain) can be used for many different things. It has the ability to synthesize information and draw conclusions from varied past events to solve new challenges. But, in terms of raw processing power, the human brain is fundamentally outclassed by computers. The human brain processes tasks at 60 bits per second. The world’s fastest supercomputer runs at 200 petaflops. Roughly equivalent to the work of 6.3 billion humans in an entire year… in one second.
Here’s how it works for strategy decision making.
AI can be applied to all steps of strategy decisions, AIMATTERS
“Analyze” is the ability of an intelligent system to gather data and generate insights without human intervention. Often, this requires “training” the machines on what good looks like, and then setting them off to do the heavy lifting. Across the organization, this could look like fraud detection, lead qualification, or candidate ranking. Each application requires a unique data set relevant to the problem at hand.
When it comes to strategy problems, the data can be varied—often including metrics from traditional accounting such as return on assets, profit, and growth, but also alternative data such as customer satisfaction or social media sentiment. An AI application will work to connect the input to data to the desired output data – which could be stock price, market share, or revenue growth. Determining the relationship between what the company does, such as capital allocation or social media activity, and the desired performance, outperforming competitors, is what AI for strategy is all about. Getting the data and relationships right is the first piece.
The second piece is education—which means convincing the human user of the AI system why the insights are correct and valuable. That is, once a machine knows something about you or your business, it provides real time, and contextual insights that will help you understand, for example, how you stack up to your closest competitors and how strongly the data that the AI uses predicts performance.
The best systems may even tie in relevant content from other sources. Educational components for the C-suite today can be either automated or led by experts, but this piece is almost always necessary given that most leaders are still in the early conceptualization phase of how to use AI in their organization.
With the fundamental analysis in place, and the leadership team oriented about its significance, the system should begin to make recommendations. Please note that this more difficult than simply analyzing the data. In order to make good recommendations, the system needs to determine that the information is actually predictive and actionable. That is how strong are the relationships between our data and outcomes, and how certain are we that if a company changes its actions, the desired result will follow.
The Recommend process often begins with classification, regression and ranking in order to determine how to change the input data (by taking corporate actions such as capital reallocation) in order to create better performance.
When your artificial intelligence has come far enough to begin taking action on its own, you will know that your organization has truly arrived in the AI era. Execution is a very advanced step, and requires that a great deal of human intelligence is applied to checking the recommendations made by the system, and even piloting their implementation.
To avoid any colossal mishaps, it is certainly recommended that organizations begin their AI journey by doing the execution piece manually. That is, looking at the recommendations made by the system, evaluating them, and determining if the course of action is the right one. However, as confidence grows, it may be possible to allow the AI autonomy to make some course-corrective moves on its own. In the strategy space, this might look life shifting budget allocation, adjusting staffing levels, or halting a marketing campaign.
One of the advantages of artificial intelligence is that it is data-based and digital, which lends itself well to ongoing monitoring. To gauge the success of AI efforts it is, of course, essential to track the analysis, the recommendations, and the execution—hopefully creating a virtuous circle where successes and failures add more information back into the process and create a smarter AI.
Often times this can be done automatically. For example, a sales organization needs to hit key lead generation metrics. This is automatically tracked via the CRM and monitored with an AI that feeds a business intelligence dashboard.
A good monitoring system also provides a visual dashboard that can keep the leadership team up-to-date, and increase their confidence that the system is working well. It can even act as an auditing system and send up alerts when important metrics get out of line.
These five pieces together define a successful strategy AI system from start to finish—and back to start again. No leader should deploy intelligent and autonomous applications without a thorough understanding of how the system works, from the variables used for analysis to the key outcomes tracked for success. Without the right data and the proof needed to recommend and execute actions, the system is actionable. Leaders cannot, and will not, move to a model of intelligent applications without the trust and transparency provided by the education and monitoring piece. But when the five components of AI are integrated and made easily available to leaders, AI can begin to guide strategy in all types of organizations.