Allied Irish Bank CIO Tim Hynes has seen many emerging technologies reach the hype cycle’s peak during his IT career, now well into its third decade. Their history tells him that artificial intelligence will only reach its potential if it’s promoted and applied to realistic scenarios.
“Artificial intelligence feels a little bit like a gold rush event,” said Hynes at the AI Congress London.
“That doesn’t mean there’s not value in it. What it does mean is that if you’re going to get real value, you have to be pragmatic.”
AI will attract greater support for its attainable and foreseeable benefits than for the transformative potential it will unleash in the future.
It’s difficult to attract interest and engagement from investors in long-term potential, as they struggle to see the immediate value.
“Human nature is that we tend to overestimate what we can achieve in one year, and underestimate what we can achieve in three,” Hynes said.
“What’s important for us as technologists is to think about that and be very clear and honest with the people who are going to be paying the money for it, because each time we get that right, we build confidence and we build credibility, which makes it easier for us to get the support going forward.”
The history of computers shows the challenges of predicting the future. In 1979, a 250MB hard drive weighed 250kg and cost tens of thousands of dollars. Ten years later, laptops, PDAs, cameras, Walkmans and walkie talkies were widely available.
By 1994, Bill Gates was extolling the virtues of the CD-Rom, but during the following decade the iPhone was heralding a new technological dawn
CIOs need to bear in mind that when they’re speaking to non-technologists their visionary predictions will sound preposterous at the time, even if later they prove to be accurate.
Today, neural interfaces and autonomous cars are making AI a reality. Hynes believes that if CIOs are to make it successful in their organisation, they need to balance pragmatism with vision.
“The trick here is to get predictions right, and one of the ways to do that is to be pragmatic,” he said.
AI use cases
The term AI is used today to describe a wide range of technologies.
Workflow automation is sometimes inaccurately called AI, but the misnomer doesn’t mean such data-driven systems lack value.
AIB has found a number of use cases for it, including a scanning system at its branches for customer documentation that digitises the information and runs it through the bank’s systems quickly and in line with regulatory requirements.
“It’s not AI,” Hynes said. “What it will do for AI though, is it’s part of the digitisation of the business, and the more we digitise the business, the more data we get, and the more data we get, the more we’ll be able to feed the AI. Because what AI will be able to do for us is take data, and if we apply context it turns it into information.”
Robotics is another form of automation often described as AI. AIB uses it to augment its human workforce, by training robots to compensate for unusual rises in demand that create a workload that is beyond the capacity of the team.
AI is the next step in the evolution of these technologies. While these forms of workflow automation and robotics rely on structured data, AI can make sense of both structured and unstructured data and develop context out of the information that can turn it into insights.
It’s already used by AIB, to mitigate errors in tax deductions on mortgages to which customers are entitled, but don’t always receive due to regular changes in the rules
“It’s usually something we figure out and fix, but it annoys customers and it annoys the regulator and it costs us effort and time,” said Hynes.
“We applied artificial intelligence looking backwards, and we discovered that if we had an AI watching what was going on and dealing with this for us, it would have identified over 90% of the errors that had slipped through the net.”
AIB is also working on using chatbots in their HR systems and call centres, to streamline both employee and customer communications.
Customers are becoming more comfortable with chatbots, particularly when it works in conjunction with a human.
Sentiment analysis can help the chatbot identify the words and phraseology that indicate when a customer is annoyed, and then engage a human being to resolve the situation. The AI can continue to watch the conversation that follows, and intervene when the human needs further information that they can’t instantly remember, such as the new interest rate on a particular loan.
Other banks have used AI to assess when customers were defaulting on loans. In the US, one augmented this assessment by giving the AI access to weather information, and learned that customers in areas affected storms were more likely to default on their loans, as they’re more worried about their house flooding or their roof blowing off. This didn’t happen, however to the most financially secure customers.
The AI now monitors both the customer’s financial stability and the weather in their area, and identifies when they’re in the path of a storm. It can then reach out to those customers that may struggle to pay the mortgage due to the damage and offers them a holiday of three months on their mortgage.
“By using AI what they’re able to do is get ahead of it,” said Hynes. “They don’t have to allocate capital, they don’t have to engage these expensive processes, they don’t annoy the customer, and frankly what they do is enhance the brand, because the customer says ‘hey, if I’m having a problem, this bank is on my side’.”
The ethics of decisions made by AI such as this still need further work and controls. This will range from a chatbot developing the intuition to identify that a customer fully understands what they’re buying, to an autonomous vehicle weighing up the risk of human lives before a crash.
The future of AI
Hynes believes that AI will have to harness future technologies to unleash its full potential, particularly the processing power of quantum computing and data transmission speeds of 5G.
They will mean that machines won’t have to carry all their compute with them. An autonomous vehicle, for example, wouldn’t need to have all of its computational power on board, but could instead communicate with a remote facility that could do the compute for it.
The combination of quantum computing and 5G will also support dramatically enhanced capabilities for robots.
“In the future, with quantum computing on the back-end and faster communication, the processing doesn’t actually have to be on the robot, so you start getting even more intelligent or clever activities and uses for the robots,” said Hynes.
“As you look at this stuff, think about where we’ve come from, understand the pace of technology, understand it’s going to keep happening and ground it in practical use with a view to the future.”