Health insurer nib has broken down the biggest roadblocks and wins in its journey to migrate 95 percent of its workloads to AWS, which started in 2015.
The insurer’s expansion into providing NDIS, travel insurance, telehealth services and B2B data services left it with disparate infrastructure, CIO Brendan Mills said at a press briefing announcing the closure of its last data centre; it had seven such facilities at one stage.
Mills said that “the growth trajectory” of the company meant “containing costs” – and this saw it move from “traditional infrastructure” to serverless computing.
Flexible, cloud-native microservices and access to other cloud-based emerging technology had been instrumental in helping nib to rapidly diversify and build new capabilities, he said.
Supporting more members required automating tasks like triaging policy coverage inquiries with chatbots and processing receipts submitted in claims with optical character recognition (OCR); more actuarial-specific AI projects like “payment integrity models” and “lapse modelling” were also in the pipeline, he added.
Cloud-enabled capabilities
The drive to explore cloud migration in early 2015 was to address “how we get a more focused investment in technology that benefits the customer as opposed to managing data centres and managing that infrastructure,” Mills said.
As well reinvesting data infrastructure expenditure into research and development, moving to AWS was “an opportunity to increase agility” with a more flexible environment.
“If you take the NDIS space for instance, we’ve made seven acquisitions there in about the last eight months give or take; that’s a testament to the fact that we’ve had a capability that’s allowed us to leverage that agility.”
Mills also credited several new features rolled out to nib’s app with the company’s ability to develop and deploy them in a cloud environment.
“If you think about the capability that we now have, just simply within our app today, you can virtually consult with a GP.
“One of our members or customers can check some of their symptoms using an AI-assisted tool; you can complete a skin check or health check; you can search and have a prescription delivered to your home or a medical certificate all within the app ecosystem.”
Trials and tribulations
The cloud migration spanned nine years in part owing to “momentum and prioritisation” variations across the period, as well as “regulatory landscape changes”.
It wasn’t until 2018, for example, that the Australian Prudential Regulation Authority (APRA) relaxed its restrictions on storing and processing customer data in the cloud.
The majority of nib’s front-facing websites and digital platforms were on AWS at that stage, but APRA’s new guidelines meant that nib could “set [a] very ambitious goal of reducing our [data centre] footprint to zero.”
The tiered regulatory system mandates financial institutions to consult APRA after their internal governance processes are completed for any migrations considered a “heightened risk” and that they demonstrate sufficient controls before migrating platforms classified as “extreme risks”.
“In terms of some of the heavy lifting, we completed the first migration of what we call a system of record – and APRA defines as an extreme inherent risk platform – in 2019,” Mills said.
“That was with the migration of GU Health — one of our acquisitions — our corporate health business.”
APRA updated its standards for financial systems’ resilience and business continuity in July last year. The revised “operational risk management” regime, first released in July 2022, added new deadlines for notifying APRA about “offshoring agreements” and incidents “likely” to have financial or operational repercussions, but it won’t affect most outsourced IT and cloud contracts until 2026.
“We spent a lot of time with APRA both consulting directly with them as well as working out various artifacts and capabilities to make sure that we were aligned with their various cloud computing guides. I don’t want to say that it was difficult or hard, but there was certainly a lot of it.”
In addition to scoring regulatory approvals and under-prioritisation, Mills said nib had encountered “various other technical aspects where there have been pieces, which, in hindsight, the product set’s evolved or our applications haven’t lent themselves to a native deployment.”
“We’ve had to refactor applications that we might not have otherwise wanted to, but we were reasonably familiar with some of the challenges we were going to come across when we went in as well.
“Ultimately, we would have liked to have finished earlier, but I think that’s why it has elongated a little; that’s probably been our biggest challenge; balancing those priorities,” Mills said.
New AI use cases in the pipeline
Mills revealed new milestones reached by its customer-facing AI use cases and said that there were also several machine learning projects underway.
The insurer’s ‘voicebot’ nibby, trained on Amazon’s natural language processing models Lex and intelligent search platform Kendra is now nib’s primary support contact and has managed more than 3.2 million customer queries as of January 2024, Mills said.
When asked about new AI projects under development behind the scenes, Mill said, “Obviously, we’re an insurance business, so in the actuarial space is where we’re looking.
“Things like lapse [rate] modelling and intelligent document processing; some of these are in the early stages of maturity, and some of them are moving through to more mature stages.
“We’ve also got a whole program of work around ‘next best action’ and how we do personalised marketing and making sure that our marketing is more on point.
“Payment integrity models as well, being an insurer that’s the core: making sure we’re paying the right amount for the things we have. We have large data sets for claims; payments; so, it’s about making sure that’s on point.”