Lessons from insurance: Living with uncertainty

Uncertainty has a habit of arriving dressed as something new.

Today it’s climate risk – not just as a scientific challenge, but as a decision-making one. Yesterday it was systemic risk. Tomorrow it will be something else entirely. Each time, the conversation sounds familiar: the models aren’t good enough yet, the data is incomplete, we need more certainty before we can act. The insurance industry has heard this before. For decades. Not because insurers are unusually brave, or unusually clever, but because they’ve never had the luxury of waiting for certainty. Hurricanes don’t pause while we refine assumptions. Floods don’t care how confident we feel in our numbers. And yet, decisions still have to be made. Prices still have to be set. Capital still has to be held.

This experience matters – especially now, as other sectors find themselves facing the same challenges.

Uncertainty is not one thing

One of the first lessons insurance learned is that ‘uncertainty’ is a dangerously vague term. It sounds singular. It isn’t.

Some uncertainty comes from what we haven’t included at all – risks outside the frame, losses no catastrophe model was built to capture. Some comes from imperfect inputs – incomplete data, wrong assumptions, rounded numbers that hide sharp edges. Some comes from nature itself – hazards that are rare, clustered, or simply unpredictable. Some comes from how assets behave under stress. And some comes from people: contracts, courts, politics, and human behavior under pressure.

These uncertainties behave differently. They grow, shrink, cancel out, or compound in different ways. Treating them as one amorphous problem leads to two equally bad outcomes: blind confidence or total paralysis.

Insurance learned, often the hard way, that the real question isn’t “Is this result right?” but “Which uncertainties matter for this decision?”.
That shift alone moves the conversation from argument to judgement.

From prediction to preparedness

There is often an unspoken desire for models to deliver ever greater predictions. Instead, Insurance learned to work with ranges rather than precise points; scenarios rather than forecasts; clarity about what drives outcomes rather than the illusion of precision. The goal was never to predict exactly what would happen, but to understand what could plausibly happen and whether that was acceptable.

This is exactly the challenge climate risk presents today. Climate models are often judged by how precisely they can predict future losses, temperatures, or events. Insurance learned that this is not the right test. The real value lies in understanding ranges, tipping points, and sensitivities to assumptions – what breaks first, what changes fastest, and which decisions become fragile under different futures.

In this way, physical risk models became decision-making tools, not crystal balls. They are very good at answering the questions we know how to ask, and open about their limitations where we have to step in with our own judgment.

And this challenge isn’t unique to insurance. Banks wrestling with climate stress testing, long-term credit risk, or balance-sheet resilience face the same issues. The question is rarely whether a loss estimate in “right” but whether the institution understands which assumptions drive the result – and how exposed it is if those assumptions change.

Best practice vs. silver bullets

Insurance hasn’t solved uncertainty. It has built best practices around it.

Best practice for how models are used – and challenged; how assumptions are documented; how uncertainty is communicated to decision-makers; and for who owns the decisions when the numbers are uncomfortable.

This work rarely makes headlines, but has been what has turned uncertainty from an abstract and paralyzing concern into something that could be governed and worked with.

Crucially, this best practice emerged through real experience. Through events that didn’t behave as expected. Through decisions that looked reasonable at the time and were painful in hindsight. These lessons were learned once – and paid for.

This is why other sectors don’t need to reinvent the wheel. The hardest thinking has already been done. What’s missing is not capability or understanding – but insight to reuse mature practices in other sectors and for other use-cases rather than reopen debates that insurance has already closed.

In banking, for example, the challenge is not the absence of models, but uncertainty over how outputs should inform lending decisions, portfolio steering, or capital buffers – and where accountability sits.

The hardest part was never the science

Numbers feel precise. Charts look authoritative. Decimal places create confidence. And yet, the most dangerous moment in any analysis is often when results are handed to someone who didn’t build the model but must act on its output.

Insurance learned to be wary of two extremes. The first is over-confidence: presenting results with such apparent certainty that no-one feels able – or willing – to challenge them. The second is under-confidence: overwhelming decision-makers with caveats until the safest option becomes doing nothing at all.

The skill of those working in risk management lies in between. Good communication of uncertainty leaves the audience clear on three things: the general direction of the risk, the key sensitivities, and the questions that matter if more precision is needed.

This balance and the associated skills have been hard-won. And they are transferable.

An invitation to stand on the shoulders of experience

This isn’t about insurance telling other sectors what to do. It’s about sharing experience – and saving time.

Many sectors are now confronting the same discomfort that insurance has lived with for years: climate risk shows the future will not resolve itself neatly into a single answer. Waiting for certainty is, itself, a decision – and often the riskiest one.

The insurance sector has spent decades learning how to act responsibly in that space. It has built habits, language, and governance that turn uncertainty from something to be feared into something that can be managed.

The hard lessons are already paid for; the opportunity now is to reuse them – not to claim that uncertainty is solved, but to raise the standard of how we live with it.