AI use by insurance carriers will eventually become ubiquitous – or at least that's the prevalent hypothesis as AI development continues in earnest. Indeed, AI-native operating models will be a necessity to fully embrace AI's potential within the insurance landscape.

But we do not need to get to a fully automated environment to identify a significant opportunity for insurance carriers.

AI patents represent the next great competitive frontier for carriers, and most carriers are completely unprepared for what comes next.

Consider that AI patents are mostly filed by a handful of carriers, predominantly in the P&C space. But an estimate is that just three carriers have filed for 77% of all patents. That is a significant concentration of assets among a small number of carriers.

What does that mean for insurance carriers?

Immediate Implications

For insurance carriers, AI investment is likely driven by at least one of three considerations:

1. Reducing cost
2. Building on an existing strength
3. Addressing a deficiency or weakness

AI investment and development is too early-stage to truly address the third item. Lack of data, an inability to successfully drive adoption, and limited resources would not reward carriers for placing initial AI bets on areas where they are weak. For example, an insurance carrier with 30-day underwriting cycle times is not likely to invest in AI in this space. Instead, they will either focus on process improvements or leveraging an out-of-box solution that can instantly reduce 30-day cycle times into 10-day cycle times.

That leaves cost reduction or developing strengths as the primary motivation for AI investment.

In either situation, the development of a successful AI tool and its inevitable patent is defensive. This means it will help to develop a moat that keeps other insurance carriers at bay.

That may seem intuitive, but the concentration of AI patents among a few carriers tells a different story. Either insurance carriers have not achieved AI results strong enough to justify patents, or the industry has chosen to pursue trade secrets to protect its intellectual property. The trade secret route is unlikely – there is too much movement within the industry, and independent development of AI tools is an inevitability.

This lack of strategy is problematic for carriers – it will only widen the gap between performers if unaddressed.

Long-Term Implications

In the long run, carriers that possess AI patents will inevitably focus first on their strengths to solidify their market position. If a carrier already has strong underwriting discipline and cycle times, leveraging AI will only improve that strength within the market. To be sure, some level of trade secret and proprietary knowledge will make the AI tool more successful for one carrier over another, but a patentable AI tool provides strong defensive capabilities to the insurance carrier.

Imagine an annuity carrier that develops an AI tool that automatically reviews new business applications for annuity exchanges that involve an income rider – typically, this would trigger some enhanced, manual review. If instead an AI tool is designed that performs this specific function and a patent were issued on it, the carrier now has a unique position. Not only can it perform this well, but it can effectively block others from being able to do the same thing.

Now as patent lawyers will tell you, there are ways around this – a carrier could create a different process from the initial patent. But the importance is not that there are other ways to achieve it; it is that one pathway has been closed. And as carriers continue to invest in AI and develop AI-native processes, you could significantly increase the cost of doing business for competitors.

Factor in that the carriers filing patents are the strongest carriers, and strengths are enhancing strengths to create chasms between these carriers and their competitors.

But the value of patent development is not just defensive – it can also be an offensive tool.

A Hidden Financial Goldmine

Carriers should not just look at their patents as ways to protect themselves. While the value of patents may first be their protective nature, there is a significant opportunity for carriers to potentially monetize patents by licensing them to competitors.

Consider that in some instances, carriers have out-innovated insurtech firms. This has spurred insurtech from being a competitor to legacy carriers to being a partner.

Developing patents could be the evolution of this relationship where carriers incubate technical solutions, apply them internally, patent them, and then seek to commercialize them.

Will every patent follow this model? No. In fact, a defensive strategy should be the primary consideration to ensure that an insurance carrier maintains a competitive market position.

But in certain circumstances, owning the right technology is only half of the equation. Consider a carrier that has developed a lead-generation AI tool and can successfully patent and defend it. That tool will only be as useful as the data that is provided to it.

The licensing carrier could license the technology to another carrier (Carrier B), knowing that Carrier B does not have access to the same level of data as the licensing carrier.

The result? Carrier B can obtain significant gains, but not as significant as the licensing carrier will see. We see this today with lead generation models, where generic data still provides a significant lift in cross-selling and up-selling efforts, but not as strong as models that leverage proprietary data. But for Carrier B, who may be a laggard in the lead generation area, they have an opportunity to significantly improve their capability's maturity. This coopetition model allows all carriers to compete but provides the lion's share of rewards to the most innovative carrier.

