(Image source: IIR/Gemini.)
There’s little doubt that we live in an exciting and vibrant time from a technological perspective. The pace of change continues to accelerate, making remarkable advances possible in what we do—and how we do it. It can be tempting to think we have never seen this before. That would be wrong.
George Santayana is often credited with observing that “those who cannot remember the past are condemned to repeat it.” The point holds. Human memory is short, particularly at institutional scale. Even recent events—pandemics, for example—show striking, if imperfect, parallels across decades.
Mark Twain is frequently paraphrased as saying history doesn’t repeat itself, but it often rhymes. Whether or not he phrased it exactly that way, the insight is apt.
New Tools, Familiar Patterns
Which brings us to what is happening with technology today. As tools such as AI capabilities from Anthropic, OpenAI, and others race to the forefront, it is tempting to view this moment as entirely novel. That would be a mistake. The tools may be new, but the patterns of adoption—and missteps—are not.
Every major technological advance promises dramatic improvements in productivity and capability. In the early stages, however, these tools are typically layered onto existing business systems, where the focus is on incremental gains rather than structural change. Real value takes time to emerge. More importantly, it requires leadership willing to rethink not just technology, but people and process.
Bill Gates famously observed that organizations tend to overestimate the impact of new technologies in the short term while underestimating them over the long term. Whether this reflects human nature or how businesses measure success is beside the point. Recognizing the pattern can help insurers prioritize more effectively—and realize long-term value sooner.
Said another way, not every stove needs to be touched to prove that it is hot.
Productivity Gains—and Their Limit
Today, a wide range of publications tout outsized productivity gains from rapid deployment of artificial intelligence, including agentic systems. This narrative underpins significant investment flows and elevated valuations for companies that prominently feature AI in their offerings—often irrespective of measurable outcomes. To be clear, some gains are real, particularly in areas such as software development.
At the same time, some productivity improvements are difficult to quantify. Code can be generated faster, but downstream work—optimization, validation, and security—may offset a meaningful portion of those gains. A recent Stanford University study highlights this tension.
Lessons from Earlier Waves
For insurers, the more instructive lens may be historical. Prior waves of “citizen computing” delivered short-term productivity gains but introduced longer-term challenges. Tools such as Excel spreadsheets and Access databases enabled rapid development outside formal IT controls. Many implementations were never fully tested or productionized, leading to issues in version control, security, and support. The resulting technical debt took years to unwind across the industry.
Governance models and testing disciplines eventually brought these tools under control, allowing them to deliver sustained value—but in ways quite different from their initial promise.
Anyone who experienced the MS Access era will recognize the pattern. AI tools are similarly accessible, quick to deploy, and evolving at a pace that shortens their useful life. Combined with workforce turnover, this dynamic has the potential to create a new generation of technical debt—possibly on a larger scale than previous cycles.
A Familiar Outcome
This is not a new phenomenon. When COBOL emerged in the 1960s, it was designed to be readable and accessible, even described at times as a step toward eliminating the need for professional programmers. That prediction did not materialize. Instead, decades of accumulated logic—often poorly documented—have made legacy systems difficult to modify or replace. The resulting technical debt continues to constrain many organizations today.
None of this suggests that emerging technologies should be ignored. Rather, it is a caution against “ready, fire, aim” adoption. The echoes are there for those willing to listen.
Insurers that pair AI investment with disciplined governance, process redesign, and long-term architectural thinking will capture its benefits. Those that do not risk repeating a familiar—and avoidable—cycle of technical debt.
Pendulum Swings Continue to Shape Insurance Technology Strategy