Watch Our Demo: 

Why We Are the Right Solution For Your Needs

ScienceSoft's AI Agent for P&C Claims Coverage Verification adapts to your existing operations and technology landscape rather than forcing you to work around the software.

• Comes ready for coverage verification. The agent can automate the end-to-end workflow out of the box, from policy research and coverage analysis to evidence-backed memos and reservation of rights letters.

• Fits your claims processes and rules. ScienceSoft can customize the agent's verification logic, workflows, and outputs to reflect your coverage policies, authority levels, and claims handling practices.

• Lets you set the boundaries for AI. Built-in controls allow you to define which findings require human review, when the agent should escalate a case, and how AI-generated conclusions are validated.

• Works with your existing systems. The solution can integrate with commercial claims platforms and legacy applications without limiting your technology stack. You can deploy it in your cloud or on-premises environment.

Watch the demo to see the agent verify coverage for a homeowners hail damage claim.

Main Benefits of Our Product: 

• Faster coverage verification. In early tests, the AI agent helped cut coverage verification time from 2–3 days to just 30 minutes by automating policy analysis and document preparation.

• Greater claims team capacity. Even with human review of AI findings, the agent frees up 20–30% of adjusters' time spent on coverage verification for standard claims, allowing teams to handle more cases without proportional headcount growth.

• Fewer ineligible payouts. The agent helps flag claim-policy discrepancies and potential fraud before payment. For 50,000 annual claims averaging $5,000 each, preventing ineligible payouts on just 0.1% of claims could save an insurer $250,000

What Part(s) of the Insurance Industry Can Benefit From Our Product: 

• Claims

• Policy Administration

• Fraud Detection

• Customer Service

Learn More

Interested in learning more about ScienceSoft and its AI Agent for P&C Claims Coverage Verification? Please visit these sites to learn more or schedule a personal demo:

• https://www.scnsoft.com/insurance 

•  https://www.linkedin.com/showcase/sciencesoft-insurance/

•  https://www.youtube.com/watch?v=iBus093ow5I 

We've built an extraordinarily sophisticated financial system around industrial failure. We can price it, insure it, reserve capital for it, and finance recovery when it happens, but preventing that same failure often requires an entirely different standard of proof.

Insurance is exceptionally good at mobilizing money after something goes wrong. Natural catastrophes caused an estimated $100 billion in economic losses in the first half of 2026. Insurance covered approximately 42%, leaving nearly $60 billion to be absorbed elsewhere. Insurance's ability to fund recovery is indispensable. The shortfall also exposes a contradiction in how risk is funded.

Premiums, deductibles, and recovery costs have established places in corporate budgets. A proposal to prevent an asset failure faces a tougher test. Someone must prove the threat is serious, determine what intervention it warrants, and defend spending money on an event that might never happen. Failure has a funding mechanism. Prevention has a business case to make.

Insurance can transfer part of a financial loss. It cannot restore missed production, repair damaged customer relationships, or undo the wider consequences of an industrial disruption. The challenge is making prevention as intelligible and investable as recovery.

Knowing the risk isn't the same as managing it

Industrial organizations rarely lack information. They receive risk-engineering reports, inspection findings, maintenance records and alerts from connected equipment. Yet information doesn't guarantee action.

A temperature change, unusual vibration, or a deteriorating component has little meaning in isolation. Its significance depends on the asset's role in the operation. Is there redundancy? What production depends on it? What would a shutdown cost? Could a failure disrupt customers, supply chains or other facilities?

These questions turn technical observations into business consequences. Without that translation, recommendations can remain in reports while maintenance teams address more immediate demands and capital committees fund projects with clearer returns.

Industry has invested heavily in detecting changing conditions and predicting potential failures. Those capabilities matter. Still, knowing what might fail doesn't tell an organization which emerging risks deserve attention first.

A relatively minor condition on a critical asset may represent far greater exposure than a severe condition on equipment with sufficient redundancy. The challenge is understanding what matters most, what the consequences could be, and where intervention will create the greatest value.

Prevention needs to compete for capital

Consider a manufacturer that identifies deterioration in a critical component on its main production line. Although the component hasn't failed, its condition is changing. The organization must compare the cost of replacing it during a planned shutdown with the lost output, repair time, and disruption that could follow an unplanned failure.

The consequences, however, may extend well beyond that production line. A shutdown could affect customer deliveries, contractual commitments, supply-chain continuity and financial performance. What appears to be a maintenance decision may represent a much larger business risk.

The appropriate response might be immediate repair, closer monitoring, a change in operating parameters or acceptance of the exposure. Someone must have the authority to make that decision and commit the necessary resources before the opportunity to intervene passes.

Now multiply that challenge across hundreds of assets and facilities, each presenting different levels of exposure. The question is no longer simply whether an individual intervention is justified. It's which interventions deserve priority when resources are limited, and the consequences extend across the enterprise.

For decades, industry has developed sophisticated ways to measure losses after they occur. We need comparable discipline in understanding the economic value of preventing those losses before they happen.

Capital should not only be deployed to absorb loss. It should be possible to direct it intelligently toward preventing loss.

The standard for resilience must rise

Risk assessments, inspections and engineering recommendations are essential. Completing those processes, however, doesn't necessarily mean an organization has addressed its most consequential exposures. An inspection can establish that a process was followed. A risk assessment can document a vulnerability. Neither, by itself, demonstrates that the organization is becoming more resilient.

Industrial risk isn't static. Asset conditions change, operational dependencies evolve, and an exposure that appeared manageable during a periodic assessment may become significantly more consequential.

A recommendation without a named decision-maker and deadline is information, not control. Someone must have the authority to accept the exposure, fund the response or explain why another course is justified. Following the process is important, although it's not the same as preventing the loss.

Insurance can help change the economics of prevention

Insurers can make recommendations easier to act on by explaining the severity and urgency of an exposure, not simply recording that it exists. Brokers can connect risk-engineering findings with an organization's insurance and resilience priorities. Operators must determine whether to intervene or accept the exposure and take responsibility for the outcome. There is also a bigger opportunity here.

Insurers and brokers possess extraordinary expertise in understanding, quantifying, and pricing risk. Much of that expertise has historically helped organizations prepare financially for potential losses. What if more of it could be applied to making prevention economically compelling before those losses occur?

Imagine insurers, brokers, and industrial operators working from a more continuous, consequence-based understanding of emerging physical risk. Rather than relying primarily on historical losses and periodic assessments, they could better understand where exposure is developing, which conditions present the greatest business consequences, and where intervention could materially reduce risk.

This wouldn't eliminate uncertainty or diminish the need for insurance. It could, however, create a stronger economic foundation for prevention and a more productive relationship between risk financing and operational decision-making.

For decades, the insurance industry has helped businesses understand the financial consequences of what can go wrong. That expertise will remain indispensable, but the next opportunity is to help make the economics of what doesn't have to go wrong equally visible.

We've become extraordinarily good at financing industrial failure. It's time to become just as disciplined about investing in its prevention.

Bill Gates recently said he is using generative AI as part of the annual, two-week strategy session that determines how the Gates Foundation spends its next $10 billion. He'll be "taking the strategy notes and actually engaging in a dialogue with ChatGPT, Claude, Copilot, and even having them talk with each other," he said.

I don't have $10 billion that I'm considering how to spend, and you may not, either, but genAI has now improved to the point where it can at least be an aid in setting strategy at almost any level. 

An old boss, Mel Bergstein, used to say: "Strategy is what you don't do." The idea worked for Mel, who founded a consulting firm, Diamond Management & Technology Partners, that pioneered the notion of digital strategy in the mid-1990s. Chunka Mui and I took Mel's adage further in our book, "Billion Dollar Lessons," for which we had 20 researchers spend two years looking at 2,500 major failures to identify the warning signs.

Our recommended solution boiled down to having a "devil's advocate" identify all the underlying assumptions for a strategy, to keep problems from being glossed over as an executive team rushed to realize the CEO's plan. GenAI, used right, can now do a pretty good job of filling that role for anyone.

Let's have a look. 

Gates told the New York Times that, "in a few of the reviews, we’ll actually have the A.I. sit in. In a few cases, we’ll tell it: Hey, only speak if we ask you. And then in a few other cases, we’ll say: Hey, if you hear something you think is wrong or you hear us thinking about, for example, what are these statistics? Please engage. So we’ve gone from it being a joke [when someone suggested using genAI a year ago] to it will be a peer — not making any final decisions, but it will be a peer in deep, complex strategic discussions, making a significant contribution."

I'm not willing to go that far just yet, partly because I'd need to test the domain knowledge that the large language models have about a subtle, heavily regulated business like insurance and partly because genAI has shown itself to be a sycophant. The biggest, overarching pattern of failure that Chunka and I found in the research for "Billion Dollar Lessons" was that people were reluctant to point out obvious problems that could derail initiatives that had the imprimatur of the CEO, and we don't want to replace human sycophants with AI ones. 

But you can describe your strategy to an AI, then pepper it with questions, going deeper each time while demanding evidence for its thinking and insisting it be blunt.

Let's say you're thinking of buying a company. You could start with a general question: What could go wrong? Then you probe. How will our competitors react? What if interest rates increase? The economy goes into recession? The war in Iran escalates? Will we actually be able to cross-sell to our customers? Will the efficiencies from scale be as great as we expect? Who's done something similar and had problems? What were the problems? And on and on and on. 

You can do the same exercise if you're thinking of a new line of business, a software overhaul, adopting AI chatbots — or if you want to explore the dangers of standing pat, of not undertaking a major initiative that is under consideration. 

Major corporations have staff do extensive analysis along the lines I'm suggesting and, in the case of a possible acquisition, draw on their investment banks for advice — but the bias is always toward acting. If the perception is that the CEO backs an idea, the staff strives to make it work, not to kill it. Investment banks only collect their hefty fees if the deal happens, so why would they recommend against it? 

When Chunka and I did a consulting project with a major company that was considering launching autonomous vehicles, we saw elaborate studies it had commissioned from consulting firms that said the robotaxi business was a slam dunk and that the first to enter would win big. That was more than eight years ago. 

It turned out, however, that the executives working on the AV project internally were aware of all the sorts of operational and safety issues that we've all seen play out and that have slowed the robotaxi rollout. Those concerns were brought to the surface in the interviews for our devil's advocate review. 

Warren Buffett used to set aside two, equal pots of money when asking investment banks for advice on possible acquisitions — one for a bank that argued in favor of a deal and one for a bank that argued against it. Whichever carried the day collected its fees, while the other got nothing. You can be sure the "anti" bank worked as hard as possible on its arguments and evidence. 

In my experience, many big ideas should fall apart at the slightest scrutiny. As we wrote in "Billion Dollar Lessons," we thought 47% of the failures we investigated were based on strategies that never should have made it out of the strategy room — and not because those involved were stupid or inexperienced. 

