Building better chatbots

by Angie Basiouny

AI-powered chatbots are being deployed by companies to handle everything from booking travel to helping shoppers choose the right outfit. What makes some more effective than others?

Chatbots are everywhere. A fast-growing communication channel for brands, AI-powered chatbots are being deployed by companies to handle everything from booking travel to refunding purchases to helping shoppers choose the right outfit.

When done right, they can drive sales and provide positive experiences that keep customers coming back. But poorly designed chatbots can frustrate customers and drive them away - sometimes for good.

“AI used to be a tool in the back office. The consumer wouldn’t know much of what was going on under the hood,” Wharton marketing professor Stefano Puntoni said. “The arrival of generative AI gave us this interactive capacity, and suddenly everybody has a chatbot. Not all delight.”

Puntoni, who is faculty co-director of Wharton Human-AI Research, has teamed up with Thomas McKinlay, founder of the Science Says newsletter, to create The Wharton Blueprint for Effective Chatbots. Based on the latest scientific research, the blueprint offers practical solutions for increasing chatbot usage, improving consumer trust, and deciding when and how to use chatbots that are more human-like or machine-like.

Puntoni said the goal is to help firms build a better chatbot by understanding the psychological factors at play when customers interact with “a nonhuman agent.” The blueprint draws on dozens of studies conducted by experts at Wharton and elsewhere; none of the research is more than a few years old.

“You couldn’t do this research 10 years ago because it wasn’t there,” Puntoni said. “We thought it would be valuable for the business community to have a guide that would take the latest research and package it in a way that would be easily digestible and useful.”

Some of the key points in the blueprint include:

Increase the use of chatbots by recalibrating them to recommend both popular and niche products. That way, suggestions feel personalised.
Improve consumer trust in chatbots by communicating that people were involved in the development, rather than presenting them as autonomous systems.
Use human-like chatbots to deliver favourable decisions, such as getting approved for a loan.
Use machine-like chatbots for situations where customers are likely to be angry (e.g., billing disputes or technical complaints).

Focus on the customer

Chatbots are lauded for their cost-cutting efficiency. They can process requests faster without the need for as many employees. But Puntoni said it’s important for firms to consider the other bottom line: their customers.

“It’s not just about saving money. I actually think of chatbots as an engine for customer experiences. Can we create new customer experiences, new lines of business, stronger customer relations, higher margins?” Puntoni said.

Puntoni studies how technology changes human behaviour and the sense of self. A study he published last year found that AI-powered companions can reduce feelings of loneliness for people. Another co-authored report revealed that weekly usage of gen AI in large organisations has almost doubled, from 37% in 2023 to 72% in 2024. He said companies need to understand the psychological components of AI if they want to get the most out of tools such as chatbots.

“The repercussions for society are huge. Imagine a world where almost at every turn you have access to an efficient chatbot. This is quite radical,” he said. “From the point of view of the firms hoping to benefit from this technology, the design matters. The main challenge for many organisations is not engineering, it’s customer insight.”

Useful resources:

Knowledge@Wharton is the online research and business analysis journal of the Wharton School of the University of Pennsylvania.

Visit our website.

Popular

Follow Us

Newsletter

Receive the free Ecommerce.co.za newsletter for the latest news and trends: