When it comes to implementing new technology often a good starting point is to look at whether it can be used to address a pain point in the business that hasn’t been solved before.
That’s the view of Vectore founder Shamima Sultana, an AI specialist with a deep practice knowledge built by working in some of Australia’s most successful financial services companies.
According to Sultana, “When you have a pain point that hasn’t been solved and you want to take a new approach that’s where probably AI becomes more useful.”
The Vectore chief says one area where she has seen a lot of promise is in the area of assisted technology. “This means augmenting a human experience with your AI and machine learning technologies. This becomes particularly relevant today because there’s a lot of push around automation, and demand for doing a lot of work with only a few resources.”
When considering whether a new technology like AI is appropriate companies should consider two points;
– Can they provide the customer outcome in a really exceptional way?
– Will they also actually generate value for your organisations as well?
“In both of those cases you do not want to remove the human element from it. It is so crucial in so many businesses and it will be for a very long time. So the best use case involves augmenting the human workforce’s experience by enabling them with a technology like AI and machine learning, and at the end, providing a value to your end customer.”
Sultana gives the example of an operational centre which she describes as the engine room of the bank and which she says often represents one of the larger workforces of groups within the organisation.
“They are processing the mortgage applications, the credit card, home loans, finances and such. Even though it gets initiated via digital channels, there is a huge amount of human effort that sits behind the process to make sure it’s fulfilled, it’s maintained, “ she said.
To add to the complexity, the work generated once an application for a product like a home loan doesn’t end once its approved, especially where there might be as much as a 30 year life cycle attached to it.
“So you need human resources and expertise, to sustain it for a very long period. But at the same time, it is very cumbersome, very process heavy as well.”
That led one bank with whom Sultana has worked to embark on a project to minimise the operational cost of the human workforce while, at the same time, speeding up the process of home loan processing.
“We wanted to bring it down from 30 days to three hours, which is quite aggressive,” she said.
The team started by breaking down the components of work to understand where in the process AI would be most applicable, as well as determining those areas where there was no real requirement for a technical improvement.
“It came down to having a right balance between AI machine learning capabilities that the bank could use plus the ones that augmented the human experience.”
Start Small
For Rebbecca Kerr, general manager, information systems at mining company Roy Hill, one of the lessons from her company’s early entry into machine learning was that it is possible to get meaningful and usable insights from a relatively small set of data points.
According to Kerr, “A core issue for us is understanding our ore body and understanding both the physical properties but also the chemical properties, to ensure that we’re selling our customers what they’ve asked for.
“We’re a relatively new company, and while we had some information about our ore body, one of the things we discovered was that there was a chemical property showing up in our shipments that we actually didn’t sample for when we did the initial drilling.”
Roy Hill used machine learning to analyse 58 samples of the data from a specific location. The data was run through a machine learning process with all the other ore body data the company possessed. “Out of that, we where able to determine where else this particular chemical property might exist in the ore body.”
The company tested the results of the machine learning and found that it was accurate.
“So from actually a relatively small sample set, we were able to use a model to help us predict where we might have some issues going forward which is hugely valuable.”
There is an important lesson from both these examples.
According to Nigel Watson, JAPAC Head of Technology Partners at Google, “One of the reasons machine learning adoption has been delayed in the past is the enormous amount of data and processing power required to train ML models.”
However he said, “Cloud offerings like Google Cloud Platform, combined with democratised access to AI and ML technology and APIs have made it easy for any organisation to build and leverage AI for everyday use. Organisations can now cost-effectively amass and process large amounts of data with machine learning technologies to generate insights for their business.”
