Writing a newsletter on a morning when the president of the United States threatens that "a whole civilization will die tonight" strikes me as a fool's errand. Whatever I write will quickly pale in comparison with what happens -- or, I hope, doesn't happen -- in Iran in the next 24 hours. 

So I'll keep it short this week, just pointing out something I've been tracking for a while: that a "super" El Niño has become increasingly likely. An El Niño increases surface temperatures in the ocean, leading to higher temperatures worldwide and exacerbating just about all the sorts of natural disasters that have been producing record claims for property/casualty insurers (with the notable exception of Atlantic hurricanes). And the El Niño that's now forming looks like it will produce a far greater increase in ocean temperatures than normally occurs -- perhaps the greatest in 140 years. 

I'll take a quick look at the likely effects -- and at why U.S. insurers need to be even more careful than usual about their public statements and handling of claims, given that the insurance industry is an easy target for politicians looking for scapegoats in an election year.

Then we can all get back to our doom scrolling. 

A Washington Post article does a thorough job of laying out the risks of the El Niño that is forming. It says a "super" El Niño is one in which a key part of the Pacific Ocean sees surface temperatures increase by more than 2 degrees Celsius (3.8 degrees Fahrenheit) above average. The El Niño now taking shape could see the temperature rise 2.8 degrees Celsius (5.04 degrees Fahrenheit) above average, breaking the record set in 2015.

The result, the article says, could be: 

• "Reduced hurricane activity in the Atlantic Ocean and possible drought in the Caribbean islands. Increased hurricane and typhoon risk in the Pacific Ocean....
• "Potential drought in central and northern India....
• "Above-average summer temperatures and humidity in the Western United States, possibly coming with unusual downpours, which may reach into the Plains and extend severe thunderstorm season.
• "Developing droughts in portions of Central Africa, Australia, Indonesia, the Philippines, some South Pacific islands, Central America and northern Brazil, particularly later in the year. Flooding downpours in Peru and Ecuador, parts of northern and eastern Africa, the Middle East and near the equator in the Pacific.
• :Higher frequency of heat waves across large parts of South America, the southern United States, Africa, Europe, parts of the Middle East, India and eventually Australia.
• "New global temperature records — especially in 2027 — probably breaking records set in 2024."

These threats come as high temperatures and low precipitation have created drought conditions across more than half the continental U.S. The problem is especially severe in the West, where devastating wildfires have become all too common. 

So, while the lack of a landfalling hurricane in the U.S. last year meant insured losses from natural catastrophes were below the average for the past decade, claims could well soar this year.

Natural disasters always draw complaints about insurers. Everybody wants the recovery to be fast and smooth -- in situations where it's almost impossible for anything to happen fast or smoothly. Politicians, always looking for a way to position themselves on the side of voters, will home in on problems and publicly scold insurers -- a week ago, President Trump made a social post calling State Farm "absolutely horrible" for its handling of the Los Angeles fires early last year, and we can expect a lot more sniping at insurers from all sides in the run-up to the fraught mid-term elections this November. 

So everyone in the industry needs to be on high alert as the El Niño develops, helping people make their homes and properties more resilient and then, when the inevitable losses occur, acting as swiftly and empathetically as possible to help people recover.

In the meantime, let's all hope and pray for a sane resolution to the man-made disaster taking place in the Middle East.

Cheers,

Paul

The race to fill senior leadership roles—and to do it faster, more competitively, and with greater diversity—has become a defining challenge for CEOs and CHROs. Across industries, organizations report a limited pipeline of "ready-now" talent, particularly at the highest levels. This raises an important question about whether the gap is one of supply or of visibility.

At the same time, women's advancement into senior leadership continues to lag despite years of focused initiatives and investments. Recent Women in the Workplace research from McKinsey and LeanIn.org attributes part of this to a shift in ambition.

But what if the issue isn't ambition at all? What if the signal is being misread?

What the Data Shows and What It Doesn't

For more than a decade, McKinsey's annual research has shaped the national conversation on women's advancement, offering critical visibility into representation and career progression. The data reveal a consistent pattern: Women enter the workforce in strong numbers, advance early, and then experience slower progression at senior levels. In fact, recent reports have sparked a broader discussion on whether gender parity will be reached in 50 years and whether there is an ambition gap.

Yet this survey research data tells only part of the story. It maps where women are and the movement, but it does not fully illuminate the experience of leadership once they arrive. That deeper understanding requires a different lens—one that brings into focus how advancement actually unfolds.

What the Experience Reveals

As part of the research for SelfPowerment: The Inner Shift for High-Achieving Women Who Want More Than Just Success, a qualitative study was conducted with 52 senior women leaders and 10 male executives across industries in the United States, including meaningful representation from the insurance sector across property & casualty and life & annuities.

Rather than focusing on representation alone, this research examined lived leadership experience—how careers begin and evolve, and ultimately how advancement unfolds once performance has already been proven.

What emerged is a pattern that is widely experienced yet rarely articulated. High-achieving women are consistently delivering at the highest levels. They are leading enterprise transformation, running complex organizations, and driving the outcomes companies rely on for both growth and performance.

And yet, many describe a subtle but consistent shift. Performance continues while advancement slows. What surfaces is not a question of capability; it is one of visibility.

As outlined in the SelfPowerment white paper, "What No One talks About—But Women Know," this dynamic forms what is defined as the Invisible Advancement Cycle—a repeatable pattern in which leaders become indispensable to execution while authority, sponsorship, and progression fail to keep pace.

The Misinterpretation of Ambition

From the outside, this dynamic often presents as disengagement. Over time, some women step back from pursuing the next role and become more selective in the opportunities they consider. In some cases, they choose to leave the organization altogether.

This shift is frequently interpreted as a decline in ambition. A closer examination of the lived experience reveals something far more grounded: When sustained high performance no longer consistently leads to advancement with greater influence or authority, leaders recalibrate how they engage. They become more intentional about the roles they accept, seeking opportunities where responsibility is matched with decision-making authority, where visibility translates into influence, and where increased scope aligns with how they define success.

