From a graduate's limit routine to the CTO's board report

Mandate

Make critical member services resilient, reduce the personal information the fund holds and adopt new technology only where its errors can be supervised.

Outcomes

• Reduced identity records for 512,000 former members to a minimal cover history
• Set and tested disruption tolerances for extras claiming and hospital eligibility
• Introduced a supervised cover-question assistant that cut handling time from 8.9 to 7.4 minutes
• Moved core systems from an owned data centre to a hosted environment by mid-2024

Decisions in context

I took the role when the fund's technology function was combined under one executive for the first time. My predecessor as chief information officer had retired, and the chief executive wanted technology risk reported to the board by the person accountable for it. The first eighteen months went on moving our core systems out of an ageing owned data centre, which reduced one category of risk while creating a dependence on a hosting provider that we then had to manage. The late-2022 attacks reshaped the agenda that followed: data minimisation, breach rehearsal and, from 2024, preparation for APRA's operational risk standard. I report quarterly to the risk committee on cyber security, resilience and supplier performance. I am not a director, and decisions on benefits, premiums and products remain with the executives who own them.

Reason for the move

Current role; seeking a first non-executive contribution alongside it.

Mandate

Replace an ageing hospital claims engine and stabilise point-of-service extras claiming.

Outcomes

• Replaced the hospital claims engine in five benefit-group stages between 2018 and 2020
• Reduced unplanned point-of-service claiming outages from nine in 2016–17 to two in 2019–20
• Required parallel reconciliation of every benefit group before it moved to the new engine

Decisions in context

I came back to health insurance after sixteen years away and found the code I had once written for small funds had grown into much larger systems with the same habits. The hospital claims engine was the most pressing problem. Two specialists close to retirement understood it, and every change carried risk. I chose a staged replacement, one benefit group at a time, with the old engine running beside the new one until results agreed to the cent. The parallel running cost about A$2.5 million more than planned and frustrated colleagues who wanted the programme finished. It also meant that no member received a different benefit because of the change. Alongside the replacement, my team reduced counter-claiming outages by fixing monitoring and release practices rather than buying new equipment.

Reason for the move

Appointed chief technology officer when the chief information officer retired and the role was redefined.

Mandate

Keep the claims management platform aligned with scheme requirements and make case managers' decisions easier to record and review.

Outcomes

• Delivered 23 scheme-mandated changes on schedule across five years
• Replaced free-text case notes with structured decision records for return-to-work plans
• Reduced case-manager time spent on payment reconciliation by about a fifth

Decisions in context

Workers' compensation taught me how long an insurance decision can stay alive. A single claim might involve weekly payments, medical approvals and return-to-work plans over several years, with different case managers along the way. The scheme set our requirements and audited our performance, so my team often implemented rules we had not designed. I learnt to explain the operational cost of a requested change early, and to say when a change would make an earlier decision harder to reconstruct. The structured decision record was the work I valued most. It replaced pages of free text with a short account of what had been decided, on which evidence and by whom. Case managers resisted it at first because it took longer. Within a year they used it to answer workers' questions about stopped payments more quickly.

Reason for the move

Returned to Adelaide and to health insurance.

Mandate

Make policy rating and renewal changes safer and faster to release.

Outcomes

• Moved home-insurance rating from hard-coded tables to a managed rules service
• Cut renewal-batch defects reaching customers from eleven in 2006 to one in 2010
• Kept flood claim lodgement open in January 2011 through a temporary intake database

Decisions in context

This role made me responsible for other people's code at scale. Renewal batches touched hundreds of thousands of policies, and pricing teams wanted changes quickly. I moved rating tables out of program code into a managed rules service, which let pricing analysts change values under review without a full software release. That shift required new testing discipline: every change was run against a sample of the previous month's renewals before release. Defects reaching customers fell steadily. My final months were dominated by the January 2011 floods. Claims arrived far faster than the claims system could accept them, so we built a temporary intake database within two days and reconciled it over the following weeks. I decided afterwards that I wanted to work on claims systems directly, where the consequences for people were most immediate.

Reason for the move

Moved to Melbourne to lead claims systems for a long-running injury scheme.

Mandate

Maintain and extend the rating and renewal code for home and motor policies.

Outcomes

• Rebuilt the motor renewal notice process after a rating defect reached customers
• Introduced regression testing against prior renewals for the rating module
• Led the team that added flood cover options to home policies

Decisions in context

I joined Marrowind to work on systems far larger than anything a small health fund needed. The policy system was older than I was in places, and much of its logic lived in code that few people read. In my second year a rating defect I had not written, but had reviewed, reached several thousand motor renewal notices. Correcting the notices took weeks and taught me that a code review without realistic test data is mostly a formality. I proposed testing each rating change against the previous month's renewals, and the practice became standard after I became a team leader in 2002. Leading six developers was my first experience of being responsible for work I did not do myself. I learnt to ask for evidence rather than assurance, without making colleagues feel accused.

Reason for the move

Promoted to policy systems development manager.

Mandate

Write and maintain claims-validation code for client health funds.

Outcomes

• Rewrote the optical and dental limit routines used by four client funds
• Built test cases from paid claims to check benefit calculations
• Supported two client funds through year-2000 date remediation

Decisions in context

Kestrelby gave me an unusually close view of how health insurance works. Our clients were small funds, often with a few dozen staff, and their claims managers would ring us directly when a benefit looked wrong. I wrote and fixed the routines that applied limits, waiting periods and benefit percentages. Each fund configured its products slightly differently, so a change for one client could break another unless we tested carefully. I started building test cases from real paid claims, with identifying details removed, to check that new code produced the same benefits as the old. The year-2000 work at the end of my time there was mostly patient checking of date logic. I left because I wanted to understand how software behaved at a scale where one defect could affect thousands of customers overnight.

Reason for the move

Joined a larger insurer to work on systems at higher volume.