What Carriers Need To Do To Unlock This Value

Insurance carriers that understand the value of these patents need to take concrete steps to be leaders in this space.

1. Identify Strengths: Patents require disclosure of the underlying model. The best patents will be areas where simply having the AI tool is not enough to win. Areas where the insurance carrier has operational expertise, unique data, or capabilities that cannot be easily replicated are good candidates for patents. This ensures that a carrier protects its competitive advantage while also having the capability to leverage its technology for monetization.

2. Develop A Patent Strategy: Not everything should be patented – some things may be internal trade secrets or rely on other protections. And most importantly, not everything can be patented. Insurance carriers should form teams that combine internal and external counsel, operational expertise, and technical leaders to evaluate which options are most likely to be legally defensible and impactful to the organization.

3. Design Commercialization Capabilities: Just as important to the patent strategy is the ability to commercialize the technology itself. Carriers need two capabilities. The first is the ability to implement the technology itself within the carrier successfully and recognize a benefit. That provides proof of concept and reaffirms what makes the tool successful. The second is to develop an incubator that can pursue partnerships with other carriers akin to how insurtech works with legacy carriers.

4. Adopt An Offensive Posture: Insurance carriers need to be aggressive with reviewing the patent landscape. When insurance carriers file patents, it provides a clear perspective on where other carriers are placing their bets. For example, a large number of patents on claims payments probably means a carrier believes they can differentiate in their claims experience. Carriers should review patent filings and prepare to be litigious, particularly mid-sized and smaller carriers. Insurance carriers cannot allow the competition to simply move unimpeded. If they do, they risk being pushed out of the competition.

AI patents provide a significant opportunity for insurance carriers to achieve strong defensive positions, with the potential for monetization in the future. But most importantly, as insurance carriers transition to AI-native operating models, controlling patents secures competitive positioning while successfully blocking others. The carriers that are able to develop the most effective strategies and execute will place themselves in a strong position as carriers begin operating in AI-native environments.

My auto insurer, Progressive, knows when I (rarely) hit the brakes hard. But why just use that information to determine my premium? Why not amalgamate data on hard braking and provide it to the people who design and maintain roads? 

If I'm the only one hitting the brakes hard at a specific spot, that's a me problem, but if loads of people are doing the same at the same spot, that's a systemic problem that some government agency can and should fix, heading off accidents.

A recent report makes the case persuasively and, I hope, will lead to the amalgamation and sharing of near-crash information by insurers. Doing so could save a lot of lives and avoid a lot injuries and property damage. 

Two researchers at Google looked at 10 years of public crash data, compared it against aggregated data on hard braking and found an extremely high correlation. They then used the hard braking data, separate from the crash reports, to predict trouble spots and again had very strong results. For instance, the intersection of Highway 101 in California with Interstate 880 in San Jose was in the top 1% of all road segments for hard braking — and police reports show a crash every six weeks, on average, for decades.

Google is making its data available to transportation agencies — and they should use it. While the U.S. has traditionally viewed auto safety as the responsibility of the individual driving the vehicle, European countries have shown the importance of system design. 

Using features such as roundabouts, protected bike lanes, lower speed limits and narrower lanes (which prompt drivers to go more slowly), European countries have far fewer traffic deaths per capita than the U.S. does. For instance, the U.K. reports 2.6 traffic deaths per 100,000 people per year; France, 4.9; Germany, 3.3; Spain, 3.7; and Italy, 5.3; while the U.S. reports 14.2.

The U.S. has such a car culture, including a love for pickups and massive SUVs, so I'm not sure U.S. roads will ever be as safe as those in Europe. But using hard braking, rather than police reports, provides information rapidly and overcomes the inconsistencies that arise because different police agencies handle traffic reports differently. The telematics also can extend the use of data to roads that, unlike the 101/880 interchange, aren't so heavily traveled and aren't such obvious outliers — simply because of randomness, an accident may not happen for a long time in a less-traveled spot, but hard braking can still alert authorities that a big problem exists. In addition, the telematics data can be more precise — you don't just notice that accidents happen in a certain spot but can see that hard braking picks up at a certain time of day, in certain weather or at a certain time of year.