Ambition remains. But it evolves, becoming more focused, more deliberate.

Where Alignment Changes Everything

One of the most important insights from the research is this: Not all women stayed in this invisible advancement cycle. Those who were able to sustain both influence and fulfillment did so because they had made a shift. They moved from endurance to alignment.

These women took ownership of their careers rather than wait for validation. They set clear, strategic boundaries around roles that expanded responsibility without corresponding authority. They repositioned their leadership from execution alone to enterprise-level impact, and they defined success on their own terms—beyond title or traditional progression.

This is the foundation of SelfPowerment—a return to purpose and a renewed ownership of one's career on one's own terms. It reflects a fundamental shift from, "Will they choose me?" to "Do I choose this?"

Alignment changes how women engage—with greater clarity, confidence, and conscious choice.

Why This Matters to CEOs and CHROs

This realignment extends beyond a women's issue; it is an enterprise leadership imperative. When this alignment pattern persists, organizations begin to absorb hidden costs that often go unrecognized until they become systemic.

Leadership capacity becomes underleveraged as the leaders closest to execution—those who understand how strategy truly operates—remain outside core decision-making circles. Succession pipelines narrow over time, with organizations looking externally for leadership while proven internal talent remains underused. At the same time, dependence concentrates within a small group of high-performing leaders who carry disproportionate responsibility for outcomes.

As alignment erodes, experienced leaders begin to step back or disengage, and gaps emerge between an organization's stated commitments to leadership development and the reality of lived experience.

As the research makes clear, these are not abstract dynamics—they are operational, financial, and strategic in their impact.

The "Operational Leader" Blind Spot

One of the most consistent insights across industries is how leadership continues to be evaluated. Strategy is often defined through vision, narrative, and positioning, while execution—where strategy becomes real—is frequently categorized as operational, tactical, or supportive.

In today's environment, however, execution is where complexity resides. Leaders who integrate technology and operations, guide transformation, and deliver enterprise outcomes often hold the deepest understanding of how the business truly functions. But when execution is undervalued in advancement decisions, organizations inadvertently overlook the very leaders they depend on most.

A More Accurate Question

Rather than asking whether ambition is shifting, organizations are better served by examining how leadership is defined and rewarded. This invites a more precise set of questions: whether visibility is being equated with leadership, whether performance is being recognized or simply relied upon, and whether the leaders most critical to outcomes are also those advancing into positions of influence.

When the answers begin to diverge, the issue becomes clear. It is not ambition that is changing. What is changing is the alignment between performance, authority, and advancement.

A Call to Action

This moment presents a powerful opportunity for both organizations and the leaders within them.

For CEOs and CHROs, the imperative is to make the invisible visible. This begins with a deeper examination of how advancement decisions are truly made, by elevating execution leadership into core strategic conversations, and by ensuring that authority aligns with demonstrated impact rather than perception alone.

For women leaders, the opportunity is equally significant. Recognizing this pattern creates the ability to respond to it thoughtfully. When something has felt misaligned despite continued success, it often reflects a structural dynamic rather than a personal one. Alignment—SelfPowerment—becomes the pathway forward, enabling leadership with greater clarity, confidence, and conscious choice.

The Leadership Shift Ahead

The next decade will require a more integrated model of leadership—one that values not only vision but also execution; not only strategy but also translation into outcomes; and not only performance but also alignment.

In many organizations, this leadership already exists. The capability is present, the experience is proven, and the impact is measurable.

The question is not whether the talent is there.

The question is whether it is fully seen—and whether women are fully choosing it.

You can preorder the book here:. You can download the white paper here.

Someone on your team just demoed how ChatGPT or Copilot can extract data from a medical report in seconds. Now leadership wants to know why you're paying a vendor six figures for document processing when the same AI is available for $20 a month.

It's a reasonable question. Personal AI tools and operational AI systems solve fundamentally different problems, even when the underlying technology looks identical.

When an adjuster pastes a claim's medical records into ChatGPT to prep for a call, that's a one-time task. They provide context, review outputs, fix what needs fixing. The AI just makes them faster at work they were already doing.

When AI processes hundreds of documents a day as part of an operational workflow, a person isn't shepherding each one through. The AI becomes one component inside a larger system that needs to work reliably at scale.

The problem is these two use cases get treated as interchangeable. Organizations sometimes spend months building infrastructure for tasks an employee could handle with ChatGPT. Other times, someone proposes using ChatGPT for a process that actually requires serious engineering. Both mistakes come from not recognizing when a task requires a product. The telltale signs are integration, operational complexity, and repeatability.

Integration

ChatGPT works with whatever you paste into it. It can't reach into your claims system, query your policy database, or update a file on its own. In production, the AI sits in the middle of a pipeline. Data has to get in, and data has to get out

On the input side, documents arrive through fax servers, email integrations, and carrier portals. On the output side, extracted data has to be written to the system of record and validated against what's already in the claim file, and exceptions need to be flagged and routed. Without the integration work on either side of the AI, the system can't function in production.

Operational Complexity

When someone uses ChatGPT to extract data from a document, they're the orchestration layer. They decide what to paste in, what to ask for, what order to work through it. If something looks wrong, they adjust and try again. That works when you're processing one document at a time.

At scale, software has to do that job instead. Documents need to be normalized into a usable format. Illegible or corrupted files need to be handled. Outputs need to be validated, exceptions routed, and results connected to downstream systems. When something fails halfway through, the system needs to know where it stopped, what succeeded, and how to recover.

There's also the question of proof. When a claim gets litigated or an auditor asks how a decision was made, "the AI said so" isn't an answer. You need to show exactly where in the document a value came from and why it was interpreted that way. Personal AI tools are black boxes. Enterprise systems build in traceability because insurance requires it.

Finally there is the file size. A single claim file can run 10,000 pages. You can't paste that into ChatGPT. Personal AI tools have input limits that make documents like these impossible to process in a single pass. At that point, you're not using a tool. You need a product.

