As today’s businesses work to adapt to the new era of AI, one of the big questions is whether to trust this technology to accurately predict the future. People have spent years working to understand business and economic cycles, laws of supply and demand, and other forces that shape opportunities and perils in order to come up with their best forecasts. Can algorithms come along and do a better job?
A slew of studies are exploring this. They often focus on predictive analytics, which Investopedia defines as “the use of statistics and modeling techniques to forecast future outcomes.” Comparing forecasts made by AI tools and humans, researchers have found striking results.
In one case, a team from the London School of Economics, MIT and the University of Pennsylvania pitted 12 large language models (LLMs) against “a crowd of 925 human forecasters from a three-month forecasting tournament.” The team asked for predictions of real-world events, including significant geopolitical ones. “Our results suggest that LLMs can achieve forecasting accuracy rivaling that of human crowd forecasting tournaments,” Philipp Schoenegger, Indre Tuminauskaite, Peter S. Park and Philip E. Tetlock wrote in their study.
Meanwhile, three researchers from the University of Chicago Booth School of Business looked into AI’s ability to predict earnings changes. Again, technology delivered. “Even without any narrative or industry-specific information, the LLM outperforms financial analysts in its ability to predict earnings changes,” Alex Kim, Maximilian Muhn and Valeri V. Nikolaev wrote. “The LLM exhibits a relative advantage over human analysts in situations when the analysts tend to struggle.”
But there is still a crucial role for human predictions in business. A study commissioned by the Association of International Certified Professional Accountants and the Chartered Institute of Management Accountants tracked one business that tried using both AI and people to predict its inventory needs. It found that in certain cases, people overriding the algorithm’s recommendations helped the business make better decisions.
“For both company-owned and independent-owned stores, analyst discretion improves retail assortment planning decisions when inputs to the model are scarce or noisy,” researchers Jen Choi, Ewelina Forker, Isabella Grabner and Karen Sedatole wrote. “Moreover, analysts are more likely to intervene by overriding the model recommendation in these circumstances, suggesting they recognize the value of their judgment when the model is likely to be weaker.”
Predictive analytics in sales
My work focuses on helping organizations transform customer experiences (CX), the most important element of the modern sales process. More than half of customers switch to a competitor after a single bad experience with a company, and about three-quarters do so after multiple bad experiences, according to Zendesk.
My team and I have explored how predictive analytics can drive better CX. Take call centers for example. In those environments, predictive analytics involve multiple components. As my colleague Joe Manna explained in a blog post, these include data mining, extracting information from large sets of data; statistical analysis, understanding and interpreting data as a basis for predictions; and machine learning algorithms, which enable “predictive models to learn from historical data, improving their accuracy over time.”
I’ve found that well designed AI-powered platforms can often do a better job predicting outcomes and customer behaviors. That’s why, in every interaction, it’s so helpful to have these tools working for the organization.
By following the language a customer is using, the questions they’re asking, the sentiments they’re indicating, and more, these tools powered with natural language processing (NLP) capabilities can search through records to determine what the best course of action is. They know with high statistical probability what the chatbot or human representative should say or do next in order to make the customer happier and achieve a sale.
Still, people are needed in this process as well. Even when chatbots are handling customer interactions, humans should be ready to jump in and improve things. There are times when a customer has a complicated challenge or a specific need that the AI doesn’t have enough information about. This is why I recommend a traffic light type of system, in which sales representatives get alerted when it’s time for them to take action.
As with so many other elements of delivering great customer experiences, organizations should see the predictive abilities of AI as a boon, not an erasure of the need for people. And employees in these areas should not fear that they’ll become irrelevant in the future. As AI tools develop new skills, people should think of these developments as opportunities. As these tools take on more tasks, people can spend more of their own time on other ways to improve CX and build the business. Ultimately, it’s not a competition between people and AI. The two can work in harmony.