In general, I wish the insurance industry had been faster to use the full capabilities of telematics. For years, they were just used to tell people after the fact that they had been recorded doing something dangerous. The incentive to do better was there but remote, because the incidents would only affect premiums somewhere down the line. It's only in recent years that telematics devices are being used to coach drivers in real time about being drowsy, following too closely, etc., and even now the focus is mostly on fleets of drivers, not individuals. 

I understand technology adoption curves, so I know the industry couldn't just wave a magic wand. I also realize that part of the issue is critical mass — you can't do something like aggregate data on hard braking if you don't have enough cars on the road using telematics that can report instances of the behavior.

But I think back to how magical it seemed 25 years ago when I wrote something about how it was going to be possible to learn about traffic jams in real time, because authorities were going to track mobile phones in cars. If phones in an area were stopped, you had a problem. If they were all going 75mph, things were all clear. 

And I think we're at this sort of place now with sensors in cars. Hard braking is actually just one example. Sensors will be able to report on potholes or other problems with roads. Dashcams can monitor for other safety-related issues, including crazy driving. (Yes, privacy will be a thorny problem.)

But, for now, I'll be happy if we can just get information on potential accidents into the hands of the right people so they can do what they can to head off fatalities, injuries and property damage. Traffic deaths in the U.S. have been falling in recent years, but more than 40,000 people lost their lives on U.S. roads in 2024, and that's far too many, even for a country that loves its cars.

Cheers,

Paul  

In 2026 and beyond, eight transformative trends will reshape strategy, technology, operations, products, and talent across the industry. From the rise of AI-native core systems and human-centric AI, to the expanding Silver Economy and the explosive growth of specialty markets and products like parametric insurance, these trends highlight the urgency—and the opportunity—to rethink traditional assumptions.

Download Majesco's full report to learn:

• The eight trends that will define 2026—and how they will reshape strategy, technology, and customer value.
• Why AI-native technology and reimagined operating models are now mission-critical for competitiveness.
• How new market forces—from demographic shifts to climate risk to InsurTech instability—will influence growth opportunities and partner strategies.

Read Now

Sponsored by ITL Partner: Majesco

Growth is getting harder as margins tighten and agency workflows stay manual. This eBook lays out 10 practical ways to speed up quoting, reduce re-keying, cut costly errors, improve follow-up, and help your team write more business.

Download the eBook

Sponsored by ITL Partner: bolt

Sponsored by: ITL Partner: PwC

Sponsored by: ITL Partner: PwC

With over 25 years of operational and advisory insurance experience, Yuri Poletto leads the Open and Embedded Insurance Observatory and co-leads the Cyber Insurance Frontier, member-led, executive communities focused on the evolution of insurance distribution, ecosystems, and cyber risk within regulated markets. Both initiatives convene senior leaders from across the global value chain - including insurers, brokers, tech providers, and non-insurance brands - to shape forward-looking strategies. These communities host executive plenary sessions, workshops, strategic roundtables, and research initiatives across Europe, North America, and Asia, helping market leaders navigate complexity with better-informed strategic decisions.

Paul Carroll 

Travel insurance has long been a great example of how insurance can be embedded into another process, especially buying plane tickets, but embedded insurance is moving far beyond that. What are your favorite more recent examples?

Yuri Poletto

One of the most interesting and inspiring recent examples of embedded insurance involves a Brazilian fintech called NuBank. NuBank launched an embedded life insurance proposition and achieved remarkable sales. What's particularly interesting to me is that nearly 50% of the buyers of this life insurance were first-time insurance buyers.

This is a remarkable example of how embedded insurance is not only about seamless distribution and innovation but is effectively a way to grow the market, enlarge the pie, and reduce underinsurance—which is an issue in every part of the world.

They sold 2 million policies in just a few years, and, as I said, nearly half went to first-time buyers of life insurance. 

In hearing care, a global leader in this niche provides hearing aids and comprehensive auditory services. They have been a pioneer in the space, offering embedded insurance since 2014. Today, over 200,000 of their clients in Europe protect their hearing aids through these integrated policies, ensuring peace of mind for a critical and expensive medical device.

Embedded insurance is repositioning itself from a distribution efficiency model into a model that can create markets.