Repeatability

Personal AI tools are inherently variable. Ask ChatGPT the same question twice and you'll get different answers. When drafting a strategy document, this can actually help. Running the same prompt multiple times gives you different angles to choose from.

At operational scale, variability becomes a liability. A diagnosis code extracted one way in the morning might come out differently in the afternoon. Tags get applied inconsistently. Provider names normalize differently across batches. A claim that gets flagged high-priority on Monday might score as routine on Tuesday. These inconsistencies create problems throughout downstream processes.

When outputs are unreliable, users lose trust. They start checking everything manually, which defeats the purpose. Enterprise implementations address this through standardization: controlled prompts, validated extraction logic, versioned models, systematic testing. When something changes, you know what changed and why. When something breaks, you can trace it back. This infrastructure is what makes production systems require real investment, but it's also what makes them suitable for production.

When You Don't Need a Product

But the opposite mistake is also common. Not everything needs a product. If the task is infrequent and doesn't need to connect to anything, a person with ChatGPT can be the right answer.

A couple of times a year, a supervisor prepares for a mediation on a complex claim. They need to review the medical records, understand the treatment history, and build their argument. Someone sees that and thinks: we should build a tool for this. However, ChatGPT can help them work through it directly, surface key details, summarize sections. That's a tool making someone better at their job, not a workflow that needs to be automated.

The same applies outside of claims. Quarterly management presentations. Strategy preparation for a renewal. Evaluating a vendor. One-off policy research. These happen a few times a year, the output goes into a document or slide deck, and nobody needs the data anywhere else. Building automation around them solves a problem that doesn't exist.

The person doing the work already has what they need. They have the data, they understand the context, they'll review and edit whatever the AI produces. The value of personal AI tools is that they require no infrastructure. Let people use the tools directly, get useful output, and move on. Trying to systematize that just adds overhead without adding value.

Different Problems, Different Approaches

Personal productivity works because a human handles everything around the AI. They provide context, review outputs, catch errors, make decisions. For these use cases, give people access to AI tools and get out of their way.

Operational automation requires software to do what the human does in the personal productivity scenario. Integration with existing systems. Repeatable outputs. An application layer that makes the AI's work usable by others. That's a product.

The underlying AI might be identical in both cases. The difference is what surrounds it. If the task requires integration with other systems, that's a product problem. If it requires orchestration across large or complex inputs, that's a product problem. If it requires consistent, auditable outputs, that's a product problem. The more of these that apply, the further you are from something a person with ChatGPT can handle. If none of them apply, you probably don't need a product at all. Give someone the tool and let them work.

Document ingestion is one example. However, the same pattern holds for triage, fraud detection, subrogation, anywhere AI moves from assisting one person to running inside a workflow. The question isn't whether AI can do the task. It's whether you need a tool or a product. Get that wrong and you'll spend six months discovering why the vendor charges what they do, or build a system for something that just needed a person with ChatGPT.

When it comes to AI in the insurance industry, 2026 is shaping up to be a defining year. Adoption is accelerating, investment is increasing, and nearly two-thirds of independent agents say they're optimistic that AI can support their work.

But putting generative AI to work inside an independent agency, MGA, or carrier doesn't happen with the flip of a switch. Successful AI implementation takes an operational shift through disciplined execution, clear governance, and a defined path to measurable benefits. Most importantly, it requires investment in your most valuable resource: your people.

That was our guiding principle when we implemented enterprise AI tools at Vertafore. In 2025, we set out to equip ourselves and our teams to use AI in the best way possible. Our goal was to learn how to make ourselves, teams, and products better for the end customer.

To accomplish that, we committed to pausing business as usual for a full week as a company to give our global team time to immerse themselves in AI confidently, responsibly, and creatively. We made a deliberate decision to treat employee confidence as the primary KPI of our AI transformation, and we measured it before and after implementation.

Before expanding access, we concentrated on three priorities: mindset, training, and governance. Here's what happened and takeaways for businesses that want to find success with AI.

Employee confidence sets the foundation

Understandably, some employees fear that companies want to use AI to replace their human workforce. That may be the case in some organizations, but many businesses—including Vertafore—see AI as a tool to help make work more manageable, improve efficiency, and create space for higher-value thinking and better customer service.

The key to managing employee fears is transparency and communication. We addressed employee concerns early and reinforced our intent to use AI to support our human teams, not replace them. That's especially true in a relationship-driven industry like insurance, where judgment and accountability cannot be outsourced to a model.

We reinforced that human oversight remains central, established clear security guardrails, and created structured opportunities for employees to ask hard questions and understand expectations. This included a simple framework: asking "Should I?" instead of "Can I?" Employees were encouraged to consider who benefits, who could be harmed, and whether AI was truly the best fit for a given scenario. The framework reinforced that AI supports professional judgment rather than replacing it.

To measure the impact, we surveyed employees before and after our immersion week. We didn't focus only on productivity metrics or efficiency gains. We measured employee confidence. We positioned AI as a productivity multiplier and asked one core question: How do I make myself, my team, and our products better for the end customer?

The results speak for themselves: confidence rose 13% and those that were "not confident" dropped 11%. Employees who saw AI as a "go-to collaborator" doubled and expected daily or near-daily use rose 27%. Among people managers, confidence using AI rose 23%, with 80% saying they're confident with helping their teams use AI responsibly.

Create space to learn

To help employees build real skills, we made a deliberate choice to set aside dedicated time for learning.

Instead of squeezing experimentation between daily responsibilities, teams were given bandwidth to explore AI. Leaders set up time for employees to test ideas, collaborate across teams, and explore ways to apply AI to their specific roles. And as an organization, we created full-company sessions to dive into AI questions and hear from experts on their lessons learned from putting this technology to work.

Prioritizing learning in this way accelerated adoption. Our teams spent time getting to know the AI tools in the best way possible: by seeking solutions to the challenges and opportunities they encounter every day. Not every idea and trial made it out of the gate. And that's okay. But many of our teams built the foundation for real AI applications.

