Welcome to Part 3 of our consideration of how to leverage artificial intelligence (AI) in customer-facing settings. We share examples and strategies for using AI as a coach or as a tool for coaches. As you would expect, coaching formats affect people differently based on their expertise. The graphed data come from Luo et al. (2020), Artificial intelligence coaches for sales agents: caveats and solutions, Journal of Marketing. Given both Utpal and I play tennis, we also offer examples from a more traditional coaching setting.

Thanks to Utpal Mangla, VP and Senior Partner, IBM, for joining me in this post. Under Utpal's leadership, his team recently achieved its mission of making "Watson AI Impact 1 Billion Consumers." Utpal and I are officers within the ISSIP (International Society for Service Innovation Professionals) organization.

Fintech Sales Agents

Luo and his colleagues examined data from two fintech organizations in Asia. In the first, sales agents are selling personal loans to current banking customers. In the second, sales agents call people who are delinquent on their peer-to-peer loans to collect overdue amounts. In both instances, and in all contact center work we are aware of, sales coaches listen to calls and give coaching. The focus here was whether and how AI could play a role.

In the first study, the comparison was AI versus human. Sales agents received an email each morning either from a human coach or from an AI system adopted by the company. The email included text excerpts from the prior day’s calls, called-out mistakes, and provided solutions to those mistakes.

Overload versus Aversion

As compared to human coaching, the researchers expected that AI coaching, as represented by this particular system, would affect sales agents differently based on the sales agent’s expertise (as measured by past sales performance ranking). They expected bottom-ranked sales agents (generally new-hires or interns) to be overwhelmed by the more extensive feedback provided by the AI. The researchers expected top-ranked experts to have aversion to AI coaching given the experts’ existing skills and expectations. Both overload and aversion were predicted to reduce the benefits of AI coaching. The three-bar chart maps the outcomes, supporting the researchers’ predictions about sales (their underlying predictions about learning over time were also supported).

In the loan collection context, Luo and colleagues dug deeper into how best to support agents. Bottom and top-ranked agents were coached in one of three conditions: face-to-face by a human coach, face-to-face by a human coach who received the AI’s feedback for the agent, or directly from the AI via email. (The host company declined to include the middle-ranked agents as they were satisfied with the results for middle-ranked agents in the prior study.)

Whatever it is, the way you tell your story online can make all the difference.

Race With the Machine

Using this AI system and these coaches, the face-to-face human-supported with AI coaching is the winner. Leveraging the benefits of AI (ability to listen to all calls, identify more mistakes, suggest more improvements) coupled with human coaches’ interpersonal abilities is a chance to do as Brynjolfsson and McAfee suggest and race with the machine, not against it (book, TED talk).