SMU Office of Research & Tech Transfer – “The future is already here – it’s just not very evenly distributed,” joked William Gibson, a prominent American science fiction writer who was the first person to use the term “cyberspace”. Gibson’s comment about the future from the 1990’s is applicable to the role of artificial intelligence (AI) in today’s workplace, as more businesses deploy AI systems to gain a competitive advantage. While there are a growing number of early adopters, many others remain hesitant about the need for such immense digital transformation.
Quoting Gibson, SMU Professor Emeritus Steven Miller explained this phenomenon, which he terms “the future of work now”, at the Working with AI-Enabled Smart Machines webinar on 25 August 2021, organised by the SMU Lee Kong Chian School of Business. Drawing from a set of 30 case studies of how employees and executives are now doing their everyday work with smart machines, Professor Miller’s talk provided insights on AI-enabled business transformations.
The case studies are an integral part of a book Professor Miller co-authored with business-IT author and thought leader Tom Davenport on the future of work with smart machines that will be published by MIT Press next year. Spanning a range of industries from banking to healthcare, e-commerce and manufacturing, the case studies include familiar names in Singapore, such as DBS Bank, Shopee and Jewel Changi Airport. All of the case studies highlight AI’s capacity to enhance productivity, and several of them also discuss how the enhanced capabilities provided by AI-support systems were able to increase inclusivity in the workplace.
Boosting productivity in big banks
More than just a lender or a safe space to deposit your funds, a bank acts as an overseer of money flows. At Singapore’s DBS Bank, the transaction surveillance team watches out for suspicious transactions that may indicate money laundering, fraud or other types of financial crime, Professor Miller shared.
To detect errant activities, banks across the world have long adopted rule-based systems. “They work well, even in this age of data-driven machine learning,” he acknowledged. But though these rule-based systems are useful, they are inefficient – as much as 98 percent of alerts may turn out to be false-positives due to the fact that more and more rules have been added over the years to deal with new situations. As a result, frontline employees were spending inordinate amounts of time sorting through false alerts.
Despite being somewhat laborious, rule-based systems remain in place as regulatory authorities in most countries still require them. To accelerate the surveillance process, DBS integrated a machine-learning system that draws upon internal and external sources of information to automatically evaluate each alert from the rule-based system. Today, this added machine-learning analysis layer enables the bank to evaluate and prioritise alerts by generating a probability score that indicates the level of suspicion.
“This machine-learning system transforms the work of the frontline worker doing transaction surveillance analysis,” Professor Miller said. “They no longer have to spend most of their time manually assembling data from various parts of the bank to check each transaction, as the process of obtaining all of the source data for analysis has been automated as part of deploying the machine-learning support system.”
Increasing inclusivity in the workplace
As productivity gains from AI implementation become apparent, demand for employees with relevant technology skills has also risen. These employment opportunities may be a boon for those who are digitally savvy, but it could also lead to firms providing fewer opportunities for entry-level workers who work on administrative tasks and simple transactions.
“Some of the firms we interviewed shared that the people they hire now need to have a lot of domain and industry experience, because they’ll be working with these AI support systems and they need to evaluate whether the output of the systems are correct,” revealed Professor Miller. Such hiring policies may work well temporarily, but could prove short-sighted in the medium to longer-term, he added. “If you keep doing that, how can you grow the pipeline of experienced people?”
Instead, Professor Miller suggests, companies should consider how technology can increase the range of suitable hires and help to make recruitment policies more inclusive. “Many of these tech support systems were designed to improve the productivity of the existing workforce. But it turns out that as a byproduct, they also provide performance support – support that can make it easier to bring in a new person and train them,” he said.
The future is here
While AI is already proving useful in almost every industry and application domain as illustrated by the successful use cases, Professor Miller noted that the technology is not evenly distributed. “The usage of these AI support systems and automation systems is highly concentrated in a very small proportion of firms, and I don’t just mean only the Amazon’s, Microsoft’s and Google’s; this also includes DBS and other [non-tech] firms at the leading edge,” he said.
“For the majority of all the other firms in each industry, they’re in a much earlier stage in terms of harnessing AI capabilities. But the future is already here, and the usage of AI systems will continue to propagate. The partnering of humans and machines will become everyday work for more and more firms, just like it already has for those firms in our case studies,” he said.
Contrary to widespread speculation and popular media comments, AI-enabled machines won’t take over the human workforce anytime soon, Professor Miller assured. “AI research and technology moves quickly, but deployment within an organisation is complex and slow. So, the threat isn’t about whether AI is going to take over our jobs, but whether or not we choose to get involved in learning how to work with AI to support our work.”