Internally, our learning and development team created a custom GPT to help employees and managers build individual development plans that support career growth. The tool guides users through identifying where to start, developing skills within their current role, or preparing for a move into another internal role. By following prompts and exploring development ideas, employees can create clear, actionable plans while discovering recommended resources such as LinkedIn Learning and other company tools.

How to measure what matters

If employee confidence with AI increases, it stands to reason that successful adoption will follow. With that in mind, we treated the mindset shift as the first KPI. From there, we narrowed our focus.

Rather than attempting to transform the entire organization at once, we identified one to two areas where return on investment (ROI) could be clearly measured within a defined timeframe. Functions such as customer support often provide the clearest early signals because call times and quality improvements are easier to quantify.

For each use case, we established a clear hypothesis to answer:

• Where will AI create value?
• What outcomes should we see within the first 90 days?

Within customer support, for example, we used AI to refine knowledge base articles so representatives can quickly access the most up-to-date product information during calls. Measurable outcomes include reduced time spent searching for information, faster turnaround between calls, and higher customer satisfaction.

Another area we tracked was product development, with defined objectives around output volume and delivery timelines. We've established targets to measure progress and ensure consistent improvement over time. Peer experts also led office hours that created a safe space to test ideas, ask questions, and build confidence.

By focusing on a specific department, establishing clear goals, and tracking quantifiable metrics, such as time saved, communication volume, and call response time, you can better determine the next steps for scaling AI across your teams.

Lessons learned

Our early efforts drove real engagement, but they also exposed some opportunities. Not every team benefited in the same way, and not every employee felt ready to apply what they'd learned. That tension helped clarify what actually drives progress. Here are the lessons we took forward:

1. AI adoption isn't one-size-fits-all. Teams across the organization have very different AI use cases and skill levels. All-company sessions, such as discussions with external business leaders, offered useful perspectives but did not give employees the role-specific guidance they needed to apply AI in their daily work.
2. Progress doesn't happen at the same pace. Some individuals jumped in quickly, while others needed more context and support. Moving forward required patience, targeted enablement, and learning experiences tailored to where people actually were rather than where we expected them to be.
3. AI success starts with mindset, not mandates. Tools alone don't drive transformation. Confidence does. When employees understand the "why" and feel equipped to experiment, adoption follows. That's where change management becomes a true advantage. Organizations that invest in building understanding and trust are best positioned to turn AI from a tool into a differentiator.

A clear path forward

AI transformation begins with people. We learned that dedicating focused time for learning helps employees step away from daily work and fully engage in building new skills.

Pairing that time with privacy and security training ensured employees explored AI tools while understanding the company's protocols for responsible use. When employees understand the purpose, feel supported as they learn, and trust the guardrails in place, AI stops being an abstract initiative and becomes part of how the business operates. In insurance, where judgment and relationships matter, that human foundation is everything.

Invest in your people first. The technology will follow.

What if your car insurance reflected how you actually drive, not just who you are? That question is no longer hypothetical. In 2024, more than 21 million U.S. policyholders shared telematics data with their insurer, according to IoT Insurance Observatory research. That reflects a 28% compound annual growth rate since 2018. Usage-based insurance (UBI) is no longer a niche; it’s a mainstream strategy reshaping our industry.  

For years, competitive pricing drove adoption. But today, something deeper is at play: Trust and perceived value are fueling the next wave of growth. This isn’t just about saving money; it’s about believing the insurer will use sensitive driving data responsibly and deliver tangible benefits in return. 

Source: IoT Insurance Observatory field customer surveys

According to a recent consumer survey by Arity and the IoT Insurance Observatory - sampling 2,059 personal auto policyholders representative of the U.S. market - 82% of policyholders would recommend a telematics app that rewards safe driving, offers feedback, provides crash assistance, and delivers other valued services. Among drivers under the age of 53, that number exceeds 90%. Positive sentiment toward telematics has steadily increased over the past decade, as shown in the chart above.

Trust isn’t a buzzword here, it’s the foundation of adoption. Consumers share data only when they believe insurers will protect it and use it to create real value. In fact, 53% of respondents expressed high trust in insurers’ handling of personal data, ranking insurers second only to banks. That trust translates into action: willingness to switch plans, share driving scores, and pay for connected services. 

The willingness to adopt UBI is strong: 60% of policyholders are open to switching, rising to 72% among younger drivers. This level is consistent with the evidence from recent TransUnion surveys showing that 60% of people reported being offered telematics opted-in.

When consumers see clear benefits, privacy concerns fade. They want pricing that reflects lifestyle, rewards for safe driving, and features like automatic crash assistance. Three-fourths are open to sharing their driving score for a personalized quote. More than half of those willing to switch prefer pricing models that offer bigger potential savings, even if it means some risk a surcharge. 

This is where telematics shines. Insurers can deliver compelling offers because telematics unlocks incremental economic value by sensing events, transmitting data in real time, and applying AI-driven analytics to understand, decide, and act. This enables smarter underwriting, faster claims processing, and more proactive risk management. By sharing part of this value with policyholders, insurers create a win-win scenario that makes UBI not just viable - but mainstream.

Source: IoT Insurance Observatory

Over the past decade, insurers have proven the power of telematics data to transform core functions: 

• Continuous Underwriting: Telematics data enables more accurate risk assessments and selection, allowing insurers to better match rates to actual risks. This leads to more sophisticated pricing, improved retention and effective acquisition of good risks, and reduced premium leakage from riskier drivers. Insurers can also use telematics-based data to make portfolio-level decisions regarding risk appetite and reinsurance.
• Enhanced Claims Management: Real time crash detection is a game-changer for claims management. Insurers can trigger proactive responses, notify emergency services, and initiate the claims process. Insights about crash events support timely claim handling and help minimize potential fraudulent or inflated requests.
• Connect and Protect: International telematics-based experiences demonstrate effectiveness in mitigating risks by identifying risky situations in real time and intervening before accidents occur.  Behavioral change programs promote safer driving, leading to fewer accidents and lower loss ratios. 

