At a time when highway deaths in the U.S. are finally in a sustained decline again, after a surge related to COVID recklessness and distracted driving, car makers seem to be set on a dangerous path. Wanting to show off their AI chops, they are rolling out and emphasizing autonomy-lite capabilities that are convenient in the best of all possible worlds but that could well increase accidents and fatalities.

Basically, many car companies seem to be heading in the Tesla direction, bragging about their cars' autonomous capabilities while telling drivers they need to have their hands on the wheel and be constantly alert — a combination that just doesn't work. Once you tell people their cars can do the driving, they let their cars do the driving, and the sort of autonomy that exists this side of Waymo and a few other providers of fully autonomous vehicles simply isn't reliable enough yet. 

I've become what's called a 410-er, and I think insurers should be, too.

410 — unrelated to 420, a coinage celebrating cannabis that seems to amuse Elon Musk — refers to three of the six levels of autonomy: levels 4, 1, and 0, which I believe are the safe levels at the moment. Car makers, meanwhile, are focusing on providing levels 2 and 3 to the mass market, levels that I believe are fraught with danger. 

Level 0, as you can imagine, involves no autonomy. Level 1 is referred to as driver-assist — your car keeps you in the center of the lane if you're starting to drift and lets your cruise control maintain a safe distance from cars around you. Both provide little enough assistance that the driver stays fully engaged.

Level 4 is fully autonomous within a certain area and under certain conditions. As long as you stay within, say, a well-mapped area in good weather in daytime, you could be asleep in the back of the car. This level is safe, too — once the technology is proven, of course.

Combine those three levels, all of them helpful without overpromising to the driver, and you get to 410. You could also add level 5, which is fully autonomous anywhere at any time, but we're not there yet, even with Waymo, et al. I think we'll get to a 5410 paradigm, just not yet.

The problem is that car companies want to brag about levels 2 and 3, which can lull drivers into danger. Level 3 is especially seductive, because it promises "conditional driving automation": The system handles all aspects of driving under certain conditions, but the driver must be available and able to take over.

Google, Waymo's parent, tried a Level 3 approach years ago and quickly gave up. The technology was so good that Google employees who volunteered as test subjects soon zoned out and started checking their phones, playing video games or whatever — but the car wasn't totally reliable. And when the car told the driver to reengage, many seconds passed before they could stop whatever they were doing, size up the issue and act. When you're traveling at 75mph, 10 seconds equals almost 400 yards of distance traveled while the driver is taking control of a situation. 

Tesla is finding out how dangerous Level 3 autonomy can be. It's faced numerous lawsuits over fatal accidents that occurred while drivers had engaged what Tesla calls Full Self-Driving (FSD) but which is really a Level 3 system. Tesla has mostly escaped liability by arguing that it warned drivers repeatedly that they were responsible for the car's actions, but it did recently lose a $240 million wrongful death judgment that should serve as a warning to Tesla, to other auto makers offering Level 3 systems and to auto insurers.

Tesla, seemingly unchastened, recently told its car owners that if they felt drowsy they should engage FSD — even though a drowsy driver would take even longer to reengage if told to by the AI driving the car. Other car companies are promising what are generally referred to as Level 2+ systems, and Telemetry says more than half of new cars will be equipped with such "hands free" systems by 2028. 

Car companies will surely promote these systems. Everybody wants to be seen as being at the cutting-edge of technology, and AI can be a real selling point these days. Autonomy is exciting.

But I hope cooler heads will prevail. 

"Customers really love these hands-free systems, especially on longer drives, but God is in the details, and... the worst of these systems may result in preventable injury or worse," Telemetry says.

Traffic deaths in the U.S. had been declining steadily for decades, reaching a low of roughly 35,400 in 2014, but increased steadily as smartphones tempted drivers with distractions and then surged as COVID somehow made drivers more reckless. Deaths peaked at approximately 47,000 in 2021. They dropped slowly from there, falling back to 44,700 last year, and declined a gratifying 8.2% in the first half of this year. 

Let's keep the progress going and push back on efforts to promote Level 2+ autonomy as anything more than an occasional convenience. Level 2+ and Level 3 are wildly impressive technology — impressive enough to be truly dangerous.

410, 410, 410....

Cheers,

Paul

P.S. What happens when a driverless car commits a traffic violation? Who gets the ticket?

That was the issue facing police officers when a Waymo vehicle made an illegal U-turn right in front of them in San Bruno, CA, just south of San Francisco, a week ago. They pulled the car over — Waymo cars pull off the side of the road when a police car turns on its emergency lights — but when they approached the car they found... no one in it.  

Under California law, tickets can't be issued to driverless vehicles until next summer — “Our citation books don’t have a box for "robot,” the San Bruno Police Department noted. Even once tickets can be issued, they carry no penalty. 

Expect the law to change — and expect further oddities as autonomous vehicles become more common. 

Paul Carroll

The insurance industry has been operating under projections that approximately 400,000 professionals will retire soon, meaning that an awful lot of knowledge as well as an awful lot of people will be walking out the door. What can we do?

Sharon Emek

The insurance industry is challenged because we don't have enough young people still coming into the industry. There's now become sort of a bias by companies against young people. I hear it from my clients all the time. Young people don't have the same work ethic. They're nine to five. They don't put in the extra effort. Within a year, they want a promotion and a raise. Then they move on.

On the other hand, with the older generation, first of all, we're living longer. People have to work beyond 65. So retirement is a conversation the industry needs to have.

If somebody is turning 65 and they say "I'm done with all the stress," it's not that they really retire and do nothing. A lot of people retire and end up being a greeter at Walmart. I see them at retail stores, see them at grocery stores because they still want to work. They just don't want the kind of stress and travel time required for a full-time job. They want to be near home.

So the industry needs to think about this looming huge Boomer population that is going to retire—they don't really want to retire fully from work.

Some will because they have medical issues, but many of them want to do something less challenging while still being productive. Because people cost more medically as they age, the industry sort of lets them go instead, but we need to figure out ways to keep them in a different capacity, maybe with lower salary.

Young people can teach older workers about technology while gaining insights from them. One group is technologically savvy while the other has all that institutional knowledge, wisdom, and work ethic.

Paul Carroll

Where do you think holding on to older talent could have the most benefit?

Sharon Emek

Brokers are probably being affected more than any other sector, because a lot of the people in brokers' offices are Boomers—especially women who never went to college and went to work when they were 18 or 19 years old at a local agency.

And brokers don't have HR people helping with all this stuff.

Paul Carroll

The move these days is to get people back in the office. How does that trend fit with what you’re describing?

Sharon Emek

Well, I've been doing this for 15 years and have placed thousands of people at brokers, carriers, reinsurers, MGAs, MGUs. While a young person may need to be in an office—they're on a career path, they need relationships, they need to be mentored—a highly experienced older person does not. And everyone is familiar enough with the technology these days to work remotely.

Older workers have an amazing work ethic. They're happy to feel that they're still engaged. They're happy that you're allowing them to have a flexible life. They can work from home and save all that commuting time. They’re no longer on a career path, so they’re not stressing themselves or my clients about promotions and raises and all that stuff.

Instead of offshoring, having a retiree is such a better option because they could do double the work because they know what they're doing. And they're less expensive because they're retired; they know they're not going to get the same money as before.

I call them the "pre-tired." They’re in their "pre-tirement."

Paul Carroll

I love the concept in theory but am curious about the details. How do you structure employment for your workers?

Sharon Emek

You can't do part-time less than 20 hours a week. It's not enough time. So the minimum is typically 20 hours. Nobody wants to just work 10 hours.

60% of our people are working roughly 25 hours, and 40% work 35 to 40 hours a week. Some people feel like 35 to 40 hours is nothing compared with what they were doing. They say, "Oh, I live outside of Atlanta and was commuting an hour and a half each way in the car. That's three hours a day."

We pay them by the hour, but this isn’t like a consulting gig. We tell clients they need to treat our people just like they would any full-time employees. Our people have to become part of the team to do the work they need to do.

We give people the option of having benefits if they need them. They're not independent contractors.

Paul Carroll

How long do people keep working?

Sharon Emek

Our average age right now is 64, but I have people in their eighties who are still working, as well as people who are in their late sixties or seventies. So it depends.

One of my managers is in her mid-eighties. She's terrific, and she's not ready to retire yet. She still goes to rock concerts. Another woman who's almost 80 is a roller skate champion and still competes.

Paul Carroll

My mother was still playing tennis three times a week at age 87 and, at 92, was getting answers that all the contestants in a round of the Jeopardy! Champions tournament didn’t know, so I hear you.

Any final thoughts?

Sharon Emek

The industry is still going to need young people, and we need to have a multigenerational workforce. Many companies are having some challenging experiences with young people, but they have to move on from that and figure out how to attract the right young people with the right culture.

Getting them mentored the right way is critical because companies need to bring them up through the ranks. Eventually, our retiring population will fully retire, and the next generation isn't as big a cohort. We simply don't have enough young people coming into the workforce.

And AI is not going to replace everything. There will always be roles that require human judgment and expertise.

But the “pre-tired” can help a lot as part of the mix.

Paul Carroll

This is great, Sharon. Thanks so much.

About Sharon Emek

Sharon Emek, Ph.D., CIC, is chairman and CEO at WAHVE, a talent agency founded with a vision to address the approaching Baby Boomer retirement and growing need for experienced talent in the insurance industry. She is a frequent speaker on the challenges that employers and “vintage” professionals are facing today.

As an advocate for innovation who has been chronicling the power of digitization for nearly four decades, I sometimes get impatient with the pace of change in insurance. But I take comfort from the latest global survey by our friends and colleagues at the International Insurance Society. 

In particular, I'm encouraged that 66% of those surveyed said artificial intelligence is their top technology priority, up from 17% just four years ago. I'd say the figure should be well north of 90%, but the latest results still show a remarkable increase in awareness of the possibilities presented by AI.

The emphasis on AI dovetails with other survey results on issues such as operational efficiency and talent, where AI can solve a lot of problems. More broadly, the survey supports my belief — or, at least, hope — that insurance can be a leader in adopting generative AI. Insurance obviously has major regulatory issues to deal with and owes stability to its customers, but, despite those constraints, insurance is a perfect fit for AI. At its core, it is a purely digital business, based on the masses of data that AI is perfectly designed to collect, analyze and act on. 

The tendency is to look at technology revolutions as Big Bangs. For instance, we all learned in high school that James Watt's steam engine produced the Industrial Revolution. True enough, but what we (at least, I) missed is that Watt unveiled his key improvement in 1776 while the economic gains didn't start showing up until the 1820s, some 45 years later. A lot had to happen in the interim, including technology improvements, legal innovations and the invention of the factory. The same will be true with AI: We aren't going straight from invention to a change in life as we know it. But we've learned a lot over the past two-plus centuries and won't need 45 years this time. A lot of the reason is the sort of focus that the IIS survey of insurance executives reflects. So, kudos.

Let's look at some of the other interesting results.

The other super-high priority in the IIS survey was inflation. It was the top economic priority for the fourth consecutive year, with 63% of respondents listing it among their top three issues.  

Concerns over an aging workforce almost doubled over the past year, as the industry faces a much-publicized wave of retirements. I think AI will play a major role in addressing the issue by capturing the knowledge that in years past would have simply walked out the door, by facilitating part-time work by those who would have otherwise retired, and even by attracting some younger, tech-hungry talent. 

The survey found: "Beyond AI, the broader emphasis on technological advancement has also experienced significant growth. Forty-one percent of respondents now view technological advancement as a top social and environmental priority, continuing its rise from only 12% in 2021." AI will be key on this issue, too, as it will for operational efficiencies, which for the second year in a row top the list of operational priorities. 

I encourage you to not only read the full version of the report available to the public but to also consider attending the IIS's Global Insurance Forum, which will explore the key issues in the report. This year, it's being held Oct. 26 and 27 near Zurich.

At the GIF, we at ITL will join our colleagues at the IIS and our sponsor, Lloyd's, in handing out the Global Innovation Awards. The finalists are:

Life/Health/Retirement Innovator of the Year

• Irish Life: AI Reasoning Assistant for Claims
  A multimodal AI solution that automates document processing, accelerates claims assessments, and enhances accuracy, efficiency, and customer satisfaction during critical moments.
   
• RGA: MedScreen+: AI Solution That Strengthens Digital Underwriting
  An AI- and OCR-powered underwriting solution that digitizes and standardizes health records, enabling faster, more accurate, and customer-friendly life insurance decisions while bridging the gap between centuries-old practices and the modern era.

Property/Casualty Innovator of the Year

• Gallagher: Global Data & Technology Capabilities
  A data and analytics platform that simplifies complex industry insights into rates, losses, and limits, empowering clients, brokers, and partners with timely intelligence to make confident, informed decisions.
   
• Tawuniya: End-to-End Motor Claims Transformation
  A fully digital motor claims platform that streamlines the entire journey from accident reporting to settlement, delivering faster processing, greater transparency, and higher customer satisfaction for over 1 million users annually.

Insurtech Predict & Prevent Innovator of the Year

• AXA: Transforming Insurance at a Time of Polycrisis 

  A one-stop, AI-powered ecosystem that unites real-time catastrophe intelligence, cyber defense, risk tools, and training to help organizations predict, prevent, and respond to today’s connected crises. 

• Tawuniya:  Vitality & Drive: A Lifestyle-Based Ecosystem
  Leveraging gamified insurance, health apps, and telematics to engage 500,000-plus users in safer driving and healthier living, cutting accidents by over 25% and reducing claims while building resilience.

They're all deserving. I'll be fascinated to see who wins.

Cheers,

Paul

The reason everyone seems to have a scam story these days isn't due to an increase in reporting; it's because scams have become a near-universal experience. As underscored by a significant increase in the past five years of both reported incidents to the FBI and mainstream media coverage, scams are more frequent, more costly, and more difficult to discern than ever before.

The State of Scams USA 2025 report, conducted by the Global Anti-Scam Alliance (GASA) and sponsored by Iris Powered by Generali, showed that 77% of American consumers encounter scams on a daily basis, with over 70% indicating they had been scammed in the last 12 months. The report also found that one in five Americans lost money to a scam in the same time frame, with an average of over a thousand dollars lost per person and over $64 billion stolen in total.

With scams and fraud on the rise, consumers have turned to institutions and communication platforms for help. Almost three in four (74%) respondents who had experienced a scam reported it to an authority or company for assistance. That is consistent with the findings of the Iris 2025 Identity and Cybersecurity Concerns survey (“ICC”) conducted in April, which found that most consumers reach out directly to companies that have been part of a data breach. However, over half the time, nothing is done – with 57% of reported incidents having no discernible action taken. Even worse, of the 82% of U.S. consumers who reported scams to payment services or financial institutions, less than half (44%) were able to partially recover money in the end, and 38% received nothing back at all.

This gap between consumer action and institutional response feeds a dangerous sense of futility: if reporting scams doesn’t lead to meaningful outcomes, why report at all? This mentality can allow scammers to gain the upper hand. Americans need an ally in the fight to defend themselves against scammers, and they’re expecting financial and communications platforms to step up.

Digital Platforms Top the List for Scammer Channels

By and large, scammers are targeting consumers digitally. Most consumers reported encountering scams via SMS messenger, followed closely by emails and phone calls. Americans reported that 82% of scam attempts occurred on platforms with direct messaging capabilities, including social media, instant messengers, online marketplaces, and even digital ads.

In terms of platforms, Gmail ranked highest in reported instances at 45%, followed closely by Facebook at 41%. TikTok, Snapchat, and X (Twitter) ranked notably lower, but consumers tended to take the longest to recognize that they were being scammed on those platforms.

Consumers are offered little recourse through the platforms themselves. Recent reports indicate that large social or digital communications platforms can take weeks to act when scams are reported. This lack of urgency contributes to an erosion of consumer trust.

Most Lose Money Through Debit Cards and PayPal

Debit cards were the most common method used by scammers, accounting for 30% of reported losses to fraud, followed by PayPal at 25% and credit card payments at 23%. When fraud occurred, most consumers discovered it themselves: 66% discovered it on their own, while only 14% were alerted by their bank or financial services provider.

Americans who were affected overwhelmingly reported the fraud to banks or payment services, with 82% reaching out for support once they realized they had been scammed. But again, this ultimately had mediocre returns for consumers. Likewise, according to Iris' ICC survey, 46% of Americans say their first call would be to their bank after receiving a notification of a data breach, making it their top choice.

These patterns make it clear that consumers view banks and payment platforms as their frontline defense. But when response and recovery prove insufficient, trust is eroded.

Consumers Blame Commercial Organizations – But U.S. Laws Don't

Consumer protection authorities are contacted only 12% of the time, compared with banks or payments services at 25%.

While one in three Americans believe that commercial organizations should be responsible for protecting consumers, U.S. laws and regulations don't agree. For instance, authorized user payments, such as those through platforms like Zelle or Venmo, have no legal requirement for banks to reimburse customers. Additionally, newer scams like imposter scams or AI/deepfake scams are not covered by older FTC regulations and U.S. laws, leading to confusion and denials from banks to reimburse victims.

Where Third-Party Identity Protection Services Fill the Gap

Consumers need stronger support in the fight to protect their identities – and wallets – online. Yet major commercial organizations and digital communications platforms are failing to provide adequate protection.

Iris' ICC survey found that most consumers want a comprehensive, all-in-one solution and are willing to pay for it. While just three in 10 Americans indicated they follow all recommended data protection practices, close to eight in 10 said they would likely use identity protection features if they were integrated into an app they already use, with banks and credit card providers being one of their top picks to purchase from.

Third-party identity protection solutions help close critical gaps. Tools that monitor for compromised data on the dark web, help to spot scams, and offer expert fraud recovery services aren't new but are increasingly sought after. These services not only accelerate resolution by managing outreach to banks and authorities but also help ease the emotional toll of falling victim to scams – a cost that's often overlooked. Additionally, they often take critical steps on the consumer's behalf to prevent further damage.

Consumers want accountability from today's institutions, but they also want protection and peace of mind. There are bills currently under review by the U.S. Congress – like the Protecting Consumers from Payment Scams Act – that are aimed at addressing accountability gaps with banks and payment providers. But responsible businesses shouldn't wait for laws to catch up with the rising threats; they should show up today for their customers and be the ally they need by offering protection.

Not only is it the right thing to do – but it's a powerful investment in customer loyalty and trust.

Walk through any insurtech conference today and you'll hear "agentic AI" mentioned at every turn. Every vendor booth promises autonomous systems that can think, act, and learn. But when you examine these solutions more closely, many turn out to be large language model (LLM) implementations with intelligent automation added. These are valuable advances, certainly, but not the autonomous agents they claim to be.

This terminology inflation creates a fundamental problem. When insurance executives hear every vendor claiming to have "agentic AI," the market becomes so cluttered that companies that invest in building these new capabilities get lost among rebranded automations.

Defining Agentic AI

Part of the problem is that there's no uniform definition of what makes AI "agentic." Different experts emphasize different aspects: Some focus on autonomous decision-making, others on learning capabilities, and still others on goal-directed behavior. But while the exact boundaries remain fuzzy, we can certainly identify what agentic AI is not.

It's not just a chatbot with a fancy prompt. It's not a series of LLMs strung together with if-then logic. And it's definitely not traditional automation rebranded as AI.

One proposal is that true agency requires at least three core capabilities:

1. Tool usage - the ability to navigate and interact with different systems
2. Memory - maintaining context and learning from past interactions
3. Real-time adaptation - adjusting approach based on results when something unexpected happens

The coding assistants like Cursor and Claude Code offer a useful reference point. These tools represent the current state of the art in AI, and most industry observers would comfortably label them as "agentic." If these are our benchmarks for genuine agency, the gap with most "agentic AI" solutions in insurance becomes clear.

This distinction matters because it reveals a spectrum. On one end, you have simple automation following predetermined paths. On the other end, you have fully autonomous systems that set their own objectives and continuously evolve. Most of what's being called "agentic" in insurance today sits firmly at the automation end, despite the marketing claims.

Current Market Examples

The evidence for this is everywhere. Take one major claims administrator's recent announcement of their "agentic AI" solution. Dig deeper, and it's a bundle of voice bots, intelligent document processing, and some alerting.

Another prominent vendor markets six different "AI agents" as part of their agentic platform. Remove the marketing speak, and you find a data layer with LLMs for document routing, a chatbot that accesses internal data, and template generation with compliance checks. These are often solid implementations that deliver real value, but they're a far cry from being truly "agentic."

The Market Distortion

The pressure to appear cutting-edge creates an arms race of terminology. When every vendor feels compelled to claim "agentic AI" to stay competitive, an insurtech that invested heavily in genuine foundations—tool usage, memory, and real-time adaptation—gets lumped in with one that simply added "agent" to their chatbot's name.

This creates an unfortunate dynamic. Insurance executives face an impossible task: evaluating solutions when every vendor uses the same terminology for vastly different capabilities. Even sophisticated buyers struggle to identify which systems will grow into true agency as AI matures versus those that are essentially dead ends with fancy names.

When implementations fall short of vendor promises, it naturally reinforces skepticism about AI investments. The insurtechs building for the future get caught in this backlash, making meaningful transformation even more challenging. Everyone loses: Buyers miss out on genuinely transformative technology, real innovators struggle to differentiate themselves, and the industry's digital evolution slows to a crawl.

The Path Forward

The insurance industry doesn't need to claim false sophistication. Current AI applications can provide tremendous value. Intelligent document processing saves countless hours. Well-designed chatbots genuinely improve customer experience. Predictive analytics enhances decision-making in measurable ways. These are powerful tools that augment human capabilities. The industry benefits when we accurately describe what these tools accomplish and match them to appropriate use cases.

For those evaluating solutions, start with a more fundamental question: Do you actually need agentic AI? If your goal is to reduce document processing time by 80%, intelligent automation might be exactly what you need. If you want to improve first-call resolution rates, a well-designed LLM-powered chatbot could be the perfect solution. These aren't agentic, but they solve real problems with proven technology available today.

Reserve the search for true agentic capabilities for problems that actually require them: complex claims that need dynamic investigation across multiple systems, underwriting decisions that must adapt to unique scenarios in real-time, or fraud detection that needs to evolve its approach as schemes change. For these use cases, ask the hard questions: Can this system actually use tools to solve problems? Does it maintain context across interactions? Can it adapt when things go wrong?

As agentic AI capabilities mature, they will transform how we handle claims, assess risk, and serve customers. But we'll only realize that potential if we're honest about where we are today and deliberate about where we're investing for tomorrow.

As buyers and builders, we all have a role in maintaining clarity about what AI can actually accomplish. This ensures that success goes to companies building on real capabilities rather than marketing claims, while preserving confidence in AI's genuine transformative potential.

In P&C insurance, delay can cost millions. Yet critical insights are often stuck in spreadsheets, buried in SQL queries, or waiting in the inbox of an overloaded data team. Underwriters, actuaries, analysts, and executives depend on timely answers, but bottlenecks in access to data slow decisions, erode agility, and hold back growth.

From Queries to Conversations: Redefining P&C Analytics

AI-powered conversational analytics, exemplified by solutions like Snowflake Cortex Analyst, represents a significant leap forward, transforming how P&C insurance companies access and use their structured data. By offering a natural language interface, these solutions empower business users to directly interact with data, providing instant answers without the need to write complex SQL queries.

Whether you are an analyst buried in SQL requests, an actuary tracking loss severity trends, or an executive making growth decisions, conversational analytics ensures you can get trusted answers instantly.

Six Strategic Benefits of Conversational Analytics for P&C Leaders

1. Empowering P&C Professionals

Conversational business intelligence (BI) tools directly enable underwriters, claims adjusters, and actuaries to obtain immediate answers to their business questions, significantly reducing their dependency on overloaded data teams.

Conversational Analytics in Action:

An actuary reviewing quarterly loss ratios notices a spike in severity but cannot pinpoint the cause without a custom report. Traditionally, this would mean submitting a request to the data team and waiting days for an answer. With conversational analytics, they simply ask: "Show me loss severity by line of business over the past 12 months." Within seconds, they have the insights needed to recommend pricing adjustments before the next underwriting cycle.

2. Targeted Executive and Operational Insights

Executives, marketing, and actuarial teams are ideal beneficiaries of ad hoc querying and discovery. They can quickly explore data, understand trends, and inform decisions without waiting for pristine, pre-defined reports.

Conversational Analytics in Action:

During a monthly leadership review, executives see an unexpected drop in renewal conversions. Normally, they would debate whether the numbers are accurate, request more data, and revisit the topic weeks later. Conversational analytics eliminates the wait. An executive can drill into renewal performance by product line, geography, or agent channel instantly, making it possible to align strategy and respond in real time, not after the quarter has closed.

When a catastrophic weather event strikes, claims volumes can spike overnight. Executives need immediate insight into incurred losses, reserve impacts, and cashflow forecasts to make operational and customer decisions quickly. With conversational analytics, leaders can instantly model claims by region, line of business, or severity—turning what once took weeks of reporting into rapid, informed response.

3. Rapid Analysis of Critical P&C Metrics

The solution enables fast analysis of profitability, growth, and risk metrics:

• Profitability: Combined Ratio, Incurred Loss Ratio, Underwriting Expense Ratio (UWE)
• Loss Costs: Loss Frequency, Loss Severity, Loss and Reserve Development (LRD)
• Growth: New Business (NB) Binds, Renewal (Ren) Conversion Ratio, Written Premium (WP)
• Cashflow: Earned Premium (EP), Collected Premium

By tying these directly to strategic levers such as pricing, retention, and claims management, leaders can act faster and with confidence.

4. Enterprise-Grade Security and Governance

Solutions built within secure data platforms ensure that data, metadata, and prompts remain strictly within governance boundaries, with role-based access controls (RBAC) intact. Every query adheres to enterprise policies, protecting privacy and compliance.

5. Streamlined Data Access and Control

Integration within a unified platform creates easier, more controlled access to data. This reduces friction, enhances trust, and ensures analysts and executives are always working from the same single source of truth.

Conversational Analytics in Action

Analysts in many insurers spend more time fulfilling one-off requests than actually analyzing data. A marketing leader might ask for retention ratio by region, while underwriting wants conversion rates by channel, each requiring new queries and validation. With conversational analytics, those stakeholders can self-serve the basics. Analysts are freed from being "SQL order takers" and instead focus on deeper projects like identifying emerging risk drivers or developing predictive models.

6. Low-Investment AI Experimentation

For P&C companies already on platforms like Snowflake, conversational analytics offers a low-investment way to explore AI-driven insights. Leaders can experiment without major upfront commitments, building both organizational confidence and future readiness.

AI-powered conversational analytics is more than a technology upgrade. It is a strategic imperative. For actuaries, it means faster insights into loss ratios and risk drivers. For analysts, it means shifting from query fulfillment to strategic analysis. For executives, it means eliminating delays and making confident, timely decisions.

In a market where every day of delay can cost millions, the insurers who embrace conversational analytics now will lead. Those who do not risk being left behind.

Enterprise success depends not merely on being data-rich but on decision quality. As insurers strive to scale beyond pilots and deliver better outcomes, it is imperative to diagnose the gap and find an approach to bridging the divide. This article demonstrates how Decision Intelligence—combining decision memory, adaptive analytics, and explainable AI—codifies tacit knowledge into institutional capability, reducing variability and enabling consistent, transparent, and resilient performance in a volatile and dynamic environment.

Problem Statement: What?

In today's VUCA environment [volatility, uncertainty, complexity and ambiguity], insurers face a paradox. Despite abundant data from broker submissions, underwriting systems, claims, third-party sources, and significant investments in analytics, better insights do not always translate into better outcomes.

Consider a commercial property underwriting scenario, where two underwriters, given the same portfolio, guidelines, data and insights, often arrive at different decisions. This divergence stems from how each interprets insights, applies them in workflows, and navigates multiple decision paths. This tacit knowledge that drives these choices remains uncaptured, limiting institutional learning and consistent performance.

Bridging this gap requires a new paradigm, Decision Intelligence, to translate the art of decision making into a science that adapts to the environment.

Understanding the Gap in Decision-Making

Commercial property underwriting is inherently complex, involving numerous factors that influence risk exposure, insurability, pricing, and profitability. Most insurers have focused on improving underwriting cycle time, but with limited focus on consistency of decisions. Current systems accelerate processes but do not capture the rationale behind decisions or institutionalize the tacit knowledge. As a result, decision making remains highly dependent on individuals, creating variability in outcomes. For instance, underwriting a portfolio with higher premium (to cede the risk), or optimal premium with certain limits and exclusions, or seeking additional information from risk assessors or actuaries on assumptions, anomalies, etc.

How to bridge the Gap?

To close this gap, insurers need to move beyond process automation toward Decision Intelligence—a framework that embeds decision memory, simulations and explainability, knowledge graph powered by a decision agent. This digital capability connects data, insights, and decision logic, ensuring decisions are explainable, repeatable, and adaptable to changing market conditions. Below are some key components of Decision Intelligence.

Adaptive Analytics Agent

Discover hidden patterns/groups and key influencing features from the captured datasets using unsupervised learning and apply contrastive learning techniques to generate scenarios that will maximize the value and minimize risk. This decision assist agent provides its recommendations in natural language to the underwriter with explanations/rationale, for their review/feedback. The feedback loop is a critical linchpin for the agent to capture tacit knowledge – such as observations with respect to anomalies in roof condition scores and high claims related to roof repairs in that region with similar property characteristics.

Scenario Modeling and Explainability

Tools to simulate various scenarios (e.g.: open-source platforms such as Oasis Loss Modeling Framework, Fathom etc.) and determine annual average loss (AAL) for the portfolio. This risk assessment/modeling is governed by COPE framework (construction, occupancy, protection, and exposure).

It involves analysis of scenario values at location level, insurance-to-value (ITV) analysis to assess the coverage sought vs. cost to replace or repair insured property. ITV is driven by factors such as inflation, cost of replacement materials, labor shortages, etc. and helps to avoid underinsurance and coinsurance penalties. The underlying exposure (AAL) insights and pricing/premium governed by ITV analysis, helps to determine/forecast the profitability of the portfolio, based on cohorts and associated features that is driven by the principle of value maximization and risk minimization.

These in silico trials, augmented by ontology and powered by knowledge graph-driven decision memory and explainable AI, help underwriters to choose the decision paths. This capture of tacit knowledge and decision rationale will continuously evolve with integration to real-world environment (sensors, IoT, spatial imagery, geocoding etc.) and foresights, guiding underwriters in effective decision making.

Decision Memory

Building decision memory involves capturing the "why" and "what happened" in the workflows, including micro-decisions. This includes risk assessor notes, pricing notes and assumptions, property characteristics, perils considered, patterns/key influencing factors such as primary, secondary modifiers, anomalies between broker submission and risk assessment, risk mitigation strategies, exclusions, learnings/feedback loops from good and bad decisions. This would also involve digitizing the underwriting decisions by creating and integrating UI/prompts to capture and validate this additional information (human-in-the-loop) in the workflows.

Potential Benefits

By capturing rationale, simulating scenarios, and closing the loop with feedback, insurers reduce decision variance, improve price adequacy, and speed time-to-quote. This will result in better outcomes such as higher quote-to-bind on target segments, more written premium per underwriter, tighter loss and combined ratios, and stronger model governance—measured through KPIs such as decision-rationale coverage, outcome variability across underwriters with similar risk profiles, ITV accuracy, technical-price adherence, and audit-ready explainability with assumptions, trade-offs and overrides.

The Way forward

To achieve these objectives, organizations should prioritize the integration of advanced analytics and AI-driven decision support tools across underwriting, risk management, and data governance functions.

This includes establishing cross-functional teams to continuously refine models and decision frameworks based on real-world feedback, fostering a culture of transparency and learning, and leveraging technology to streamline workflows while upholding rigorous controls.

