The evolution of insurance: People, products and promises

In my last column, I wrote about the year so far in insurance, and how climate-related disasters and public policy have put the industry on a path toward crisis much faster than even dire predictions had forecast.

But all is not lost yet. Insurers have an opportunity to position the industry for a brighter future – from finding real value in AI to investing in more resilient communities. In the process, they may also strengthen one of their most valuable assets: their customers’ trust.

Better together: Working with AI to better serve humans

Agentic AI is garnering much of the current hype cycle. However, decisioning itself is not a new concept, or even a new technological development. In 2017, OneConnect introduced a machine learning claims accelerator for “automatic damage determination”; the result was repair costs assessed within a few seconds.

Ten years ago, this use case didn’t represent the vast majority of customer experiences. With the power of today’s AI, speed is the standard, and table stakes for any insurer’s success. As decision autonomy grows, insurers will begin experiencing what’s known as the “interpretability paradox”: As model performance (agentic or otherwise) increases, explainability decreases.

It's hard enough reporting to a regulator on a policy acceptance or claim settlement decision rendered by a human. The report often takes months, if not years, and errors will be found under even light scrutiny. It would be foolish to expect AI to perform flawlessly.

Furthermore, troubling stories like that of OpenAI’s o3-mini model “refusing to shut down” outline how chaotic agentic AI projects can become without governance, explainability and observability.

Rather, combining AI capabilities (traditional, generative or agentic) into existing processes with various levels of human oversight not only adds explainability to the overall process, it perfectly demonstrates Kasparov’s Law – the notion that an average human working with a machine can beat either another lone machine or a more skilled and experienced human without one.

In 1997, chess world champion Garry Kasparov lost a match to IBM supercomputer Deep Blue. The next year, Kasparov returned for a rematch, this time assisted by a machine of his own – and won.

Just as Kasparov adjusted his strategy, insurers will begin infusing AI agents into steps in larger processes to augment the work of underwriters and adjusters to deliver superior customer value – especially if doing so creates a memorable and personal customer experience, making an account more sticky (and profitable) over time.

As insurers lean into decisioning technology and governed agentic AI capabilities, flexible cloud computational power will be critical in balancing costs. Data generating capabilities (like synthetic data) will allow for faster model training and deployment, making those same agentic AI models even more effective.

Some of these investments will seek to bring certain aspects of insurers’ data and AI capabilities back on-site. By 2035, on-premises AI spending in one year is projected to surpass total AI spending from 2015 to 2025 combined – growing over 500% during that period of time.

In other words, reliance on hyperscalers for access to computation prowess, expertise and infrastructure will dwindle, but it will never go away. Furthermore, maintaining data on-premises versus replicating the data in a cloud environment limits its vulnerability to theft or breach.

An insurance policy is a promise. Machines will never entirely replace the people delivering those promises. However, as various steps in risk acceptance and claims settlement processes are automated and overall time to render those decisions becomes ever faster, a single human will be able to manage a larger and larger span of control. The adjuster of the future may look more like a factory supervisor, monitoring various claims settlements in real time from a massive control panel of blinking lights and status indicators, and only intervening when the AI system – employing tens of thousands of agents – alerts the supervisor that a process is out of tolerance. Or an underwriter handling a $1 billion portfolio through a no/low-code model deployment workbench, adjusting rates using daily data ingestion and immediate model retraining.

Those jobs will become AI jobs, because every company will become an AI company by 2040.

About half of all leaders reportedly are looking to a) keep existing staff and b) hire digital workers, or AI, with the goal of increasing productivity. Fortune reports tech layoffs spiking in July to 140, and Simon Sinek predicts that blue-collar work will continue unaffected while knowledge work will be impacted.

In short: The AI revolution is coming.

Insurers can take part by becoming AI-native. Decisions to rebuild homes or repair vehicles occur in fractions of seconds and no longer represent edge cases, but industry standards.

The dawn of resiliency and fortified communities

Indemnification is dead.

2026 will usher in an age of preparedness, proactivity and prevention for the insurance industry.

New product development, like parametric flood policies, can directly address gaps in traditional insurance policy offerings. The event trigger associated with the coverage allows for critical financial assistance to be provided much sooner than a traditional claims process.

And traditional claims processes and insurance products increasingly fail as the world changes around us. We saw the result of 1% flood insurance penetration in Buncombe County, NC, following Hurricane Helene last fall: A devastated community unable to rebuild from meager FEMA payments. 

In the wake of disaster, communities need insurers. The innovation of parametric policies suffers an issue of distribution; most people don’t understand what it is or why they need it. Without a bank requiring hazard insurance to approve a loan, many homeowners would not likely think to purchase traditional insurance in the first place, let alone flood coverage (not typically included in your homeowner policy).

Parametrics are gaining traction, and some pioneering companies are exploring offering those products through employee benefit plans, addressing both the distribution and cost obstacles.

And the business benefit of investing in resiliency has already been quantified. One study found that for every $1 invested in resiliency efforts, $13 in benefits are returned from damage cleanup and economic costs.

If funding for federal programs and disaster relief faces further scrutiny and potential cuts, state and local governments have an imperative to work with insurers to fortify homes and communities and make them more resilient to natural disasters through loss control services and risk assessment.

Efforts such as 3D printing homes or the head-turning story of Babcock Ranch, FL, may sound like fiction, but they demonstrate real-world ingenuity in this space.

What insurer would not be interested in structures that resist mold, wind or water damage?

And technology can now track weather events in real time and alert people in advance. Take Jakarta, the largest city in Southeast Asia, completely connected and capable of warning residents of flood within mere minutes.

Investing in efforts to work with state and local governments and community leaders ultimately will be a wise investment, as costs of materials to fix vehicles or repair homes and businesses continue to increase amidst continued global economic uncertainty.

Most importantly, innovations like these could prevent the loss of life that comes with these tragic events.

The future is what you make of it

While SAS and Economist Impact’s Insurance Forward research does not discuss the cascade failure of the global economy or the complete collapse of the insurance industry, there is a non-zero chance either or both of those events will occur.

As inflationary pressures and economic policies drive up the cost of raw materials, and increasingly severe weather events feed claims costs into loss models, insurance rates will push already marginalized people into uninsurable/unaffordable situations. In this scenario, climate-related deaths and climate refugees displaced from disasters will likely increase to alarming levels, placing severe burdens on public services and weaken aid providers ability to meet the need.

It merits restating: The insurance industry views AI with “cautious optimism.” And maybe it’s in this moment that technology – whether agentic, synthetic, causal or otherwise – holds the key to finding a path forward for our communities struggling to survive.

The insurance business has been around for centuries; it has been well-equipped to weather the storm (pun intended). And right now, it’s looking pretty cloudy. Taking steps to shelter our communities in their times of need not only represents the very best of the insurance industry, doing so would build immeasurable trust for decades to come.