Policyholders are willing to reconsider their insurer when pricing reflects how they actually drive and live: More than half of policyholders would switch for a product with the premium defined by telematics data. 

Consumers aren’t just looking for lower premiums; they want features that matter. Rewards for safe driving and automatic emergency assistance in severe crashes rank among the top preferences across all generations - from Gen Z to Traditionalists. And the appetite for innovation doesn’t stop there: More than half of policyholders would pay $4.99 per month for a connected dashcam service that offers emergency assistance, video recording for protection against unfair complaints, and real-time safety feedback.

The impact goes beyond individual policies. When the usage of telematics data is holistically adopted across the entire insurance organization, this improves pricing accuracy, reduces losses, and makes insurance more affordable, all while promoting safer roads. Fewer accidents mean fewer injuries and lives saved. That’s why telematics is more than a business strategy, it’s a social good.

The time to invest in telematics mastery is now. Insurers that fully embrace the connected paradigm in all their core processes and responsibly use data with customer consent can unlock greater value—delivering fairer pricing, personalized experiences, and safer roads. This broader usage of data enables higher value creation and sharing, benefiting policyholders and society as a whole.

Fraud is hardly a new problem, but it is a serious issue, and recent fundamental changes in societal norms are exacerbating fraudulent conduct and making detection and deterrence less of a priority than warranted. The scope and scale of fraud are truly shocking, especially among government-funded medical and social programs currently under scrutiny, where enormous costs are somehow tolerated.

Fraud not only creates significant economic loss but also undermines confidence in its public and financial institutions, including insurance. Yet preventing and combatting fraud is seemingly episodic and random. 

All of this serves to bring renewed attention to the long-standing concerns about ever-expanding fraud in general – and specifically insurance fraud. Insurers need to heed the wake-up call.

COST OF INSURANCE FRAUD

Quantifying insurance fraud's impact is difficult and spans from premium fraud to claims fraud, whether opportunistic or through deliberate scheme. According to the Coalition Against Insurance Fraud (CAIF), insurance fraud costs American consumers more than $300 billion a year. This amounts to an individual policyholder $900 annual “tax,” as insurer costs are passed on in form of premiums. Claims fraud is said to occur in about 10% of property-casualty insurance losses. Medicare fraud alone is estimated to cost $60 billion every year.

There are also several limitations when it comes to detecting fraud. According to the National Association of Insurance Commissioners (NAIC), there are key differences between “hard” and “soft” types. Soft forms of fraud are widespread and can be a common exaggeration of a legitimate claim. Hard types are described as intentional acts to create or fabricate “damages” and claims. Still, these general headers fall short of telling the whole story. 

Claim fraud can be perpetrated by an individual or involve others including organized crime rings recognizing there are entire ecosystems designed to inflate, embellish and even fake an accident. Billing for unperformed medical procedures pales in comparison to fake “victims” being paid to undergo surgery. A single case in New York uncovered a $31 million scheme between a doctor and lawyer in trip-and-fall “accidents,” paying "victims" to endure surgery, simply to initiate a claim, justify damages or both. So-called runners are paid finder-fees to produce participants.

Further, many frauds go undetected for long periods or are missed altogether because there is much reliance on the “honor system,” whether at point of sale in which premiums are based or when making a claim. Although any healthy system checks and verifies, it is impractical, unnecessary and risky to deeply investigate a large percentages of cases. Insurers balance customer service, state regulatory requirements involving timeliness and potential complaints that can escalate to lawsuits. 

Meanwhile, internal special investigative units (SIUs) likewise have finite resources and bandwidth, only concentrating on the most actionable cases. Law enforcement agencies have similar constraints, and insurance fraud is a lesser priority than other crimes. Altogether, this dilutes the efficacy of combatting fraud, leading to uncaptured and under-reported figures.

Instead, anecdotal case examples tend to do the best job of illustrating the magnitude of fraud. Phony medical clinics, staged auto accidents, even faked deaths demonstrate the amounts at stake and the lengths fraudsters will go. More frustrating is how obvious some of the schemes are, revealed as in the infamous empty day care center stories. 

But what happens when technology pushes the boundaries beyond such traditional fraud methods?

The Yin and Yang of AI and Insurance

The rapid emergence of artificial intelligence has brought greater business risks, and the financial services industries are among the largest victims of related fraud. Ironically, business is quickly learning to harness the power of AI to fight fraud more effectively – but so are the fraudsters. 

The potential of AI in claims fraud detection is among the most powerful applications, and particularly so in life & health and accident, according to a February 2026 report from Gallagher Re and CB insights: "Global InsurTech Report."

AI has many benefits. It can improve efficiency, help make better decisions, and encourage innovation across different industries. But these advantages also come with serious risks – especially the potential for misuse in fraud or deception.

Like any powerful technology, AI can be used for both helpful and harmful purposes. This makes strong and thoughtful governance essential to maximize its benefits and protect against misuse.

Hackers and other criminals can easily commandeer computers operating open-source large language models (LLMs) outside the guardrails and constraints of the major artificial-intelligence platforms, creating security risks and vulnerabilities, researchers said.

Hackers could target the computers running the LLMs and direct them to carry out spam operations, phishing content creation or disinformation campaigns, evading platform security protocols, the researchers said. Roberto Copia, director at IVASS Inspectorate Service, spoke about this issue at the 4th National Congress of the CODICI Association in 2025. He pointed out a growing concern: While AI can improve the efficiency of the insurance industry, it can also give fraudsters more advanced tools to commit fraud.

AI and Insurance: An inseparable alliance

AI is cautiously becoming an indispensable tool in the insurance sector. Its applications range from risk assessment to customer services, claims processing and fraud detection. Predictive algorithms, neural networks, and machine learning models allow the processing of vast datasets, improving underwriting accuracy, accelerating claim settlements and strengthening insurers' anti-fraud capabilities.