The fifth-biggest cement maker in the world reasoned that it was really in the home products business because so much of its cement was used there. It began also making furnaces and lawn mowers... then went bankrupt and was bought by a competitor. Avon decided that its core competency was its culture of caring — not, you know, its line of cosmetics and its devoted salesforce — so it bought a maker of medical equipment and an operator of retirement homes. It had no idea how to operate the new businesses and quickly sold them at a major loss. In the early days of fax machines, FedEx spent hundreds of millions of dollars to set up a business that would collect a document from you, drive it to the nearest office, fax it to the office nearest your destination, then drive the fax to the recipient. The company somehow missed that Moore's law would soon make the devices so inexpensive that every company would have its own.

It may take time to learn how to use genAI to avoid costly strategic mistakes — and that's even before we start using it as broadly as Gates is planning to. You'll likely also need to introduce to the LLM information specific to your market, including regulatory issues that it likely doesn't have even as these models try to Hoover up everything available on the internet. Insurance is a complicated market.

But I think we're now at the point where genAI can, if managed correctly, start to catch strategic errors. We won't get that 47% figure for obviously bad ideas down to zero, but even heading toward 40%, maybe even 33%, would mark real progress.

Cheers,

Paul

Imagine the customer of a 20-year-old life insurance policy trying to figure out if it makes sense to surrender the policy. Do they call the insurance company first, hoping to talk to a knowledgeable agent? Probably not. They will upload the policy contract, riders and most recent statement to the AI assistant with a list of questions: What benefits would be lost by surrendering the policy? What happens to the outstanding loans? Are there other alternatives to surrendering the policy? Which policy provisions are the most important to consider before deciding?

This AI might be right, or might be partly right, or might be very wrong. But what is becoming abundantly clear is that this customer arrives at their insurer with a level of analysis and a body of knowledge that once cost time, or cost money, or both.

This leads me back to the question insurers have been asking for years about how AI will change what we do: underwriting, servicing, operations, all inside the enterprise. So what happens when the policyholder is AI-enabled as well?

The information gap is starting to change

The Society of Actuaries Research Institute is now seeking research on policyholder behavior in the artificial intelligence era. According to the proposal request, it is possible for consumers to use AI to understand a policy's provisions better to improve their premiums, loans and withdrawals, and claims through improved product features.

That isn't proof that AI has changed lapse rates, loan behavior, or benefit usage yet. It's on the research agenda. The difference is significant. But the formal asking of such questions by the actuarial community indicates that consumer AI requires attention before the results become apparent in experience studies.

Up until recently, this has created a bit of a predicament: an insurer has product specialists, actuaries, underwriters, policy administration systems, and decades of institutional knowledge; a policyholder has a lengthy contract written in technical language and maybe the occasion to review it a couple of times.

Generative AI won't remove that imbalance entirely, but it certainly lowers the cost of asking sophisticated questions about a complex contract. That alone changes the dynamic of that interaction.

A long-lived contract can meet a very different customer

Life insurance is unique because the contract could go on for decades, and the technology available to the user would be different from the present one.

A policy issued in the age of paper statements and telephone calls may one day be managed by a consumer carrying a persistence layer of AI that summarizes documents, compares alternatives, remembers every previous conversation, and prepares detailed questions before every interaction.

The policy's bones haven't altered one bit. But a policyholder's ability to kick the tires and check them out has.

All this matters because many decisions and actions can occur throughout the lifecycle of a single life insurance policy: premium payment policy, loans, withdrawals, changes of beneficiary, conversion, surrender, and more. By making those decisions easier to understand and execute throughout the life of the product, there's no guarantee that customer behavior in the future will continue to be like the historical behaviors that informed product assumptions.

Policyholder behavior may become an AI issue

SOA's current research request explicitly identifies potential effects on lapses, loans, withdrawals, and benefit usage. A separate 2026 SOA call for essays on AI and the potential for antiselection in life insurance distribution asks whether AI could help applicants or intermediaries anticipate underwriting outcomes, compare carriers, select underwriting pathways and apply sophisticated placement strategies at greater scale.

Once more, these are questions under investigation, not established outcomes. But, one thing appears clear: certain presumptions in life insurance are based on how people behave.

Making it easier to evaluate the timing and consequences of a lapse, loan, withdrawal, conversion, surrender, or benefit election changes behavior in ways that historical behavior doesn't predict. The negative reaction to that realization is to predict material disruption. The right reaction is to begin measuring whether AI-assisted behavior is emerging, where it's concentrated, and if it's big enough to matter.

From a more strategic perspective, for our actuarial and product teams, consumer AI could one day become another behavioral variable to study, rather than just a way to service customers better.

AI-enabled policyholders can also be good for insurers

This should not be framed only as a new risk.

When a customer understands the policy, it allows them to ask the right questions, know what benefits they have, prepare the right documents, and know what decision they are making. Beneficiaries will use AI to organize materials for a claim, point out relevant policy language, and understand what the insurer is asking for. A policyholder who walks up to the window to service their policy will know the difference between a loan, withdrawal, and a surrender.

This is consistent with the financial services research. In July 2026, LIMRA reported that AI is increasingly being studied not only as a tool that assists advisors but as a potential source of analysis and financial guidance directly to consumers.

While more knowledgeable buyers are good for insurers, they also bring heightened expectations. A customer who has just received an instantaneous explanation of a policy provision may be less patient about waiting for days to get a generic response from a service agent.

The customer's AI is not the insurer's AI

Yet there is a boundary—the policyholder's AI is outside of the insurer's control.

The model was not chosen, trained, fed, or even provided with its system prompt by the insurer. Your AI could be reading an outdated contract, skipping an endorsement, not grasping a rider, getting guaranteed vs. non-guaranteed values mixed up, or just making up a provision out of whole cloth that does not exist.

The assertion could be false, or at least inaccurate.

The enterprise response is to make verifying what the authoritative policy information says more accessible. Suppose a customer says, "My AI says this rider gives me a different option." It should be easy for a service person to say, "Let's review what your current contract and its effective endorsements, transaction history, and applicable rules say about this."

The more external AI becomes part of customer decision-making, the more valuable clear, accessible, and authoritative policy context becomes.

Soon, the customer may send the AI instead

The next step is not difficult to imagine: the policyholder does not merely use AI before contacting the insurer. The AI contacts the insurer.

That scenario is beginning to appear in insurance discussions already. An October 2026 Insurance Thought Leadership article described consumer AI agents calling insurers and other businesses on behalf of users. At the same time, NIST is developing work around identity and authorization for software and AI agents as agent-to-agent interactions move from concept toward practical infrastructure.

This brings us to a much bigger question for life insurers: what if a chatbot does work? How would your service channel know that it's talking to software working for a policyholder? What data could be returned through that channel? What happens when a customer AI and your system don't agree on a policy provision? When does it become important to switch the interaction over to a person?

There's nothing in those two scenarios that necessitates banning the use of AI in customer service. It just means that we, as users of these services, will need to create different channels for them to go through.

Build policyholder AI readiness

Insurers do not need to predict exactly how fast consumer agents will mature, but they do need to prepare for the direction of travel.

First, make authoritative policy context easier to retrieve and explain. A policyholder's AI will only be as useful as the documents and information it can access. Clear digital contracts, current riders, transaction history, and understandable definitions reduce the chance that outside AI starts from incomplete context.

Next, test servicing journeys with customers being assisted by AI in mind. When customers arrive having 10 specific questions, they should not follow the same process as someone who has only general questions.

Third, watch how behavior changes. Teams — actuarial, product, servicing — that rely on lapses, loans, withdrawals, conversions, benefit usage, and channel usage need to understand what constitutes a significant deviation. We want to avoid blaming AI if another force is at work. The goal is to build evidence that indicates that AI is a cause before rewriting our assumptions about behavior.

Fourth, prepare digital channels for external agents. Identity, consent, disclosure and escalation will become practical. We do not want to automatically trust an outside agent. We do want to ensure legitimate customer assistance can reach us and not diminish customer protection.

The next AI transformation may happen outside the insurer

When asked about the insurance AI strategy, carriers invariably identify the same things: automated underwriting, automated servicing, automated claims, automated document processing, and automated decision support.

However, when it comes to the impact of artificial intelligence on enhancing service innovation, customer capabilities are also affected.

The increased understanding could allow them to compare things to other products, which could open them up to asking difficult questions.

However, what I think we have not talked about and probably is an elephant in the room is that insurers are not going to be in control of all this.

They can only prepare for it.

Will the next phase of life insurance AI be defined by how intelligent the insurer becomes? Or by what happens when the policyholder becomes intelligent at machine speed?

Paul Carroll, Editor-in-chief of Insurance Thought Leadership.

Pricing gets far less attention than claims or underwriting, yet is important. How is the pricing function being modernized in insurance? And how should it be?

Kumarappan Sundaram, Managing Director, PwC US

Pricing gets less attention partly because a lot of the work is infrastructure: platforms, data flows and deployment processes. It is not a front-office workflow that a new tool can visibly transform.

Modernization can be uneven. Personal lines are well on their way. Commercial lines are earlier in the journey and still rely heavily on underwriter judgment and rating engines embedded in policy administration systems.

How pricing should be modernized can largely be a question of sequence. The foundation comes first: decoupling rating from the policy admin system. At the moment, actuaries produce pricing logic, which they hand to IT, which embeds the logic in a rating engine that sits within a policy admin system. So, making a change to the pricing logic means waiting for a new release of the policy admin system. If you decouple, you can make changes to ratings as and when you want.

Once you’ve decoupled, you should build reliable data flows and establish clear governance.

You can then go to the next level of capabilities, like advanced analytics and AI. Let’s say you have a $10 billion book. You’re considering changing a particular rate factor from X to Y, maybe related to driver age. Advanced analytics can help you understand the impact of the rate change on the whole book. The actuaries can then decide whether to make that rate change. Carriers can help deploy and monitor pricing changes quickly and safely.

Paul Carroll

From a technology standpoint, what are the most important shifts in P&C pricing and rating?

Kumarappan Sundaram

Pricing is moving from a periodic project to a continuous capability. With catastrophe volatility and shifting market conditions, treating pricing as an occasional exercise is hard to sustain.

Two shifts stand out. First, dedicated pricing platforms. Rating, models and rules increasingly sit on purpose-built platforms rather than inside the core policy system. Second, modern software discipline. Version control, automated testing and continual monitoring of live models are becoming standard.

Paul Carroll

What is driving carriers to invest in sophisticated pricing platforms now?

Kumarappan Sundaram

Five things.

First, speed-to-market of rate changes. Moving from quarterly or annual releases, gated by IT and policy admin deployment, to something near-continuous is the key value. Decoupling rating logic from the core policy system is also a practical starting point for core modernization, delivering value without an entire core replacement.

Second, loss cost volatility and trend uncertainty. Social inflation, repair and parts costs, tariffs and secondary-peril catastrophe losses make trend assumptions go stale faster. Carriers want faster feedback from emerging claims experience that they can incorporate into price.

Third, segment-level control. Carriers want to grow where they are well priced and hold the line where they are not, which means testing, deploying and monitoring changes at a granular level.

Fourth, regulatory and governance demands. Filing, documentation and explainability requirements keep growing. Audit trails can be far easier to build in than to retrofit.