Paul Carroll

Where do you see embedded insurance breaking out in the next few years?

Yuri Poletto

I see several areas. Some of these opportunities are led by digitalization of markets. For example, home insurance and households is an area where we already see some activity, but we'll see much more thanks to the growth of proptech platforms. They see embedded insurance as a clear opportunity to gain additional margins.

Telcos, particularly in Europe, are increasing their commitment with embedded insurance. In this case, they’re driven by competition. They're suffering because of low-cost competitors, so it becomes fundamental for them to retain customers, and selling embedded insurance is probably the best way to increase their chances.

Similar dynamics apply to auto insurance. We know that car manufacturers and dealers basically don't gain margins from selling the cars, but they gain margins from services. Insurance, after-sales services, extended warranties, and maintenance are a fundamental part of this.

Another area for growth is the gig economy. Today, there are niche providers of embedded insurance for the gig economy. I think we'll see more in the future because the gig economy is huge. In the U.S., nearly 40% of workers are freelancers, and they don't have access to insurance for their freelance work.

Cyber, for me, is one of the largest opportunities for embedded insurance because attaching cybersecurity and cyber protection tools has proved to be the most effective way to sell cyber insurance. 

I think we'll see lots of evolution in embedded insurance and many areas of growth. 

Paul Carroll

Embedded insurance makes sense because it reaches people when they're already thinking about potential risks, but one significant concern is customer ownership. If a bank offers life insurance to its customers, who ultimately owns that customer relationship—the bank or the insurer? If it's the bank, doesn't that put the insurer at risk of being replaced at any time?

Yuri Poletto

It’s a critical question, Paul, and frankly, the one that keeps legal teams up at night during the implementation phase. In the U.S., the hurdle isn't just "who owns the customer" but "who holds the license." Because insurance is regulated at the state level, scaling a national embedded program across over 50 jurisdictions requires a sophisticated licensing strategy. Often, the answer lies in the MGA model. By using an MGA structure, the partners can clearly define customer ownership and data rights in the contract while ensuring that the entity facing the customer, whether it’s a tech platform or a bank, is operating within the strict boundaries of state licensing laws.

The relative strength of the partners is also important. Insurers are used to being the "big guy," but when they partner with a Big Tech firm or a major retailer, they are meeting an equal. The smart way to solve this is through modular compliance: building a tech stack that can handle different regulatory requirements state-by-state, so the insurer can remain the "manufacturer" while the distributor owns the "experience" without the insurer fearing they are becoming an interchangeable utility.

We see massive volumes of this working in Europe and Asia, and it's happening in North America, too. The compliant paths exist; it’s just a matter of designing the legal and licensing "plumbing" as carefully as we design the user interface.

It can be done.

Paul Carroll 

People generally don't like to buy insurance, so they would prefer to purchase one comprehensive policy to cover everything. However, with embedded insurance, consumers are being offered individual policies when purchasing specific items like expensive watches or jewelry. What is your view on the tension between these individual policies for specific purchases and the idea of an overarching policy that would cover everything?

Yuri Poletto

That's a good point. I see it this way: Embedded insurance isn't a replacement for traditional distribution, it’s an expansion of the market. It "enlarges the pie" by reaching customers exactly at the point of need, often during significant life milestones or transactions where insurance might otherwise be an afterthought.

For example, many SMEs are chronically underinsured because the traditional broker model often struggles to efficiently handle complex risks when premiums and commissions are low. Embedded solutions bridge that gap by automating the process within the software SMEs use to run their business.

Furthermore, in the home insurance market where the comprehensive policy model is the standard, we can now integrate coverage directly into the mortgage closing process, a home maintenance subscription, or the moment a tenant signs a lease. This context-driven approach makes protection a natural extension of the purchase rather than a separate, secondary chore. Even the most comprehensive homeowners or renters policies have blind spots, like coverage limits on high-value jewelry, specialized electronics, or specific liabilities related to remote work. Embedded insurance also serves as a vital complement here, filling the gaps that a "one-size-fits-all" policy misses.

Ultimately, it isn't a battle of "one policy versus many." It’s about utility. If a customer finds a bundled policy more convenient and in line with their needs, they’ll choose it. But when a frictionless, point-of-need offer provides targeted protection, that is where the market truly grows. In the end, the most frictionless customer experience will prevail.