But these very tools – powerful, scalable and increasingly accessible – are also being weaponized by fraudsters. “Those who seek to commit fraud are often skilled innovators – frequently one step ahead of those tasked with stopping them,” Copia has said.   

A quantum leap in criminal sophistication

Insurance fraud has always been a structural problem in the sector. Yet today, it’s undergoing a qualitative shift. We’re no longer dealing solely with fraudulent damage to property or fictitious claims. Modern fraud is digital, automated and highly sophisticated. AI has become a powerful enabler for those seeking to manipulate data and images, forge documents or create false digital identities.

A March 2026 report, Verisk State of Insurance Fraud Study, finds, based on surveys of 1,000 U.S. consumers and 300 insurance claims professionals:

• 55% of Gen Z say they would consider editing a claim photo or document
• 98% of insurers say AI editing tools are fueling digital fraud
• Only 32% of insurers feel very confident about detecting deepfakes
• 69% of consumers believe fraud will raise premiums for all policyholders

A paradigmatic example is the Ghost Broker scam: insurance websites that appear legitimate, often employing advanced social engineering techniques, real logos, and data stolen from unwitting intermediaries. AI allows these fraudulent portals to appear increasingly credible, complete with chatbots simulating customer service, AI-driven profiling of potential victims, and the delivery of highly personalized fake offers. The result is a seemingly flawless customer journey. But the buyer is left uninsured and unknowingly defrauded until subsequent inspection reveals the deception.

Another example involves "synthetic" identity fraud, in which fraudsters create an identity with a mix of fabricated credentials. According to Lexis Nexis, fraudsters may create synthetic identities using potentially valid Social Security Numbers (SSNs), with accompanying false personally identifiable information (PII). This newer challenge raises the bar for insurers to validate identity at point of sale and other policy lifecycle stages.

THE FRAUD FIGHTING IMPERATIVE

We believe that insurers have an obligation to prioritize fraud detection and avoidance in this growing, too-big-to-ignore dynamic. This obligation is moral, economic and legal. An insurer’s duty to its policyholders includes protecting their investment while managing fair and accurate premiums alike.

Paul Carroll 

With AI transforming underwriting, some say the function is entering a new phase. Do you agree? 

Katie Klutts Wysor 

When you look at what property and casualty carriers are saying and what brokers are starting to say, there’s broad recognition that generative AI could reshape risk assessment and underwriting in meaningful ways. It’s becoming an enabler. It can help underwriters make better decisions and work more effectively. 

Carriers are focusing on underwriting as a function and investing heavily in it. They’re talking about that focus in investor days and earnings calls, and they’re doing significant work internally to organize data and update processes—improving speed, increasing automation, accelerating turnaround times and supporting more informed decisions with better data. 

What’s even more significant than the process improvements is what AI could do more broadly. Think of the underwriter as managing capital and trying to direct it to the ideal place. How can AI help define the parameters around what the carrier wants to write so it can deploy capital where it is targeting stronger returns? From there, you can align appetite and business mix with the underwriting process. 

For example, if I’m a regional workers’ comp player and I want to expand into other lines of business or other states, how can I use AI to support that portfolio management decision and direct capital more effectively? Then, how do I identify the necessary distribution partners to find the business I want? How do I create the proper incentives for distribution partners to bring forward that business, so I have a submission to underwrite? And then, how do I make sure the underwriting process, and decision-making aligns with my appetite? 

I think that’s where a great deal of value could be created from an underwriting perspective, looking at how AI can help inform research on the front end, and how you then use something like a GenAI-enabled underwriting platform to begin systematically embedding strategic capital decisions into appetite, process and guidelines at scale, so underwriters evaluating risks are working from more relevant information. 

Then you can use AI to respond to new business decisions more quickly, respond to renewal decisions more effectively, and potentially take certain actions during a policy’s term to support risk mitigation conversations. 

If you can start mastering that link—how you’re deploying capital and setting appetite, all the way down to those micro process decisions—that represents a new level of maturity. 

Paul Carroll 

Speed has become a significant competitive factor in insurance underwriting. If you’re slow to quote, even with a slightly better price, you may lose the business. What’s happening in terms of speed in the underwriting process, and how do processes need to change—not just the technology—to take advantage of the speed AI can offer? 

Katie Klutts Wysor 

Speed to quote and bind means something very different across varying lines of business. In auto insurance, you need to be able to deliver a quote in seconds. So, you see many personal lines players focusing on quote simplification. In auto, it is close to a mature problem, and many carriers are following established market patterns to stay competitive. 

But the speed question gets harder as you move into more specialized or complex lines in personal insurance and small commercial, middle market and large commercial. 

What we’re seeing there is impressive. Capabilities are now emerging to triage submission intake. From a technology perspective, carriers increasingly can take a submission no matter how it comes in—via email, a platform or another channel—and combine it with what they already know about that risk, along with relevant third-party data. 

At that point, it becomes an execution challenge. How can you more systematically get to a quick yes, no, or maybe on appetite, and then move effectively  toward a quote? 

How fast that can happen depends on the line of business. But that point is right: Carriers should move as quickly as their competitors. If you’re slow, distribution may not be willing to wait. 

And the benchmark will continue to move. 

Paul Carroll 

How far along is the insurance industry in using technology to allocate capital more intelligently, and what needs to happen to reach the next level? 

Katie Klutts Wysor 

It’s less a technology challenge and more a business decision-making challenge. Some players in the market are especially strong at this, and you can see that by looking at underwriting returns over time. Those companies have consistently used technology and data to manage their portfolios and allocate capital with greater precision, and they will likely continue to adopt new approaches as the tech capability improves. The timeline for broader adoption of newer technology, including generative AI, is harder to predict because it comes back to how quickly "the pack" of carriers can  evolve how they manage the profit and loss across the portfolios. 

Paul Carroll 

What is an example of how insurers can improve their capital allocation? 