Fifth, data ecosystem expansion. Telematics, geospatial and property-level data, cat model output and third-party enrichment should flow into rating at quote speed, which can often be the real integration challenge.

A CFO will often ask how the benefit is measured, so programs should define measurement up front.

Paul Carroll

Underwriting and pricing are tightly linked conceptually, but loosely coupled in practice. Why is that, and can they be brought into phase?

Kumarappan Sundaram

The link is strong in theory but weak in practice, mainly for technology and operating-model reasons.

Start with systems and data. Pricing logic sits in a rating engine or pricing platform, while underwriters work in a different workbench based on the data in the submission, so the two rarely share a view of the risk. Then there are different rhythms. Pricing changes on a release cycle, while underwriting decisions are made submission by submission. Overrides and discounts help fill the gap. Finally, the feedback loop is weak. Why underwriters override a price rarely flows back to pricing in a structured way, so a valuable signal is lost.

Modern pricing platforms increasingly aim to close this gap. Bringing them into phase is mainly an integration and governance task. Pricing can be exposed as a service that the underwriting workbench calls in real time, connected to appetite, distribution, retention and claims signals. Overrides can be captured as data, and authority and guardrails can be made explicit. In complex commercial risks, underwriter judgment legitimately plays a larger role, so the aim there is visibility and learning, not automation.

Paul Carroll

How can modern pricing technology help carriers price emerging risks?

Kumarappan Sundaram

Emerging risks, such as cyber, climate-related perils and new auto and energy technologies, may not come with deep loss history. A modern platform helps in four ways.

First, new data sources as proxies. Outside-in cyber scans, geospatial and property-level data, and vehicle and telematics data give carriers signals where claims history is thin. The platform brings them together with internal experience and subject matter specialist judgment, with clear rules for how the weighting shifts as claims data builds.

Second, scenario testing. Teams can test prices against stress scenarios, such as a major catastrophe or a systemic cyber event.

Third, fast, controlled learning. Small, measured changes show what a new risk is worth sooner and with less exposure than a book-wide change.

Fourth, flexible product design. Configurable rating makes it easier to adjust deductibles, sublimits and endorsements as understanding of the risk develops.

The limit can be that technology cannot create data that does not exist. For novel risks, the platform helps carriers learn and adjust quickly, alongside accumulation controls, reinsurance and underwriting oversight.

Paul Carroll

What impact is AI having on pricing and rating, and where should leaders be skeptical?

Kumarappan Sundaram

I would separate what is emerging with clear value from what is still unproven.

The clearest near-term value is GenAI around the pricing process. The more underrated opportunity is in the work surrounding the price: drafting and reconciling rate filings, translating pricing specifications into platform configuration, documenting changes, summarizing competitor filings and preparing data. These steps slow the path to market, and errors are easier to review than in a pricing decision itself.

What is still unproven is AI acting on pricing decisions. The idea of agents that monitor performance, propose rate actions and test them is compelling. But pricing is regulated and auditable, and many carriers are not ready for autonomous rate changes. For now, I expect AI to propose and people to approve, so the quality of that review is the real control.

As AI capabilities become widely available, proprietary data, speed of deployment and governance are likely to matter more as differentiators. Carriers that treat AI as a standalone upgrade, without the platform and operating model underneath it, risk building capability they cannot put into production.

Paul Carroll

What would you advise a carrier to do first?

Kumarappan Sundaram

Measure your cycle time, from the point a pricing change is approved to the point it is live throughout the various channels. Many carriers have never seen that number, and it shows where value is being lost. Then prioritize platform and process changes that shorten it. Put governance in place before scaling AI, and start with one or two lines where faster pricing has a clear, measurable payoff. Resist modernizing everything at once.

Paul Carroll

Thanks. This is all super helpful.

--

Kumarappan (Kumar) Sundaram is a Managing Director in PwC’s Property and Casualty Operations and Technology practice and a leader of its Insurance Core Transformation and Pricing/Rating offering. With more than two decades of consulting experience, he partners with US and UK P&C carriers, from large international insurers to small regional ones, across personal, commercial, and specialty lines, with depth in commercial and specialty. He has led numerous large-scale programs modernizing policy, claims, and pricing/rating operations, and works with PwC’s strategy, management, technology, and actuarial teams to deliver integrated solutions, thought leadership, and sales.

--

Sponsored by PwC

Collaboration. It's what every leader hopes for in her team. It's the purpose behind much of our communication at work. And it can be a challenge.

There is a collaboration platform in the digital workplace for almost everything. Salesforce. Asana. HubSpot. Figma. Git. ServiceNow. Too many to name them all. Each tracks and assigns work items and enables us to collaborate through a system. The systems sit between us. They enable us to tell one another what to do online. Want something done, open a ticket. Want to feel productive, pound through your task list, and assign "work items" to the next person in the workflow. Fire, then forget.

Most of the time, this works just fine. The efficiency can't be disputed. But ask yourself a couple of questions. Are there times you assign a task in Salesforce and get back a result you didn't expect? Have you ever opened an Asana task and not known what you were being asked to do? Our experience tells us it happens all the time.

Communication scholars and system architects make a distinction between synchronous and asynchronous messages. Synchronous communications occur in real time and involve immediate back and forth between participants. Asynchronous communications allow for time delays. Workflow assignments are asynchronous. They happen independent of an immediate response. Sometimes, in fact, responses don't happen at all. Workflow makes it easy to make an ask of a colleague, assign a task to a subordinate, offer feedback to a collaborator. They also eliminate the synchronicity of a person-to-person "handoff."