Paul Carroll

Back-of-the-envelope calculations suggest embedded insurance can save 15% to 25% of the first year premium on policies like life insurance by engaging customers already active in another environment. What other numbers best demonstrate the effectiveness of embedded insurance?

Yuri Poletto

This is one of the big selling points of embedded insurance, the fact that the distribution chain is simpler, with fewer intermediaries, resulting in lower commissions and lower costs for end customers. Sometimes it's true, but other times it's not. Particularly in the past, before there was more attention around this issue, B2B2C affinity insurance was sometimes sold with commissions of 70% or 80%, which didn't make it any cheaper than insurance purchased through another channel.

I have seen significant savings in commercial lines in the U.K., for example, where embedded insurance sold through SaaS companies achieved savings of up to 40% in price. I don't have particular numbers for personal lines, but I can speak to commercial lines because I've worked directly with one of the members of the observatory in the U.K.

The theory and the model are sound—once you're dealing with a brand that already has a customer base and you manage to set up a commercial agreement that works well and is streamlined, you can definitely achieve significant savings. Obviously, if you charge huge commissions, you eat into all those savings.

But I see the market moving toward more transparent settings and lower commissions. We've also seen some cases where prices are reduced thanks to profit sharing with customers. 

Paul Carroll

Technology clearly has enabled embedded insurance. Is it now as good as it needs to be, or can we expect meaningful improvements as technology continues to advance?

Yuri Poletto 

Unlike in the past, the technology for embedded insurance is mostly available today. It's no longer niche and super expensive, but mainly commodity. So the issue is not so much the availability of technology but the fact that often organizations prefer to try to build the technology internally rather than using technology that has been developed by specialists. This is a bit of a blocker.

APIs, obviously, are the key kind of technology when we talk about integration of insurance in non-insurance flows. There are a lot of enabler companies that provide these kinds of insurance services. We are already seeing a consolidation from the vendor standpoint, and we probably will see more now that capital is no longer cheap, because many enablers rely on venture capital.

Paul Carroll

This is all super-insightful. Thanks, Yuri.

Valentine's Day brought a flurry of reports about devastating romance scams perpetrated by organized, remarkably patient cyber criminals. 

While insurance doesn't cover these individual scam victims, they dramatize what insurers are up against as criminals become more sophisticated and tap into AI's capabilities as they go after the companies that insurers do cover. 

It's a scary sight — and may get worse.

Fortune magazine told the tales of three women who have come forward to warn others as part of their efforts to combat romance scams. They described great patience by the scammers and careful wooing; one even invested $185,000 to fool his victim.

After meeting a man on a dating site, one of the women received numerous trinkets from him — then a $100,000 check. She worried the check was fake, but it was real, and she deposited it. He then began planning a birthday bash for her on a yacht and laid out $85,000 for deposits and various purchases. He texted her constantly — she counted 20,000 text messages, in all — and asked her to pray with him daily. Eventually, he suggested she withdraw her life savings and invest in a sure-thing crypto venture, which she did. She received email weekly reports, likely generated by AI, that showed thousands of dollars of gains each time — but those were fake. She lost $1 million in savings, as well as her condo.

Another woman felt she had bonded with a man and was preparing to meet him, when he asked her to help him log into his bank account while he was abroad. He supposedly needed to pay a translator as part of a deal that would net him a $10 million payment. When the woman logged into the bank account, she saw $700,000 in cash. The next day, both were locked out of the bank account, and he said he needed $20,000 for an additional payment. Reassured by his impressive bank account, she sent him the money. He pocketed it. 

The third woman was wooed by a man who not only seemed to care for her but for her six rescue dogs. When she'd talk with the scammer on the phone, if one made noise in the background, he'd ask, "Is that Duffy? Is that Trixie?" He gradually talked her out of her life savings, as well the life insurance payment she'd received when her husband died. She was left with so little money that she she couldn't repair her air conditioner, a decision that led to a fire that burned her house down and killed all her dogs. To escape, she had to flee the house in her underwear at five in the morning, in a neighborhood where she'd lived for 42 years. 

