This piece is by Keith Viverito, Managing Director of EMEA, Clearwater Analytics
It is 8am on Monday morning, and a portfolio manager at a mid-sized insurer is reviewing a portfolio where nearly half the assets now sit in private credit and infrastructure. Two years ago, getting a clear picture of total risk exposure across that portfolio would have taken three days, four analysts and a web of spreadsheets. Today, it takes a matter of minutes. This difference goes beyond speed – it fundamentally changes activity and the ability to respond.
Portfolio complexity will only continue to increase. Technology evolution now enables a real shift to bring together critical portfolio data in one place and instantly model changes.
For decades, spreadsheets have been accepted as the universal language of insurance investing, and for many years they worked well enough. But the industry has changed. Insurance portfolios are no longer dominated by government bonds and investment grade credit. Increasingly, they include private credit, infrastructure and other alternative assets that generate far more varied and complex data. In some portfolios, these assets now represent 40 to 50 per cent of total allocations. That shift has made something simple far harder. Investors need to understand their true exposures across the entire portfolio, not just within traditional asset buckets.
Many insurers are therefore moving towards what is known as a total portfolio approach, where the portfolio is managed according to overall risk exposures rather than individual asset classes. The challenge is that achieving that view requires every asset, public and private, to be captured in a consistent data structure.
This is where modern data platforms and AI are beginning to make a real difference. When portfolio data is properly consolidated and governed, AI tools can analyse exposures across thousands of positions simultaneously. They can surface concentrations, model how shocks might ripple across asset classes and identify risks that would previously have been hidden in separate spreadsheets. But that capability only exists where the underlying data is reliable and properly structured. Without it, the output is only as good as the data feeding it.
For an insurance investor, that ability to see the whole portfolio clearly is transformative. Speed is the visible gain, but the more important shift is that investors can now ask far better questions of the whole portfolio and get answers quickly enough to act on them.
Questions that once required days of manual work can now be answered in minutes. How exposed is the portfolio to a change in interest rates across both public and private holdings? Where are the hidden correlations between infrastructure debt and corporate credit? How would a shift in spreads affect solvency capital requirements across multiple capital regimes? Each of these questions shapes how insurers allocate billions of dollars of long-term capital.
In that sense, AI is doing something more fundamental than improving workflow. It is expanding what is practically knowable. It gives investors a way to see relationships across complex portfolios that were previously too slow, too fragmented or too labour intensive to uncover in time to influence decisions.
Where that foundation exists, the results can be dramatic. Across the portfolios we work with, this shift can often return 20-30 per cent of an investment professional’s time back to actual investment work, time that was previously spent locating, cleaning and reconciling data before any analysis could begin. Spreadsheets still have a place, but they were never built to serve as the backbone of what modern insurance investment operations now demand. That role is rapidly fading.
As portfolios become more complex and investors adopt a total portfolio approach, the firms that can see risk clearly and early will have an advantage over those still piecing together yesterday’s picture. AI is already changing how portfolios are managed. That question is settled.
The more pressing question is whether their data infrastructure is good enough to make that capability real, or whether they are running a 2026 portfolio on foundations that were never built to support it.
Be the first to comment