Evidence suggests synchronous interaction makes us feel connected at work. It even fosters relationships, which make asynchronous communication on workflow platforms more effective. An intriguing study of software engineers proves it. The engineers worked in two buildings several blocks apart. In some instances, teams worked together in the same building. In others, the team members were spread across the locations. As is common in the software industry, the engineers used a platform to share code, give one another assignments, and provide feedback. Researchers found that teams of engineers working in the same building—that is, colleagues who had frequent synchronous interactions with one another—received 23% more online feedback—an asynchronous interaction—on their code than did those on multi-building teams whose members had less frequent face-to-face contact with one another. "Sitting near workers increases not only in-person but also digital collaboration," the authors write.

The collocated software engineers enjoyed the benefits of social capital: mentorship, coaching, learning, and collaboration. All were achieved through a digital tool that, when coupled with the opportunity for synchronous interaction, worked for and not against social capital. Everyone benefited, especially those newest to the team. Retention among newer joiners was significantly higher among collocated teams than dispersed teams. Young workers and women, the authors tell us, are "substantially more likely to quit the firm when they lose proximity to coworkers." Less feedback. Less connection. Lesser results. It's a lesson the system architects at Amwins, led by Ben Sloop, took to heart when they designed their collaboration platform.

Ben knows his way around a spreadsheet. A boardroom, too. Harvard undergrad. Stanford MBA. A stint as a management consultant at Bain, then a few years in private equity. The perfect pedigree for his current role as president and chief operating officer of Amwins, the nation's largest wholesale insurance broker. But there's more to Ben's success than his stellar resume. Ben told us, "One of the core requirements to gain more responsibility here is having deep, real relationships with our salespeople. I spend a disproportionate amount of my time talking to and staying connected with people in the field. Of all the people that I'll respond to, first is if one of our brokers reaches out to me. That's the top of my inbox."

Amwins' clients are retail insurance brokers and agents looking to find homes for complex, hard-to-place risks—think workers' compensation insurance for circus performers. The insurance carriers who advertise on the Super Bowl don't want the hard stuff. And so, retail brokers work with wholesalers like Amwins to get complex risks placed with one of a large legion of specialty insurance carriers. Each occupies a well-defined niche within the commercial insurance market. Each is expert in a particular type of risk. This creates a classic many-to-many problem. Thousands of customers, thousands of carriers and thousands of Amwins brokers sitting between them.

The challenge is having the right expertise at the right time. "Our brokers have to be specialists," Ben told us. "Our retailers don't want to work with someone that does everything a little bit, because that's kind of what they do. They want to deal with someone that's a deep expert." Enabling his brokers to always find that one expert within their large, decentralized organization became a key part of Amwins' founding mission. It's the hardest part of what they do, Ben told us. "In order to be a deep expert, what that means is that, well, if my client comes to me and they ask me for a construction wrap policy in Scottsdale that I know nothing about, I've got to know where to get them what they need inside our own firm. So, we don't lose that client, right? And so, this created this huge coordination problem."

The company built Amlink to solve it. Amlink is the wholesaler's proprietary relationship management system. It is a workflow platform, but it's different. Ben beamed with a creator's pride as he told us how it works. "I can look in the system and say, well, actually, we have a guy who lives in Scottsdale and does only construction wraps, right? And so, I can reach out to that person who I usually know personally, because we have all sorts of sales events. The point is to build connectivity with an internal coordination mechanism."

We both have seen lots of systems that tried to solve similar problems. All did it the same way. Type in "construction wrap in Scottsdale" and you'd get deal listings, PowerPoint presentations, maybe a white paper or two. Lots of stuff to wade through, maybe some good slides you can reuse, but probably not the answer you need. Amlink is different. It delivers names and contact information and forces conversations to happen. It's effective at both delivering expertise when it's needed and building relationships across the firm. Knowing who to turn to can make or break a deal. Amwins brokers don't just compete on expertise—they compete on connectedness.

Amlink is a unicorn. "I've never seen another system in the industry like this. Everyone says it's a people business. but they tend not to track [data] at the people level," Ben told us. He was clear to put the tech in its proper context: "We are a technology-enabled business. The thoughtful use of technology makes our people better. But we are fundamentally a person-to-person business."

Social capital matters at Amwins. It drives market results. It also creates an environment where people want to stay. "You know, it's funny. I joined 19 years ago, and I still think of myself as one of the newer people," Ben confessed, explaining that Amwins was the kind of place "good people stay for a long time. People typically don't leave just for money. People tend to leave because they just don't feel connected, don't like who they're working with, don't feel compelled by the strategy."

We probably should have introduced you to Ben at the very beginning.

"Social Capital at Work," by John Burrows and Seth Rachlin, will be published by Simon Acumen on Nov. 10. This article is excerpted from their book.

AI is mitigating risks that are already modeled, already priced, and already reserved for. It is creating risks that are none of those things. Cyber loss, claims fraud, catastrophe exposure, and AML false positives all have loss histories behind them. Algorithmic discrimination, model drift, hallucinated advice, and correlated foundation-model failure do not.

That imbalance is why the industry's behavior looks contradictory. Carriers are deploying AI aggressively inside their own operations while moving just as fast to exclude it from their customers' policies. Both are rational. They simply belong on different ledgers.

WHAT AI CREATES

Three new exposures — and no existing control framework built to support them. 