Katie Klutts Wysor 

The fundamental approach is to look at your underwriting returns against the capital you’re deploying to the business, map that out, compare outcomes, and decide where you want to grow and where you may need to pull back. Improvements in technology may allow carriers to do that analysis more frequently. Instead of doing it once a year as part of strategic planning, you could be looking at a refreshed view every month using more current data. 

Many carriers may be able to move from annual reviews toward monthly or weekly review cycles, depending on how they make decisions. They may also be able to do the analysis in a more automated way and make decisions more intentionally on micro-segments of the business (by geography, class, line, etc.) that would have been too time-consuming to identify and react to previously. 

Maybe a competitor enters the restaurant space aggressively and undercuts on price. You may decide not to follow them down that path because you believe the pricing is unattractive, so you slow growth in restaurants. 

Or take the opposite scenario: A trend affects restaurants and causes the market to become more cautious. You may conclude that the market reaction has gone too far and decide if this is the right time to pursue restaurant business. 

Today's leadership reviews may only look at class code-level data monthly or quarterly, and at frequency and severity trends in a backward-looking way. But if you can automate how, you assess that information at a portfolio level, then you can decide whether to lean in or lean out of a class like restaurants more quickly.  

Paul Carroll 

What about the execution side of this—how do insurers actually act on these faster insights once they’ve identified an opportunity? 

Katie Klutts Wysor 

The second half of the equation is exactly that. Say you’ve been able to automate and generate more timely underwriting data, so you can make portfolio decisions weekly or monthly instead of quarterly or annually. That’s a meaningful shift. The next step is execution. 

Say you decided to lean into restaurants. You want the market to know. You want your agents and distribution partners to know you’re interested in that business, particularly if another carrier has started to pull back or take rates. That’s the business you want to enter the pipeline. 

Then you want to set up your underwriting process so you can pivot quickly. Maybe you were not prioritizing that business to get it to an underwriter’s desk and streamline escalation paths to support faster turnaround. 

Of course, once submissions are flowing in, and the process is in place to evaluate and price the business you want competitively, you also need the proper governance and controls, so you don’t end up writing risks that fall outside appetite. 

The big difference this year versus a year ago is the ability to put agentic AI workflows in place and that support faster transaction-level decisions. Humans are still in the loop, but they are not necessarily slowing down the process in the same way they did when carriers relied more heavily on manual referral and escalation processes to respond to market changes. The next frontier I expect to see in the coming year is using agentic AI workflows to help improve portfolio-level decisions.

Paul Carroll 

Would you talk a bit about some of the process efficiencies from generative AI as underwriters make their decisions? While those efficiencies aren’t as strategic as the portfolio-level decisions you’ve described, they still seem substantial.  

Katie Klutts Wysor 

Underwriters face a series of yes, no, and maybe decisions, and much of the friction sits in the maybes. You can automate obvious yes-or-no decisions. The maybes are the gray area where you bring in a person. 

Over time, we may be able to bring in a person less often because of agentic AI and other decision-support tools, while maintaining appropriate human oversight. 

Some maybes exist simply because a piece of information is missing. A file gets routed to an underwriter to obtain one additional data point. Once that data point is available, a rule can be applied, and the case can become a yes or a no. In many cases, that is increasingly solvable today.  

You should identify those cases in your portfolio, then use AI to obtain the data point and apply the rule. 

There are also maybes that are more judgment-based, where you’ve created a manual review because you want someone to look at it who has seen this kind of case many times before. Maybe they’ve seen a six-figure loss in a similar situation, so you ask, “Would you write this again knowing what you know now?” 

Agentic AI workflows can help by bringing more context to the situation and supporting more informed underwriting judgment.  

Paul Carroll 

Based on what you’re seeing, how much are underwriters working with brokers and clients to provide guidance on risk reduction—essentially telling them, “you’re getting dinged for this, why don’t you fix it to reduce your risk?” 

Katie Klutts Wysor 

Right now, it’s predominantly brokers and distribution partners that are providing that first line of risk management advice. But there’s also a meaningful role for underwriters and carriers. 

The concept is there, the question is how consistently it can be translated into actionable guidance. 

Paul Carroll 

What final advice would you offer to readers? 

Katie Klutts Wysor 

What my clients care about is taking some of the bigger ideas and translating them into what to do right now and how to respond in practical terms. So, what I’d leave readers with is this: Keep thinking about the art of the possible but also focus on what you can implement now to strengthen performance this year, and start bringing those two together. 

Understand what technology, data, and AI capabilities are available to you. But more importantly, identify which ones you can deploy quickly while you continue building toward the more complex architecture and data challenges you should address over time. 

Paul Carroll 

So you can create a cycle: make targeted investments that create near-term efficiencies, then use those gains to support the next wave of investment. 

Katie Klutts Wysor 

Exactly. Don’t spend three years trying to build the perfect solution. There’s a lot you can do right now. Deploy something practical that can create value, then use those gains to support larger investments over time. 

Paul Carroll 

Thanks, Katie. 

About Katie Klutts Wysor

Katie Klutts Wysor is a Principal with PwC who advises insurance leaders on strategy, growth, and transformation. She focuses on analyzing evolving market dynamics to shape perspectives on the future of insurance and translating those insights into practical, outcome-driven growth strategies and transformation programs for carriers and brokers/distributors.

Here's a thought experiment for you: What if Elon Musk ran an insurance company?

Just imagine how regulators would react to his brash, visionary ideas wrapped in disdain for tradition and a belief that rules don't apply to him. 

But what if you could bottle the good parts of his iconoclasm and leave out the parts that would scare policyholders about the reliability of their insurance carriers? A former president of Tesla just published a book that might allow for that. It describes the five-part "algorithm" that he and Musk used to manage the company during a transformative stretch in the mid-2010s. 

I don't think insurers should go full force, a la Musk's "hardcore" mode--you could wind up with an embarrassment like DOGE and never recover--but his algorithm does offer a playbook for radical simplicity and for what I think is the right way to approach automation. 

