AI and machine learning are touted as a way to generate value from the troves of data organisations have stockpiled. Proponents of the emerging technology argue the benefits can flow almost immediately, especially in industries like financial services.
However, there is a problem and it is particularly prevalent in Australia. A dearth of data expertise in the local market and the resources required to develop AI and machine learning models has meant that while the data exists, organisations struggle to leverage value from it.
- Leadership Webinar: Which-50’s 2019 Outlook and Business Transformation Drivers webinar is set for November 27. Register today!
DataRobot, an automated machine learning company believes they can help close the data scientist gap with software and, in the process, weed out some of the “fake data scientists”.
Armed with $100 million in series D funding DataRobot has turned its attention to the Australian market where they will help existing local customers like Virgin Velocity and one of the major banks and new clients to automate their machine learning models.
According to Greg Michaelson, DataRobot’s general manager of banking, organisations have done a good job at collecting data but a relatively poor job at extracting value from it.
Michaelson told Which-50 that in most large organisations around 80 per cent of data is not utilised. With no shortage of AI and machine learning use cases, that is essentially money being left on the table, he said.
“Most organisations have only scratched the surface in terms of the potential utilisation of their data,” he said.
“They need to find a way despite all the challenges of not having the right people and the data being a mess and there being no easy solution to that problem. They need to find an organisational willingness to put their nickel down and say ‘we’re doing this. This is important.’”
But often that conviction is lacking, Michaelson says, as the area of data science remains relatively new and the “buzz” of things like self-driving cars and computer vision is obscuring the value of “less sexy” AI applications.
Michaelson, who specialises in banking, suggests use cases like fraud modelling, credit risk modelling, employee attrition and customer churn can be improved almost immediately with AI and machine learning models. As the models improve over time the benefits do as well with potential savings in the hundreds of millions, according to Michaelson.
For those that have begun exploring AI and machine learning the process has been slow.
Most organisations attempt to code their machine learning and AI solutions by hand, according to Michaelson, a resource intensive process that relies on data expertise.
But Gartner expects the demand for AI talent will likely outstrip supply. Even now, securing top talent is costly – the top data analysts in Australia command salaries near $300,000 per year.
“It turns out there are dozens, maybe hundreds, of use cases an organisation could build. Each of which has huge potential,” Michaelson said.
“The hand coding thing is slow. Finding people who can do it is an even bigger problem.”
According to Michaelson, the resurgence in AI is so new there is a lag between industry’s demand and the tertiary system producing the talent.
“It’s everywhere. It’s developing so fast so finding someone who graduated from school with a degree in anything, they’re probably not data scientists. That’s how new it is. None of the professors really know about this stuff except in isolated pockets.”
Another problem has arisen, as candidates clamour to get a share of the high salaries – “fake data scientists”.
“There’s so many online programs you can take. You get [people] signing up for a Coursera course and then put data scientist on their LinkedIn. All of a sudden they’re in the job market.
“Part of [the challenge] is finding people that are actually what they claim to be and the other part is there just aren’t that many of them.”
“Some estimates are that there are a million data scientists in the world. I think it’s much closer to 100,000.”
It is a problem DataRobot is grappling with first hand as it seek its own digital talent for their expansion into Australia.
“Finding people is a perennial problem,” Michaelson said noting the firm’s expansion will focus on customers in banking, transport, resources, retail and telcos – the “usual suspects” for AI and machine learning.
DataRobot provides customers with an automated machine learning platform, allowing organisations to automate the technical parts of the model building process.
The solution still requires humans with knowledge of the data and domain expertise but saves them from the advanced mathematics and coding typically required in machine learning. Essentially the platform allows business analysts to perform some of the work the more in demand data scientist do.
“DataRobot’s value-prop is that we take care of the math and the coding. So that a domain expert can do all the [other] work.”
For organisations requiring more specialised models the DataRobot solution also serves as a productivity tool for traditional data scientists.