The first is model risk migrating into regulated decisions. Roughly two dozen states and D.C. have adopted NAIC's AI model bulletin, requiring a written AI program, board-level accountability, and discrimination testing. New York's Circular Letter No. 7 goes further, naming fairness metrics and requiring written explanation of adverse decisions within 15 days. Colorado has extended its AI governance regulation to auto and health; nine states now restrict AI in prior authorization and claims review. In the litigation that follows — Huskey v. State Farm, the NH Predict class actions against UnitedHealth — courts have shown no appetite for treating an algorithm as protection. This is model risk with lingering negative consequences and a civil-rights overlay that traditional model risk management never had to price.

The second is third-party AI risk, which is concentration risk wearing a new coat. Swiss Re Institute warned in January that "concentration among cloud and AI service providers creates accumulation and systemic risk concerns"; the Bank of England's July Financial Stability Report modeled correlated disruption through shared providers. Gallagher Re made the insurance point: unlike a windstorm, an AI failure spreads through the shared use of a widely adopted model, producing simultaneous losses across industries and jurisdictions. That is aggregation, and aggregation is how insurers get hurt.

The third is the silent exposure sitting inside policies written before any of this. Gallagher's 2026 benchmarking survey found one in five insurance professionals reporting a client loss or claim tied to AI risk, only 51% of them fully covered. The market is closing that gap by subtraction. ISO's three generative AI exclusionary endorsements — CG 40 47, CG 40 48 and CG 35 08 — took effect Jan. 1, and Berkshire Hathaway, Chubb, Travelers and W.R. Berkley have all filed exclusions. Affirmative capacity exists — Armilla, Chaucer, Mosaic, Munich Re's aiSure — but tops out around $25 million to $50 million per insured. Anthropic settled a single copyright matter for $1.5 billion and paid it off its balance sheet. That is the protection gap in one line.

WHAT AI RETIRES

The mitigation case is strongest where the loss data is best. IBM's 2026 Cost of a Data Breach study, across 602 breached organizations, found firms using security AI and automation extensively cut breach costs by nearly $2 million and contained incidents 65 days faster than those using none. HSBC reports its AI transaction monitoring identifies two to four times more financial crime than the rules-based system it replaced, with 60% fewer false positives — the rare compliance result where accuracy and efficiency moved together.

In claims, Allianz Partners cut its travel claim lifecycle from 19 days to four, and Intact Financial has disclosed more than C$150 million in recurring annual AI benefits. And in catastrophe risk, the gains are peer-reviewed, not self-reported: Google DeepMind's cyclone model, published in Nature in August, delivers three-day track forecasts as accurate as the two-day forecasts that preceded it — a full extra day of warning for every property book on a coastline.

THE MEASUREMENT PROBLEM

Risk managers have registered the shift. In January's Allianz Risk Barometer, AI entered the global top three for the first time, at No. 2, up from No. 10 — the sharpest single-year climb the survey has recorded. But registering a risk and pricing it are different things.

Measurement makes the imbalance worse. Evident's 2026 AI Index found just 65 publicly documented AI use cases with disclosed outcomes across the 30 largest North American and European insurers — 75% productivity gains, 2% revenue uplift. FactSet reviewed seven insurers that discussed AI on second-quarter earnings calls; none disclosed a dollar or basis-point figure tying AI investment to results. IBM's own data adds a caution: even as security AI saved money, average time to contain a breach rose to 247 days, reversing five years of improvement, with one in four malicious breaches now AI-enabled. AI is arming both sides.

That is not an argument against adoption, but for treating AI the way the market eventually treated cyber. The precedent is on the books: Lloyd's Market Bulletin Y5258 forced underwriters, in July 2019, to state affirmatively whether cyber was in or out. No equivalent mandate exists for AI yet, and until one does, exclusions will keep arriving faster than coverage.

That correction may come. If it does not, if the exposure proves too correlated for loss history to tame, then the conversation goes somewhere the industry has been before. Nuclear got Price-Anderson, terror got TRIA, flood got the NFIP, each after the market judged a risk real but not privately bearable. One proposal would apply that logic to frontier AI: a mandatory, risk-priced federal indemnity pool with a sunset. The sharpest objection is that it solves the wrong problem. Ian Adams argues the bottleneck is not capital but legal uncertainty: no one knows what courts will hold AI developers and deployers liable for, and mandating cover over an undefined duty of care subsidizes that uncertainty rather than resolving it. His remedy is clearer statutory duties and safe harbors. Insurers can price a loss once they know what it is.

For risk managers, the work is unglamorous: an inventory of every place AI touches a challengeable decision, a close reading of the AI language in each D&O, E&O, cyber and GL renewal, and vendor contracts written on the assumption that the model will fail. The coverage conversation has started without them.

SOURCES

Allianz Commercial, Allianz Risk Barometer 2026 (Jan. 14, 2026; 3,338 respondents, 97 countries).

NAIC, Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, and NAIC implementation map (as of April 2026).

NY DFS, Insurance Circular Letter No. 7 (July 11, 2024). Colorado DOI, 3 CCR 702-10, as amended effective Oct. 15, 2025.

Huskey v. State Farm Fire & Cas. Co., No. 1:22-cv-07014 (N.D. Ill.); Estate of Lokken v. UnitedHealth Group (D. Minn.).

Swiss Re Institute, sigma insights 01/2026: AI adoption is reshaping the risk landscape (Jan. 13, 2026).

Bank of England, Financial Stability Report (July 7, 2026).

Gallagher Re, Global insurtech Report: Artificial Intelligence — Risks and Opportunities in (Re)insurance and Beyond (March 2026).

Arthur J. Gallagher & Co., 2026 AI Adoption and Risk Benchmarking Survey (Feb. 23, 2026; 1,200+ businesses).

Verisk/ISO endorsements CG 40 47, CG 40 48, CG 35 08, effective Jan. 1, 2026.