Vox describes AI as a "force multiplier" for these romance scams, which occur all year but pick up around Valentine's Day. Someone who might have been able to run a few scams at a time pre-AI can now operate 20 or more simultaneously.

"On the dark web," Vox says, "fraudsters can purchase romance scam toolkits complete with customer support, user reviews, and tiered pricing packages. These toolkits come with pre-built fake personas with AI-generated photosets, conversation scripts for each stage of the scam, and deepfake video tools, [Chris Nyhuis, the founder of cybersecurity firm Vigilant] told me. 'The skill barrier to entry is essentially gone.'"

Richard Graham, the practice lead for financial crime at Moody's, told me these scams are becoming more ambitious. 

"Five years ago, six years ago, [scammers] were just people online, trying to get your information, get a couple bucks and move on. Now... instead of asking you for money on day one, they're building rapport. They're spending a lot of time with you to better understand who you are.... sometimes over weeks, but usually months.... It's not just a $3,000 payment they want. They want everything."

Graham says it's hard to know just how many billions of dollars a year are lost to these scammers, because the vast majority of victims are too embarrassed to report the crimes. But he says a UN report found that some 235,000 people worldwide were working in professional organizations, largely in Southeast Asia, two or three years ago to perpetrate these romance scams, and he assumes the number has grown since then. 

He says AI helps these professionals develop better scripts to use as they groom their victims. It can also can help those who aren't native English speakers to smooth out any issues with their language skills. AI certainly helps gather information from social media sites as scammers try to learn as much as possible about their marks. AI also makes it easier and less expensive to cast a wide net of messages that could begin interactions with potential victims. While the techniques used to woo victims are referred to as "love bombing," the thieves refer to their goal with a crasser term: "pig butchering."

Graham says AI isn't being used much at the moment to generate deep fakes as part of romance scams — but only because they aren't needed just yet. 

"That actually has been a surprise to me. I would have thought at this point that deep fakes would have been a much bigger problem," he said. "They are a problem, but they haven't really scaled yet, because these other scams [based on dating sites, social media and text messages] have just been so low-effort and so successful."

That's not a happy thought for cybersecurity, in general: Thieves have enormously powerful tools at their disposal — but don't even have to use them much just yet because people are still so easy to fool.

The only solution is to escalate our vigilance as fast or faster than the thieves are escalating their capabilities. The stories of corporate victims aren't as dramatic as those of the poor women Fortune described, but the financial damage is orders of magnitude greater. 

Cheers,

Paul

P.S. Graham offered a pointer for anyone who thinks someone they know might be getting courted as part of a scam. He says thieves always want to get victims into a chat app as fast as possible. That way, they don't put at risk a social profile they've spent a great deal of time crafting and won't be caught in any safeguards set up by the social site. So you can simply ask whether the person you're concerned about has any chat apps on their phone. That question won't raise hackles in the way that a statement like, "I'm worried about you," would. If the person has a chat app, you can explore further and perhaps head off a financial catastrophe. 

Epic Systems dominates healthcare IT—over 35% of U.S. hospital market share, trusted by most major health systems, and increasingly positioning itself in the insurance/payer space with Tapestry (health plan platform) and Payer Platform offerings. If you work in health insurance and haven't heard of Epic, believe me, you will.

Founded by Judy Faulkner in 1979, Epic built its reputation on customer obsession, deep integration, and never selling out to private equity or going public. For health insurers evaluating Epic's growing footprint in claims, care management, and member engagement, this trust matters. But Epic's AI strategy introduces a dependency chain that should concern any COO or CDO betting on long-term operational transformation.

Epic's AI road map runs almost entirely through Microsoft Azure and OpenAI. Ambient documentation, predictive analytics, revenue cycle automation, clinical decision support—all built on the Epic-Microsoft-OpenAI stack. This isn't a vendor partnership; it's architectural dependency. And Microsoft just confirmed how deep that dependency runs: in its latest earnings call, they reported that fully half of Azure's AI inference load runs on OpenAI models.

For Epic customers, this creates compounded risk. You're not just betting on Epic's execution—you trust Judy Faulkner, and rightly so. You're betting on Microsoft's sustained healthcare commitment and OpenAI's organizational stability.

