People make a lot of decisions in today’s organizations. Take pay, for example. Pay rates often reflect management discretion and intangible contributions valued by managers. And yet only 40% of employees believe their pay is fair. Would injecting AI in decision making about pay improve the outcome? More on that later.
Let’s consider first why it’s so hard to make good decisions today, and why AI could help. A recent Gartner survey found that 65% of decisions made are more complex — involving more stakeholders or choices — than they were two years ago. In short, decision making can’t keep up with the fast-changing context in which business decisions are being made today.
“With continually more dynamics and complexity in modern-day business — especially digital business — our capabilities must improve to make the best possible decision in the shortest possible time, in a scalable, risk-conscious, consistent, adaptive and personalized fashion,” says Pieter den Hamer, Sr. Director, Analyst, Gartner. “Moreover, the decisions that we make today can’t be based on yesterday’s situation awareness; they must reflect the here and now.”
Related webinar: Leverage AI to Boost Decision Intelligence for Better Business Outcomes
Different degrees of AI in decision making
Humans may not be totally reliable or consistent in decision making, but they still bring important competencies to the table. Similarly, AI in decision making has its place.
Decision automation, decision augmentation and decision support represent the degrees to which AI and analytics can be deployed to pursue faster, more consistent, more adaptable and higher-quality decisions at scale.
The differences lie in the analytics techniques used at various points in the decision process, and who (or what) ultimately makes the decision:
- Decision automation. The system makes the decision using prescriptive analytics or predictive analytics. Its benefits include speed, scalability and consistency of decision making.
- Decision augmentation. The system recommends a decision, or multiple decision alternatives, to human actors using prescriptive or predictive analytics. Its benefits lie in the synergy between human knowledge and the capability of AI to rapidly analyze high volumes of data and to deal with complexity.
- Decision support. Human employees make the decision, supported by descriptive, diagnostic or predictive analytics. Its main benefit lies in the combined application of data-driven insights and human knowledge, expertise and common sense, including “gut feel” and emotions.
Download eBook: The Future of Decisions
Know when to deploy AI in decision making
Whether a decision can be or should be AI automated, augmented or supported depends on two key variables: time, or how quickly the organization needs a decision; and complexity.
The dimension of time refers to the span between when the organization recognizes a threat or opportunity, and when it decides what to do about it and acts. The time span varies between microseconds, in the case of high-frequency stock trading; weeks, in the case of pay decisions; and months or even years, in the case of a strategic merger or acquisition.
Complexity likewise operates on a continuum — mapped by the so-called Cynefin framework, for example, as extending from simple to complicated, complex and chaotic:
- Simple situations are stable and predictable, and operate according to clear cause and effect. Examples include payroll processing or call center routing.
- Complicated situations require expertise or analysis to identify cause and effect, often using expertise with a known problem-solving practice. Examples include insurance fraud, asset management and marketing campaigning.
- Complex situations involve multiple relationships and interdependencies, and effective analysis requires a systemic or holistic approach, with simulations to see how decisions can affect far-flung elements. Supply chain disruptions are one example.
- Chaotic situations have unknown causes and effects, with unclear or dynamic interdependencies. Small changes may have seemingly disproportionate impacts. Decision making is very difficult and requires experimentation and learning by doing. Examples include stock market crashes, battlefields and natural disasters.
AI in decision making depends on time and complexity
Applying the dimensions of time and complexity together can enable leaders to assess individual decisions and determine the value and feasibility of automating, augmenting or supporting them.
Automation is an appealing option for simple decisions that need to be made within a few seconds up to 15 minutes. Decision augmentation is an option for complicated decisions, or those that need to be made within minutes or hours. For complex or even chaotic decisions, and those that aren’t urgent, leaders can explore decision support.
AI applies in all of these situations. Over time, as technology advances, leaders can expect the bounds of what can be feasibly automated to move further along the axis of complexity.