The insurance industry thought it understood hurricane risk in Florida. Then, in 1992, Hurricane Andrew made landfall and revealed how much the industry could not see. Twenty years later, the Insurance Information Institute reported in a white paper that Hurricane Andrew had produced $15.5 billion in claim payouts—more than triple the industry forecast—and left seven domestic insurers and one foreign insurer insolvent, while other carriers required parent-company support to pay claims.
Andrew was not just a catastrophe. It was also a knowledge failure. No one was blasé about hurricanes. No one believed coastal risk was imaginary. Yet the industry had written, priced and accumulated risk across a fast-growing market without the systems, models and institutional knowledge infrastructure it needed to see the full picture. What followed Andrew was one of the most important learning moments in modern insurance history. Andrew forced the industry to build better ways of seeing catastrophe risk. It turned loss history, exposure data, reinsurance strategy and modeling assumptions into a more disciplined knowledge infrastructure. For a time, the lesson seemed unmistakable: Memory needed to be part of the operating system.
Then came Hurricane Katrina less than 15 years later. Katrina was not a repeat of Andrew, but it tested many of the same institutional muscles: catastrophe preparedness, claims capacity, capital adequacy and the industry’s willingness to apply lessons from the past. Yet within a decade, underwriters were already warning that those memories were fading, creating the risk of “unintended complacency” about future catastrophic events. This is how insurance forgets: Hard-won knowledge becomes new practice, then old practice, then somebody else’s problem.
The Memory Paradox
Insurance may be the oldest continuously practiced risk management profession on earth. Lloyd’s traces its beginnings to a coffeehouse in 1688. The Philadelphia Contributionship, which Benjamin Franklin and his fellow firefighters founded in 1752, describes itself as the nation’s oldest successful property insurance company.
And yet, for all that longevity, all too often the industry treats history as just a marketing and communication asset or, worse, merely decorative—for example, history may only appear in the form of portraits of long-forgotten industry captains adorning boardrooms. But while a strong heritage can and should be a brand and cultural asset, that’s just the tip of the iceberg when it comes to how the industry’s deep and well-documented history can serve today’s leaders. Properly structured, preserved and activated, insurers’ archives are a key source of memory that can improve judgment.
Insurance has an underwriting cycle. It also has a memory cycle. The Insurance Information Institute describes the property and casualty cycle as moving between soft markets, when premium rates are stable or falling and insurance is readily available, and hard markets, when rates rise and coverage becomes harder to find. Every CEO in the room knows the pattern. The question is why the industry keeps mistaking the next turn of the cycle for something new.
Part of the answer is that lessons learned in hard markets often lose force in soft ones. Discipline that feels obvious after a major loss can begin to look overly cautious when capital is abundant, competition intensifies and growth targets reassert themselves. What began as institutional learning slowly becomes institutional folklore: respected in theory but easy to dismiss in practice.
The industry doesn’t just forget but also does so predictably, cyclically and at great cost. Unlike many sectors, insurance has the receipts—documented in archives, actuarial records, claims files, underwriting manuals, catastrophe reviews and loss histories going back centuries. Often, the information it needs to challenge the cycle already exists. It just is not always accessible, connected or used.
The Cost of Forgotten History
The insurance profession is facing not merely a talent shortage but also a memory crisis. Insurance Thought Leadership recently reported, citing Bureau of Labor Statistics data, that by the end of this year, an estimated 400,000 insurance professionals will have retired in the U.S. since the beginning of 2021. Datos Insights puts the stakes even more bluntly, estimating that by 2036, that lost expertise could cost the industry up to $124 billion annually.
The danger is not only that there will be fewer people in the roles. It is also that the people leaving often carry tacit knowledge that has not been written down, indexed, connected to training or made usable to the next generation. Underwriting judgment, claims instincts, market memory, broker relationships, regulatory context, catastrophe assumptions, product lessons, pricing scars and more: These do not live only in systems. They live in patterns of experience.
The cost of forgetting is not theoretical. Consider asbestos: Milliman notes that asbestos and pollution continue to affect general liability books written decades ago. As of year-end 2019, the U.S. insurance industry recognized approximately $92 billion in asbestos liabilities. That is the brutal nature of legacy losses. Liabilities written in one era have financial repercussions in another.
The broader pattern should sound uncomfortably familiar: long-tail casualty exposures, long-term care pricing assumptions, catastrophe accumulation blind spots, social inflation and emerging risks that look manageable until they do not. Insurance is very good at modeling uncertainty. It is less consistently good at remembering how former certainties have aged.
AI Needs More Than Data. It Needs Memory.
A handful of carriers have begun connecting historical loss data directly to current pricing assumptions using AI. Berkshire Hathaway Homestate’s wildfire underwriting model, trained on 20 years of loss history, is one documented example. But external catastrophe data is not the same as institutional memory. The deeper archives—how your organization priced that risk, what your underwriters believed, what your claims teams learned and what your leadership decided—remain largely untapped.
AI can be a powerful level setter for lost institutional memory. It can help surface what new employees don’t know they don’t know. It can make archives searchable. It can connect decisions, losses, claims, correspondence, research, product history and oral histories in ways that were impractical even a few years ago. Used effectively, it can codify and reduce the risk of losing knowledge such as how to access relevant information and use archaic software or processes from decades past.
But AI is not the solution by itself. It learns from data. If the most valuable institutional knowledge is scattered across shared drives, paper files, retired employees’ memories, uncatalogued archives and disconnected systems, AI will not magically convert it into wisdom. It will produce generic intelligence, not competitive advantage. I made a related argument recently in Fortune: AI may transform how organizations operate, but it still needs records of how they have made decisions, navigated crises and earned trust over time. The age of AI requires infrastructure that preserves and activates institutional memory, including archives, internal documentation, oral histories and digital knowledge systems.
For insurance, that means archives are no longer simply a heritage function. They are also data infrastructure, training data and risk intelligence. They are the corporate DNA that can help AI understand not just what a company knows but also how it gained that knowledge.
From Archives to Operating System
So what should carriers do? Inventory the records that explain consequential decisions: underwriting guidelines, claims reviews, catastrophe response reports, product launch postmortems, board materials, market-entry analyses, regulatory correspondence, internal publications and more. Capture oral histories with retiring leaders and technical experts, especially those whose expertise is so fundamental it has become invisible. Then connect those materials to the systems where work actually takes place.
And finally, start treating institutional memory as an enterprise asset, not just a marketing and communications one. Corporate history and archival line items generally find a home in marketing and communications budgets because these activities are viewed as supporting brand storytelling. But these budgets rarely have the resources necessary to support business transformation projects that require year-over-year commitment.
More than 15 years ago, the CMO of a Fortune 100 insurance company said to me with no sense of irony, “We’re in a period of transformation: We’re trying to make the transition to the 21st century from the 19th.” The comment has stuck with me after all of these years, because I’ve seen firsthand how slow the industry can be to change.
AI raises the stakes, because knowledge infrastructure is now competitive infrastructure.
Humans forget. Risk management is supposed to make sure we remember. That’s why historical archives and institutional memory are not a soft asset in insurance—they’re part of the operating system. Lloyd’s has records dating back to at least 1734. What are you doing with yours?
