‘What’s next in AI, Red?’

“What’s the best season for growing sunflowers?”

It’s not a question typically asked of a banker.

But it’s come up a few times during customer chats with Westpac’s virtual assistant, Red, which last week clocked up its one millionth conversation since being switched on a year ago.

Powered by IBM’s Watson artificial intelligence conversation engine, Red is designed to help with basic customer enquiries, its bread and butter being things that would once have needed a call, web search, email or branch visit, such as, “how do I activate my card?”, says Westpac’s Dan Sweeney.

“But Red does still get some quite amusing questions as well,” he says of the online assistant.

Aside from who it barracks for in the State of Origin or telling jokes, Red has been taught by Sweeney’s team of conversation analysts to recognise more than 5000 different ways customers ask about almost 100 banking tasks, a number that grows daily. Indicating the quality of Red’s chat, Sweeney says on average the virtual assistant resolves three in four queries without the need to escalate to a person – a better than expected result given Red’s still in its infancy – and repeat visits are steadily growing.

Westpac’s head of automation Nick Munro adds the questions that do require escalation are usually related to more complex or sensitive topics, such as concerns about potential fraud. 

“Red is continually learning and improving,” he says.

“And the bank has its eye on what else it can do with augmented intelligence to stay future-fit.”

Westpac is by no means alone as organisations increasingly realise the strategic benefits of robotics, along with other forms of AI-powered machine learning, natural language processing and automation.

Virtual assistants, or “chatbots”, have become common at most financial services organisations in the last few years while Jetstar has launched “Jess”, NIB Health Fund unveiled “Nibby” and Domino’s is offering “DRU Assist”, to name a few. Indeed, by next year at least 25 per cent of customer service operations globally will use virtual assistant technology, according to industry analyst Gartner.

It’s not surprising considering online “chat” is becoming preferable to picking up the phone for many consumers with 24/7 availability and eliminated phone waiting times the top benefits, according to a study last year by software company Mulesoft. Although, the study also found consumers’ top frustration was having to resort to other channels when virtual assistants couldn’t resolve their query – a problem that will, however, lessen as the technology becomes more “intelligent” over time.

While chatbots are often an organisation’s most visible “augmented intelligence” application, the use of robotics and machine learning is growing exponentially behind the scenes across many industries globally.

Earlier this month, the UK’s central bank said two thirds of UK financial firms use machine learning applications and its use is expected to double in the next three years. Its most common application was in the detection of anti-money laundering and fraud, making use of machines’ ability to mine huge volumes of data to flag anomalies much faster than humans. Customer service applications were also on the rise, according to the Bank of England report, with key benefits cited being increased operational efficiency, better customer personalisation, improved compliance and access to new analytical insights.

“The promise of machine learning is to make financial services and markets more efficient, accessible and tailored to consumer needs. At the same time, existing risks may be amplified if governance and controls do not keep pace with technological developments,” the BoE says.

While the consumer benefits of new technologies is driving rapid adoption, a quick web search shows trepidation still abounds across the population around AI ethics and its impact on the future of work.

It’s hardly surprising as age-old cinematic portrayals of “rogue AI” have been supplanted by a growing number of real-life incidents, like last month’s attempt by thieves in the UK using voice-mimicking AI software to imitate a company executive to commit fraud. Well-known technologists – most notably Tesla’s CEO Elon Musk – also regularly broadcast their worries about AI’s implications for humanity; and ambiguity often surrounds legal liability for AI-driven decisions.

But as more organisations globally focus on retraining workforces in the face of technological disruption, many commentators agree a range of previously non-existent “new collar” jobs were opening up, ranging from “conversation analysts” who review and improve the quality of virtual assistants’ responses, through to “robotics governance managers” and “chief ethics officers” who provide AI oversight and address potential biases of algorithms.

Munro – who confirms Red uses “supervised learning” that only allows it to say something a person has approved – says while demand for AI applications is increasing exponentially across Westpac as understanding grows of what’s possible, the bank takes a conservative approach – not by fear but necessity.

“We're dealing with the financial future of our customers. We need to carefully assess every single use case to ensure it’s the right thing to do rather than just because we can,” he says.

“What we're trying to do is create rich experiences sensibly, safely, repeatably, by using humans in that chain. At this point in time we won’t let the technology run free, and I don't see that changing in the immediate future.”