Microsoft has tried and abandoned healthcare repeatedly: HealthVault (2007-2019), healthcare cloud initiatives that quietly deprioritized. Healthcare represents less than 5% of Microsoft's revenue. Their actual priorities: Azure infrastructure, Office/Copilot, Gaming, LinkedIn. If OpenAI's healthcare AI underperforms or faces regulatory barriers, what's Microsoft's incentive to double down versus pivot Azure AI resources to more profitable verticals?

The OpenAI dependency may be more concerning. Microsoft has invested $13 billion for 49% ownership of OpenAI's for-profit entity, but that doesn't buy control over strategic direction, safety culture, or talent retention. OpenAI's November 2023 board crisis—where the CEO was fired for trust issues, then reinstated via employee revolt within 96 hours—revealed governance dysfunction that never fully resolved. The safety-focused board members and researchers who prioritized responsible development over shipping speed have largely been sidelined or left to found competitors like Anthropic. What remains is a growth-at-all-costs culture increasingly optimized for investor returns. Two years ago, OpenAI laughed off the notion of ads. Last month, they started running ads.

For insurers deploying AI into prior authorization decisioning, claims adjudication, clinical documentation, and fraud detection, governance matters. This isn't consumer chatbot territory where failures mean embarrassing screenshots. This is financial exposure, regulatory risk, and potential patient harm. If OpenAI faces safety incidents, regulatory sanctions, or capability degradation, Epic's AI roadmap stalls and you bear the operational consequences with zero recourse.

Then there's pricing. OpenAI's current API costs are venture-subsidized loss leaders. Post-IPO pressure or when Microsoft demands ROI on their $13 billion investment, inference pricing will spike—potentially three to five times current rates. Epic will pass these costs through as "AI-enhanced module" increases. Your negotiating leverage? Approximately zero. You're a third-order customer with no direct relationship to the entity setting prices.

Microsoft's earnings revelation—that half their AI load runs on a single vendor with documented governance issues—should trigger every Epic customer's (or potential customer's) risk management protocols. Epic has consolidated AI strategy into Microsoft; Microsoft has consolidated its AI capabilities into OpenAI. Three single points of failure, any of which could spike pricing, degrade capabilities, or shift strategy in ways misaligned with your operational needs.

When Epic comes knocking, you can trust Judy Faulkner's execution. But Epic's AI future depends on Microsoft's healthcare commitment and OpenAI's organizational stability—neither of which has a reassuring track record. The health insurers negotiating on this dependency will maintain leverage. Those that don't may find themselves funding Microsoft and OpenAI's margin expansion while paying for the privilege.

At least 94 million families own a pet in the U.S., yet just one in 10 rental properties allow animal companions without restrictions. In the context of a housing affordability crisis that is driving more Americans to rent, many face the inevitable question of whether to choose convenient housing or their pets.

Property managers and insurers alike face continued uncertainty around pet-related liability in this scenario. Many are left with no choice but to fall back on blanket restrictions; when lacking documentation reigns, these stakeholders operate blind.

With the average dog-related insurance claim now hitting nearly $69,000, managers and insurers can no longer afford one-time approvals or all-encompassing measures; dog bite claims specifically have risen over 48% in the past decade.

Meanwhile, communities increasingly demand both pet-friendly environments and safe, fair and forward-looking processes from their insurers.

Nuance in underwriting

Risk models rest on flawed assumptions, policies may not match real-world conditions, and when claims do arise, they depend on interpretation more than on established facts. Pet risks get obscured, and are quietly shifted to insurance carriers.

A property, for example, allows for all dogs under 25 pounds, because managers assume they pose less risk than a 100-pound Rottweiler. A tenant then signs a lease, and with them comes a Chihuahua with a long history of aggression: it has bitten humans and, once, latched onto the upper lip of a neighboring Rottweiler.

The lease insurance policy does not capture such nuance. Months later, when the Chihuahua bites a visitor, an insurance agent is contacted, and the claim is evaluated: the insurer could interpret the situation as "small dog equals low risk," or not. Without consistent and accurate record-keeping, coverage might be disputed or delayed.

Alternatively, if a blanket ban is imposed, the Chihuahua will probably still be living in the property – out of sight, until a problem occurs.