By aligning strategic goals with operational practices, insurers can unlock greater efficiency, consistency, and resilience in an evolving marketplace.

Cyber Awareness Month may occur only once a year, but cyber risks are ever-present for businesses, regardless of their size. With the frequency and sophistication of these attacks growing, and costing the U.K. economy an estimated £27 billion per year, it's vital that organizations understand the full cost of a cyber breakdown and how they can minimize this.

Among SMEs, ransomware attacks, data breaches and phishing incidents continue to rise, often leaving these businesses with fewer resources to respond and recover effectively. The average cost for businesses to remedy a cyberattack is estimated to be £21,000, an amount that can even bankrupt smaller companies.

U.K. businesses are facing a sobering reality: The actual cost of a cyberattack extends far beyond immediate recovery. While initial costs cover detection and response, hidden costs, such as lost business income, data restoration, regulatory fines, and reputational management, can linger.

Why Recovery Costs Don't Tell the Whole Story

For businesses, it's not a matter of if a cyberattack will hit you, but when. Just over four in 10 businesses (43%) reported experiencing a cyber breach or attack in the past 12 months, equating to approximately 612,000 businesses.

Reports suggest that 60% of small companies go out of business within six months of a cyberattack. For those who survive, they face major setbacks. For small businesses, a cyber breach or attack can set them back £65,000. However, this can be an underestimation of the full scale of the impact, as there is other fallout from cyber breaches, not limited to economic losses, but also reputational damage. The full cost of a cyber breach can involve challenges such as:

Business Interruption & Income Loss

The cyber landscape risk is changing, and increasingly, attackers aren't just looking to steal data but also to disrupt business.

A cyber breach or attack isn't an isolated incident. Once that happens, there are knock-on effects on businesses, from downtime that disrupts sales, issues with supply chains and eroded client trust. Many businesses will lose weeks of income, which can cripple small operations, especially in industries heavily reliant on online sales and client management.

For instance, if a business is targeted with ransomware and a demand for payment, accompanied by a threat that the company's data will not be restored unless payment is made, the business would be unable to conduct its day-to-day operations. Having a cyber policy that covers Direct and Dependent Cyber Business Interruption would be essential in this case to minimize losses.

M&S, for example, estimated that its cyberattack, which started in April 2025 and resulted in subsequent downtime, would cost them around £300 million in profit due to lost sales and increased operational costs from suspending online orders.

Data Recovery & System Restoration

Rebuilding technical processes, whether it involves systems, restoring backups, or investigating vulnerabilities, creates additional costs. Businesses may require specialist security experts to investigate and mitigate the loss. In the case of SMEs, they often don't have this expertise in-house.

Regulatory Compliance, Legal Fees & Penalties

Cyber breaches that result in a business's loss of personal or confidential information can lead to customer claims, breach of contract disputes, or regulatory fines under the Data Protection Act (GDPR).

Fines and legal fees under GDPR can push recovery costs even higher, particularly for SMEs that may lack in-house compliance expertise. The highest maximum an organization can be fined for this is £17.5 million or 4% of the total annual revenue in the previous financial year. Moreover, depending on the type of attack, there may be requirements to report to the Information Commissioner's Office (ICO).

Legal representation costs and external consultancy fees are high. However, with the right insurance policy in place, these costs can be covered.

Reputational Damage

The full cost of a cyberattack isn't always financial; it's often reputational, as well. If customers' data is stolen, it can affect future relationships, lead to customer churn, and ultimately affect the brand's value.

The Insurance Safety Net

Cyber insurance provides (indirect and direct) financial protection and access to expert legal and risk management support, enabling businesses to improve their operational resilience, defenses and adopt a proactive security approach.

Given the prevalence of cyber attacks, with AI making them more sophisticated, it's more critical than ever that businesses of all sizes invest in cyber insurance. It shouldn't be an afterthought; it needs to be a key priority for business resilience.

Encouragingly, many small businesses are taking note, with an increased uptake of cyber insurance from 49% in 2024 to 62% in 2025. There has also been an increase in security risk assessments and business continuity plans that address cybersecurity.

Coverage now extends far beyond simple data breaches. It can include ransomware payments, business interruption, legal fees, and even the cost of notifying affected customers.

Building Resilience Beyond Insurance

Insurance is a vital part of the cyber risk puzzle, but it's not a one-stop fix. The best insurance policies will be those that embed risk management into their policies, offering advice on how to appropriately train staff, create a risk management plan, implement multi-factor authentication, conduct regular audits, and more. Increasingly, the role of insurers is evolving to enable them to act as partners in prevention, not only by paying claims when things go wrong.

The insurance industry finds itself at a fascinating crossroads. While AI dominates board meetings across every sector, the claims space tells a more nuanced story: one of cautious optimism tempered by legitimate concerns about trust, accuracy, and regulatory compliance. A recent survey commissioned by our team at Wisedocs and conducted by PropertyCasualty360 reveals this paradox in detail, offering insights into how claims professionals view AI adoption and what it will take to gain trust across the industry.

The 2025 survey, "AI in Claims: The 4x Trust Effect of Human Oversight," polled claims professionals from PropertyCasualty360's audience, including adjusters, carrier-side claims managers, and third-party service providers. What emerged was a clear picture of an industry ready for technological transformation, but only under the right conditions.