Jon McNeill, author of "The Algorithm: The Hypergrowth Formula That Transformed Tesla, Lululemon, General Motors, and SpaceX," says the method has five steps:

• Question every requirement.
• Delete every possible step in the process.
• Simplify and optimize.
• Accelerate cycle time.
• Automate.

Question Every Requirement

McNeill writes about how Tesla, for instance, questioned China's requirement that it own a piece of any company operating in the country and eventually negotiated a deal that let Tesla own 100% of its Chinese subsidiary. He also writes about deciding that cars didn't need to be assembled out of so many parts, even though they had been since the days of Henry Ford. Instead, Tesla began experimenting with casting bigger and bigger pieces of the car and eventually succeeding, greatly reducing the need for assembly.

For insurers, though, I'm thinking the real benefit would come in more modest ways that track more closely with an anecdote McNeill told in a podcast with the Wall Street Journal. He talked about how much trouble Tesla had designing and manufacturing a part that was supposed to sit between a battery and the chassis. The problem became so important that Musk got personally involved and haunted the factory for weeks. Eventually, Musk and McNeill asked if the part was really necessary, and the battery people told them it had been mandated by the folks responsible for damping noise. When McNeill went to them, he was told that, no, the battery folks had mandated the part to minimize danger in the case of a battery fire. McNeill decided to track down the engineer who had signed the order requiring the part -- and learned he couldn't reach the person because he was a summer intern who no longer was at Tesla.

Insurers already question what they believe to be undue regulation, but I think they could benefit more broadly from asking employees across the business to question everything they're told to do, whether by someone inside the organization or outside it. Even if you just do this as a one- or two-month exercise, I'd bet you'll find you're doing lots of things just because they've always been done that way, not because they deliver any value.

Delete Every Possible Step in the Process

At Tesla, McNeill said in the podcast, he deleted several steps, and Musk asked whether he'd broken the process as a result and received some severe pushback. When McNeill said he hadn't, Musk told him he hadn't gone far enough. He needed to keep pushing until he not only got close to the bone but cut into the bone -- at which point, he should back off and find a sustainable approach.

McNeill said the rule of thumb was to only deliver what the customer directly paid for: the car. Customers didn't pay directly for manuals, for documentation, and so on, so Tesla would spend as little effort as possible in those areas.

Again, I don't think that approach would survive at an insurance company. Cut-until-you-break-something can happen in a manufacturing process, behind the scenes, but it didn't even work at the Department of Government Efficiency (DOGE), which Musk ran in the early days of the second Trump administration. Even with the slash-and-burn ethos of Trump 2.0 a year ago, Musk cut too deeply and caused problems both for those receiving government services and for Trump. 

Still, insurers can suffer from a sort of data and process bloat. Given the industry's abundance of caution, it's easy to ask for more questions, to gather more data, and to require another guardrail in the process. Life insurers have shown that it's possible to do the same with less, given the success of fluidless underwriting, and other lines could surely scale back some requirements, too -- becoming more efficient while making customers happier.

Simplify, Accelerate, and Automate

I'm combining the last three parts of the Tesla algorithm because, at least for insurers, they all fit under one mandate: "Automate last."

McNeill said Tesla learned the value of these three steps when it was having so much trouble manufacturing the Model 3 that it was running out of cash and was in danger of bankruptcy. The company stopped running its highly automated manufacturing line, set up a big tent outside the factory and started making the cars by hand. Once management figured out the best process, it began speeding up. Only once they saw that they could run the process at speed did they start bringing in the machines that would automate it -- scrapping the entire production line that they'd set up before fully understanding what was needed.

"Automate last" fits with the approach the computer industry has taken for decades: Once a manual process is fully mapped out, it can move into software and then, when you're sure you have everything nailed down, you can hard-wire the work by moving it into the silicon. 

That approach makes sense for insurers, too. When you see the possibilities of AI, for instance, you should map out a potential new process, implement it manually, speed it up -- and only then let the machines take over.

There are plenty of things about Musk's approach to business that I wouldn't recommend. For more than a decade now, I've been mocking his annual claims that he'll have millions of Teslas functioning as robotaxis, that he's going to colonize Mars (we won't even land someone on Mars in his lifetime), that he's about to unleash an army of humanoid robots, and so on. Those of us without his massive wealth would lose all credibility overnight if we pushed a similar sort of sci-fi dream. Insurance, as an industry built on trust, can't afford anything close to the wild claims that Musk makes routinely.

But I do think it's worth giving his algorithm serious consideration because it can reduce complexity and lead to effective automation. If nothing else, reading about the bold moves at Tesla might inspire some new thinking and resolve in the insurance industry. 

Cheers,

Paul

P.S. "The Algorithm" reminds me of one of my favorite geek jokes:

Q. How do we know that Al Gore actually invented the internet?

A. It runs on Al-Gore-ithms. 

Digital insurance agency platforms are the cornerstone of the insurance ecosystem. With digital first consumers demanding seamless experiences, agencies and carriers need technology that delivers speed, insight, and engagement. Zywave received the highest rank of all vendors in the strategy category, and maximum possible scores in the innovation, vision, and roadmap criteria in this report. 

The report provides a comprehensive evaluation of insurance technology platforms, analyzing current offering and strategy. See how Forrester evaluated 10 top platforms, and why Zywave earned the highest score in the strategy category and the highest possible scores in the criteria of vision and innovation. 

The Forrester Wave™ report noted:

• "Zywave’s vision is to facilitate the growth of insurance distributors by becoming an integrated, open API software suite powered by agentic AI".
• “Its impressive roadmap and innovation aim to bolster quoting and carrier connectivity as well as use AI-powered agents to automate workflows.”
• “Zywave’s SaaS platform offers agencies a robust marketing toolkit, quoting and proposal tools, and product analysis and comparison for personal, commercial, and benefit lines — aided by a highly intuitive UI and extensible platform architecture.” 

Access the Report

Sponsored by ITL Partner: Zywave