Organisations need to trust machines to do a job, sometimes better than humans can, and prepare for a future where bots will provide better customer engagement.

Specifically, businesses and governments would have to decide what defined people as human beings and establish a social contract outlining what was acceptable, and unacceptable, in a society where artificial intelligence (AI) played an integral role.

This would be essential as humans increasingly trusted machines to perform tasks better than people, said DBS Group CEO Piyush Gupta, who was speaking at Bloomberg LIVE's Sooner Than You Think forum in Singapore.

He noted that 50% to 60% of customers today processed banking transactions on their own, via their mobile, bypassing any interaction with another human or a banker. Another 75% traded online after the Singapore bank moved equity trading online and more did so after it removed remisiers from the equation.

Foreign exchange trading also climbed 95% after DBS moved the service online.

Recalling a conversation with Ping An Group, Singh said the Chinese financial services provider--which had 80,000 agents manning its call centres--was able to direct 70% of its calls to computers after introducing chatbots. Moreover, customers would not realise they were being serviced by a machine 20 minutes into a call.

More interestingly, he noted, during a trial with outbound calls to generate potential leads, Ping An saw a 1.5 conversion rate when human agents handled these sales calls. In comparison, the conversion rate climbed to 6% when machines did the calling.

Why were machines able to achieve a better sales conversion rate? He quipped that Ping An theorised this might be due to the computers' ability to stick to a given script and execute the right prompt or call-to-action as stipulated. Humans, on the other hand, were likely to stray from a given script and, as a result, would fail to convert a potential sale.

Nielsen's CEO and chief diversity officer David Kenny also revealed similar findings when the research company used AI systems to predict the weather. Over time, the machines were able to learn and reduce their error rate from 12% to 4% and, more importantly, could improve their accuracy because humans were advised against meddling with the predictions, said Kenny, who also spoke at the forum.

He noted that 75% of the time when meteorologists intervened and altered the AI projections, they introduced errors and pulled down the accuracy rate.

Piyush Singh, CEO, DBS Group

"Machines are actually better at predicting what people are going to watch," he said. "What we have to train humans on is to trust the machines, don't override them even if you don't like the answer."

He encouraged humans, instead, to let machines carry out the grunt work so they could focus their time on innovating and being creative.

The ability to generate more accurate predictions also was especially important in identifying potential audience, so marketers would be better able to target the right customer segments in their advertising campaigns, he said, noting that Nielsen used AI to power its audience measurement systems.

Better customer engagement with machines

In addition, AI increasingly is tapped to improve service delivery to and engagement with customers.

Asked if customers should be notified when they were dealing with a machine, Singh described how DBS itself had begun using chatbots in its AI-driven operations in India three years ago.

As an experiment, the bank played down the fact that it was engaging customers with a chatbot three months after it had done otherwise following the launch. He said call service usage increased by 15% during this time when customers were left to assume they were speaking with a human agent.

The test suggested that the general perception consumers were more likely to engage with a brand if they were speaking to humans might not be entirely accurate.

Just as a corporation or brand was merely a shell until its employees give it life, the same would apply for machines and robots, he said.

He added that machines increasingly were being trusted to perform tasks better than humans could and expressed his belief the former--as machines' computing and learning capabilities continued to increase--would be widely tapped to assist people in making decisions in the next five to eight years.

This then underscored the need for societies and industries to establish the role of humans and address the issue of ethics in relation to AI.

Singh said: "The journey then is to think about what defines us as human beings and how we want to construct a social contract on what [AI] is acceptable and what is not. The challenge isn't how we can code better, but how we can answer questions about what we are, what makes us exist in society, and what's right and wrong."

He urged the need to look at the "soft" issues and this, he added, required the participation of professionals from outside the technology realm, such as philosophers, psychologists, and behavioural scientists.

Kenny also stressed the need to ensure all pockets of the world's population were properly represented, especially when AI models were trained on data made available to these systems. This would further enable advertising dollars to be developed for all audience groups including the LGBTQ community, he said.

"You can't have a true dataset unless you've counted everyone," he added. This was particularly critical for an organisation such as Nielsen, which data was used to guide advertising and marketing decisions.

"If your data is being used to determine what shows are getting produced, where are ad dollars are going, what products are being developed, and where stores are being built, you need to be sure that you're counting everyone," he said.

"And in this part of the world, it's critical we're counting genders and not having any racial bias and that we're counting rural areas as well as urban areas. The risk is you don't want AI to be an elite platform. You want it to be a human platform. And we have a chance of bringing all humans into all the decisions using this technology."

Organisations also should look at how they could apply AI beyond personalised user experience and the point of purchase, said Ankiti Bose, CEO of Singapore-headquartered fashion e-commerce platform, Zilingo.

With consumption habits significantly different from what they were five years ago, where consumers now might consider a purchase after seeing a post on Instagram, it was essential brands not look only at the last step in the buying journey, Bose said.

"[Think about] the entire supply chain and how AI can be used to make it more efficient, more predictive, and more effective in selling someone what they want to buy," she said.