I’ve had good and bad experiences with chatbots.

When they are used to triage my customer service needs before connecting me with an appropriate agent, I cannot speak highly enough of their efficiency.

But when they fail to understand the nuance of my query and leave me reaching for the telephone (metaphorically speaking, anyway) the experience is frustrating at best.

“I can’t stress enough the importance of using chatbots as an agent companion, not as a substitute [for human assistance],” Adam Lindsey, senior director, global service operations at eCommerce giant Groupon, told DTX Manchester recently.

“[It should be] part of the ecosystem – not a replacement for it.”

With over 10 million customer and merchant contacts a year, and 18m active customers in 13 countries, Groupon is a hugely successful global eCommerce marketplace connecting subscribers with local merchants by offering activities, travel, goods and services.

Established in the UK in 2010, the business has spent over a decade experimenting with a variety of chatbot applications to understand properly what success looks like.

“One of the key learnings has been to embrace chatbots as agent companions rather than substitutes and be mindful of their purpose – where do they help and where do they hinder,” continues Lindsey.

“What we realised very early on was that automating infinite shades of grey results in endless uncontrollable outcomes. Machines do, in fact, lie – and automation only really works if it is at least equal to human support.

He adds: “In the end, we simplified the ecosystem – fewer tools, fewer policies, fewer variables – and encouraged the customer service delivery teams to develop strategy alongside product and engineering colleagues to arrive at better outcomes. 

“And we all agreed to boil the ocean only one cup at a time – in other words, get one thing right first before tackling the next!”

Lindsey says chatbot success eventually came when it worked out how to accurately capture customer intent and resolution and provide effective automated responses. 

“Savings made through automating repetitive tasks are used to reinvest in our wider toolset and most importantly, our people. Sentiment is far harder for AI to understand and reciprocate, by upskilling our teams and building in recovery processes ensures we retain the emotional, human response to queries,” he explains.

The experience with chatbots has provided an important blueprint for how it can utilise the next generation of AI – generative platforms based on large language models such as ChatGPT.

How AI is exchanging delays for speed in mortgage industry

With all that experience to draw on, we are now exploring how Groupon could use ChatGPT to improve customer services,” says Lindsey.

“Whilst automated customer support remains on the distant horizon, we could integrate ChatGPT into our website or mobile app to allow customers to get quick answers to their questions, like how to redeem a voucher or how to change their order, without having to wait for a person to respond – essentially a more intelligent chatbot. 

“ChatGPT could also be trained to offer personalised recommendations to customers based on their purchasing history or browsing behaviour. For example, if a customer frequently buys restaurant vouchers, ChatGPT could suggest new restaurants in their area that they might be interested in trying.

“Reducing customer wait time on queries is proven to lead to higher customer satisfaction and loyalty so ChatGPT deployed on quick, automated support when helping customers is an obvious win.”

These days 20% of queries to Groupon are fully handled by chatbots, with a roadmap to automate over 50% of volumes by 2024. 

“I see a similar journey for ChatGPT to improve the customer experience, increase satisfaction and reduce costs associated with customer support and make an intelligent collaborative, important contribution to business growth in the next 10 years,” adds Lindsey.

Understanding generative AI crucial to addressing risks