Precedents attesting to the complexity have long been set. In California's landmark 1995 Donchin v. Guerrero case, Alpha Donchin and her Shih Tzu were attacked by two Rottweilers that escaped from a rental property four blocks away, leaving her with a broken hip. The property manager denied knowing the Rottweilers even existed in his property, yet they had escaped through a damaged fence, resulting in the court ruling liability; he knew, or should have known, about his tenant's dogs and their aggressive behavior.

Risk does not need to be unknown. New approaches, including standardized pet screening, continuing documentation and compliance, data-driven risk insights, proactive mitigation measures, and risk-based insurance policies can make liabilities quantifiable.

Solutions combining pet screening with dog bite insurance are new, but very much needed. Consistency in processes can be tracked – including everything from emotional support animal verification to vaccination monitoring and document management –, flexibility poses wins for all parties involved, fast screening offers convenience, and new revenue is generated through responsible pet ownership programs and curing manual screening costs.

Why is the insurance industry lagging?

One size does not fit all. Restrictions, including banning certain breeds, sizes, weights, behavioral requirements, or vaccinations are not born from previous negative experiences, but rather indirect sources of concern. Because of a lack of information, managers and insurers alike also tend to overestimate costs associated with pets.

Bottom line, there is no way to analyze risk without getting the complete picture. In looking at pet insurance, this industry golden rule rings even truer. For too long, rentals have relied on vague and outdated restrictions that leave tenants and managers in the dark; traditional safeguards like breed discrimination are not strong predictors of a pet's behavior.

Legislation is already catching up. Last year, a Florida County Council required owners of "dangerous" dogs, excluding specific breeds but including those who had severely injured other animals without provocation more than once, to carry at least $500,000 of liability insurance. But even here, terminology like "trained to attack" or "bred for fighting" are poorly defined.

Standard homeowner policies, however, continue to provide from $100,000 to $300,000 on average in liability coverage. To mitigate this gap, stakeholders have turned to exclusion rather than innovation: blanket bans from tenants or limitations in coverage, with some carriers refusing to write insurance for renters with animals. It is now time to catch up.

Measurable pet injury liability

Accurate pet screening must be combined with liability coverage to make dog bite risk quantifiable. Teams must have access to all-encompassing systems that help verify pets, track vaccinations and documentation, manage compliance and maintain consistent records.

For example, traditional policies recognize that multi-family properties face higher liability exposure than single-family homes because of their communal components: shared walls, common areas, multiple tenants, and manager responsibility for maintenance of spaces. Yet, cases like Donchin v. Guerrero demonstrated high risk also exists by merely living in residential areas, regardless of whether tenants rent apartments or single-family houses.

Vulnerability should be analyzed individually, proactively, and consistently: pets screened singly, risks minimized both before animal companions move in and during their tenancy, and liability coverage must reflect these nuances.

There is already evidence of managers setting up these good practice pillars by setting up pet meetings prior to lease signing, for instance, but these have proved futile in many cases; pet bite claims remain alarmingly high.

Expecting stakeholders to manually follow up with pet records, vaccination requirements, and bookkeep audit-ready documentation as data becomes more granular is also increasingly unrealistic.

This is where technology, in the form of AI-powered automation, centralized real-time insights, and integrated application programming interfaces (APIs), comes into play. Such a measured approach supports premium adjustments, deductibles or coverage caps, rather than relying on traditional – and often ineffective – pet bans.

Five pillars that guide technological adoption in pet-related insurance claims:

• Centralization: Portals that integrate pet approvals and bundled dog bite coverage reduce breed bias, ensure compliance and reduce liability.
• Customization: Managers can set their individual pet approval criteria that align with insurance policies.
• Automation: Streamlining of resident onboarding processes reduces staff time spent on pet approvals.
• Derisking: Minimize liability and discrimination risk with standardized and transparent processes and records.
• Inclusivity: Objective scoring model that reduces bias and promotes transparency.

This change in paradigm offers more than the peace of mind for managers and insurers; it also supports community-building, the promotion of longer tenancies, and better communication.

In the end, pet bite liability and provisions to support all stakeholders involved have shown the real power insurers can leverage, for the benefit and betterment of communities.

The next step, then, will come from them, as processes are smoothed out through technologies, responsible parties assume the eagle-eyed view that is now required, and seize the opportunity to move from reactive to proactive.