Advice for CIOs on getting started with AI
- 07 May, 2018 15:53
Despite the huge amounts of hype surrounding artificial intelligence – AI – the technologies the term covers undoubtedly present a huge opportunity for enterprises across all industries.
But what is the best way to get started and find the value?
Speaking as part of an AI NSW Summit panel hosted by Melbourne technology services firm DiUS last week, three leading Australian technology chiefs shared their tips and advice on starting out with AI.
Here’s what they had to say.
NIB Health Funds CIO Brendan Mills
NIB Health Funds launched an intelligent chatbot called Nibby in December, to answer common queries and provide customers with guidance when self-serving online.
Key to the success of the project was engaging with employees early and getting the business on board, explained NIB chief information officer Brendan Mills.
“We run the business on a culture of experimentation, it’s how we operate. Then the experiments that work we amplify, the experiments that don’t work we shoot…We got the business on board. We engaged our contact centre early. We brought them close to the project; we had developers working directly with call centre staff for the chatbot work,” Mills said.
“There’s a real opportunity to take people on a journey. The call centre employees we worked with on the chatbot project were probably five or six of the most engaged people we’ve got in the business now. They were super energised by what we did with them and despite the fact that they could see that at some point this would mean there would probably be less strain on the contact centre, they were supercharged by what we were doing, getting close to the technology, and feeling like they were making a difference,” he added.
Nibby leverages Amazon Web Services and utilises a completely serverless architectural approach. Amazon Lex provides the core chatbot functionality which is supplemented by AWS Lambda for integration and fulfilment, AWS Cognito for authentication and AWS S3 for static hosting.
NIB worked with Melbourne based technology services firm DiUS to stand-up the chatbot within four weeks.
The work was made somewhat easier by the fact NIB has undergone a major cloud transformation over the last three years and has rationalised its vast amounts of data.
“A lot of the foundational layer we’ve invested in, on AWS and others, set us up well for ultimately spring-boarding into our first AI venture,” Mills said.
“It all comes back to data, so without the data you can’t really get the output you want anyway in terms of the cognitive power of the machine learning algorithms. So data and making sure your house is in order is really important. There is a danger of people running off down an AI path and trying to be AI first without understanding the foundations need to be in place first,” he added.
Such back-end work is essential before embarking on AI initiatives, Mills said.
“It’s not something you go and buy from Amazon and plug into the wall and it just works. It doesn’t just happen, it’s not magic,” he explained.
NIB became the first Australian health insurer to introduce AI technology to assist customers. The company is now exploring how AI can be used to improve operational efficiencies and augment existing teams for smarter decision making in areas such as customer retention, claims and fraud detection.
“Continued investment in this space for us is where we’ll see continuous wins,” Mills said.
Stockland CTIO Robyn Elliott
The company is beginning to explore how AI, particularly advanced analytics and machine learning can be used to improve the ‘liveability’ of its residential projects. In the future it is hoped AI will help with the predictive maintenance of the company’s assets. There is also a robotics pilot underway.
“There’s some really great indices were building about how you design for liveability and walkability. Things that add to the quality of life, and how design elements in a community or housing estate or shopping centre can contribute to that,” she said.
The CTIO said offerings from existing vendors were a good place to start with AI initiatives.
“One of the ways to get started, and Stockland has taken a quite pragmatic view with that so far, is to use embedded vendor offerings. How do you actually start in a lean way? If you’ve got a customer systems based around say Salesforce – well you can use tools in that. There’s simple tools in Facebook, same with AWS, there’s lots of tools there that you can use if you need to get started and prove the value,” she explained.
AI initiatives were like any other technology proposal, and had to be presented to boards in terms of business value, Elliott said.
“My experience of working with boards is you’ve always got to tie it to a business problem. If you go to a board and say ‘we need to build an AI capability’ – there’s only a couple of board members who probably would relate to that. But generally you’ve got to tie it to a specific problem, which is a business problem you’re trying to solve,” she said.
“If you use AI as a tool in that mix to solve that problem, that’s the best way to get started, then ramp it up.”
AMP CTO Chris Bell
Having set out on a cloud-first strategy a number of years ago, financial services firm AMP now has its sights set on increased automation and innovation.
Like any new technology, the opportunity of AI is framed in terms of its business benefits.
“We have lots of ways that we roll out innovation across the group. We don’t treat AI as any different from any other kind of opportunity in terms of the way we assess the opportunity. At the end of the day, you have to go back to the business benefits around doing it, the do-ability of the opportunity,” said AMP chief technology officer Chris Bell.
The company has rolled out a number of chatbots over the last 18 months including Rosie, an online chatbot which assist customers by answering questions and guiding them through AMP’s Flexible Super superannuation product. Rosie is a cognitive virtual assistant from fintech firm Flamingo.
Another bot, from Spanish AI firm Inbenta, is being used internally to assist AMP contact centre staff.
“Our proof of concepts we got up and running quickly. My observation now, because we have a few of these running now, is ok what’s next?” Bell, who joined AMP in July last year, said.
The success of early AI proof-of-concepts is now being used to raise awareness of AI’s potential at the board level, kick starting the formation of a cohesive strategy.
“AI was something that our business was keen for us as a technology organisation to get involved in and start exploring. So having a quick win up there like that has really enabled us to build a broad strategy around AI as well. What we’ve been doing over the last six months is trying to further educate our executive committee and board around what the opportunity for AI is and really bring a bit of structure to that,” Bell said.
Like Mills, Bell emphasised the importance of getting back-end data sorted before pushing forward on such projects.
“We’ve been focusing a lot around getting the foundational aspects around data sorted out, that really can’t be underestimated. That’s an absolute non-negotiable for any of the things we’re doing around AI and machine learning,” Bell added.