Key Insight #1: The Trust Deficit

The survey's most striking finding centers on trust, or, rather, the lack thereof. Only 16% of respondents expressed medium or high trust in AI-generated outputs when used independently, with a mere 2% indicating high trust. This skepticism isn't born from technophobia but from practical concerns rooted in the high-stakes nature of claims work.

The primary barriers to AI adoption reveal why claims professionals remain cautious. Accuracy concerns topped the list at 54%, followed closely by compliance and regulatory risks at 49%, and integration challenges with existing systems at 45%. These aren't abstract worries – they reflect the reality that claims decisions carry significant legal, financial, and reputational consequences.

This cautious approach becomes even more apparent when examining current adoption patterns. A substantial 58% of respondents either don't use AI in their claims process or are uncertain whether their organization employs AI tools. This uncertainty itself is telling, suggesting that AI implementation in many organizations remains fragmented or poorly communicated.

Yet this trust deficit doesn't reflect a wholesale rejection of technology. Instead, it reveals an industry that understands the stakes involved and demands proven reliability before embracing new tools.

Key Insight #2: The Human-in-the-Loop (HITL) Solution

The survey's most compelling discovery lies in how dramatically trust levels shift when human oversight enters the equation. When respondents were asked about their confidence in AI outputs validated by expert reviewers, the percentage expressing medium or high trust jumped to 60% from 16%. Those reporting high trust soared from just 2% to 22%.

This trust multiplier effect varies by current AI usage. Among occasional AI users, 33% report medium or high trust in the technology, compared with 0% among those who don't use AI and have no adoption plans. This suggests that familiarity breeds confidence, but only when paired with appropriate oversight mechanisms.

Key Insight #3: Efficiency Over Everything

While trust in AI's decision-making capabilities remains limited, its value as a productivity enhancer is widely recognized. An overwhelming 75% of respondents believe AI can boost efficiency through improved speed and resource optimization, with nearly half (49%) also citing productivity gains via increased work volume capacity.

The areas where claims professionals see the most potential for AI impact align with administrative processing tasks rather than strategic decision-making. Document automation and data extraction led the way at 69%, followed by operational efficiency and workflow automation at 57%. Claims decision support, while still significant at 31%, ranked lower – a telling indication that professionals want to handle the groundwork, not the judgment calls.

This pattern extends to perceived benefits for claimants, as well. Respondents believe AI's primary contribution will be reducing administrative delays (71%) and enabling faster claims resolution (60%). They're less optimistic about AI improving accuracy (25%) or transparency (18%), suggesting a realistic understanding of current AI capabilities and limitations.

Key Insight #4: Industry Readiness for Adoption

Several broader trends emerge that paint a picture of an industry poised for significant change, albeit on its own terms. The claims sector's approach to AI adoption reflects a mature understanding of both the technology's potential and its limitations. The emphasis on efficiency gains over accuracy improvements reveals a pragmatic strategy. Claims professionals recognize that AI's current sweet spot lies in handling repetitive, time-consuming tasks that don't require complex judgment. By automating document processing, data extraction, and workflow management, AI can free human experts to focus on the nuanced work that genuinely requires their expertise.

This division of labor – AI for processing and humans for decision-making – represents a sustainable path forward. Rather than replacing claims professionals, AI becomes a force multiplier, enabling teams to handle larger caseloads while maintaining quality standards. The survey data suggests this approach resonates strongly with practitioners who see AI as a tool to enhance rather than replace their capabilities.

Key Insight #5: The Regulatory Reality Check

The highest ranking of compliance concerns (49%) in the survey reflects the claims industry's unique regulatory environment. Unlike consumer-facing applications, where AI adoption can move quickly, claims processing operates under strict regulatory oversight. Any AI implementation must meet not only operational requirements but also legal and compliance standards that vary by jurisdiction and line of business.

This regulatory awareness actually strengthens the case for HITL approaches. Expert oversight provides a vital compliance layer, ensuring that AI-driven efficiency gains don't come at the expense of regulatory adherence. The combination offers a way to modernize operations while maintaining the defensibility and auditability that regulators demand.

Key Insight #6: Building Toward a Broader Adoption

The survey reveals an industry transition, with 37% of respondents occasionally using AI and a further 38% considering adoption. This represents a significant cohort of organizations actively evaluating how AI fits into their operations. The key to converting consideration into implementation lies in addressing the trust concerns identified in the survey.

Organizations that lead in AI adoption will likely be those that successfully implement HITL processes from the start. Rather than viewing human oversight as a temporary bridge to full automation, successful adopters will likely embrace it as a permanent feature that enables both efficiency and trust.

Looking Forward to a Collaborative Future

The survey findings point toward a future where AI and human expertise can work in tandem. This collaborative model addresses both the operational pressures facing the claims industry and the trust requirements necessary for sustainable adoption. As AI continues to advance and demonstrate reliability in controlled environments, we can expect to see a gradual expansion of its role in claims processing. However, the HITL principle identified in this survey is likely to remain a priority in AI implementation in the claims space.

The claims industry's thoughtful approach to AI adoption may well serve as a model for other high-stakes sectors grappling with similar questions about balancing innovation with responsibility. By insisting on human oversight and focusing on efficiency gains over decision-making authority, claims professionals are charting a course that could accelerate AI benefits while maintaining the trust and reliability that the industry demands.

The survey data make clear that the question isn't whether AI will transform claims processing but how that transformation will unfold. The answer appears to lie not in choosing between human expertise and AI, but in finding the optimal combination of both.