IBM Security / Ponemon Institute, Cost of a Data Breach Report 2026 (July 29, 2026; 602 breached organizations).

HSBC, "Harnessing the power of AI to fight financial crime" (June 10, 2024).

Allianz Partners (Sept. 2025); Intact Financial Investor Day (May 21, 2025).

Google DeepMind, WeatherNext cyclone model, Nature (Aug. 6, 2026).

Evident Insights, 2026 Evident AI Index for Insurance (June 16, 2026).

FactSet Insight, "Insurance AI Investments Emerge So Far as Qualitative Comments vs. Quantitative Drivers" (Aug. 14, 2026).

Lloyd's Market Bulletin Y5258 (July 4, 2019).

Gregory C. Allen, The Insurance Industry's Retreat from AI Threatens to Slow Innovation and Adoption, CSIS (Sept. 4, 2026).

Cristian Trout, Insuring Uninsurable Risks from AI: Government as Insurer of Last Resort (arXiv 2409.06672).

Ian Adams, AI Insurance and the Backstop Reflex: A Policy for Every Risk, a Taxpayer for Every Loss, Truth on the Market (Oct. 8, 2026).

Bill Gross recently announced a free database of AI-generated ideas for innovative companies, based on the 31 methodologies that he's developed over the past three decades and that have powered his Idealab incubator. 

The list of ideas — currently totaling a whopping 1,208,580 — may or may not contain an idea that you'll be able to turn into big bucks, but I think that exploring the database and interacting with the ideas is worth a bit of your time. 

Many of the notions I see for the insurance market are provocative. They can at least suggest where there are holes in insurance offerings today that you should work to fill, even if you don't launch the sort of standalone company that Gross would love to see join his pioneering incubator.

Let's have a look. 

Here is a link to the database. And here is a link to a Fortune article that describes the effort in detail, including Gross's argument that innovations driven by AI will create far more jobs than it will eliminate. The article says Gross's prompts to the AI break down into three categories: 

"Opportunities, which he says are things such as solving a business pain point, removing friction, removing waste, removing danger, and improving design, among others; enablers, such as new technologies, scientific advances, and new materials, processes, or changing cost curves that make solutions more feasible than before; and shifts, such as regulatory changes, as well as behavioral, economic, cultural, or geopolitical movements that make the timing for a particular business seem right. "

Gross also fed his AI "some 17 million source documents that included news stories, research reports, expired patents, public company filings, and threads from online community discussion boards" and came up with his 1.2 million ideas for new businesses.

But are the ideas any good?

I certainly found them thought-provoking. 

For instance, I selected the "fintech/insurtech" option, specified that I didn't want to invest more than $250,000, wanted to get to market within a year and hoped to "make the insurance claims process smoother for customers." The top suggestion, which the AI suggested was a 57% fit for me, was for a "claims trust operating system that... gives policyholders visibility into the evidence shaping settlements and repairs."

I'm not sure that's a business. It would take an awful lot of work to assemble the huge amount of settlement and repair data that would be needed to be truly definitive, but the idea raises an intriguing notion: Somebody will surely try to use an AI to give policyholders a quick handle on what the insurer will likely offer after a car accident, water damage to a home, etc. If the insurance industry provides the initial take, it will be able to lower expectations and control payments. If plaintiff attorneys fill the need, they'll be able to raise expectations, influencing policyholders to be disappointed... and perhaps hire an attorney to contest the offer. 

The competing interests would seem to create an opportunity for a trusted intermediary, but I, at least, have become skeptical about whether it could stay unbiased. I subscribe to the notion laid out by Cory Doctorow in a seminal blog post republished by Wired magazine in 2023 that describes the process of — excuse my French — enshittification. 

As I wrote at the time, Amazon is a great example. It initially built a huge base of customers by doing everything possible to make life simple for them. Once it had the customer base, it courted businesses — and the customer experience started to get gunked up by ads, attempts to keep customers from easily finding the best deals, etc. Once Amazon had built a marketplace where companies had to have a presence, it started milking them as much as possible on behalf of investors — a fee here, a fee there, maybe Amazon would even use its insights into sales data to launch products or businesses that would undercut a company with a hit offering on Amazon's platform. (Amazon has consistently denied doing so but has faced many lawsuits alleging it does.) Just about everybody still uses Amazon, but the experience is very different than it was a decade ago.

Even if I'm right that there isn't a business in a "claims trust operating system," the Idealab database has still pointed to a sore spot that the insurance industry needs to address. Somebody is going to use AI to shape policyholders' thinking about what to expect as a settlement, by providing near-instant estimates. So insurers should at least be thinking about how they show their work, and quickly, to keep plaintiff attorneys from setting unrealistic expectations and accelerating the trend toward more and more litigation. 

So I'd recommend that, whatever your role in the insurance ecosystem, you poke around in the database to see what friction points and opportunities it has surfaced that you might be able to address. There are more ideas than I can count about underwriting, pricing, services from agents and brokers, and on and on. 

If you're like me, your eyes glaze over after half an hour or so of an exercise like this, so maybe make a visit to the database an occasional excursion over the next month, perhaps at the end of the day as you're winding down or before a weekly meeting with colleagues so you can stir up discussions. We all get locked into our mindsets and into our day-to-day routines, and Gross has provided a whole host of prompts that can challenge us to do ever more than we are now.

Cheers,

Paul

One of my takeaways from ITC Vegas this fall was that a lot of vendors are playing defense. A lot of carriers are looking at what AI can do now and asking some version of the same question: Hold on a minute, the world changed, so should we just build everything ourselves?

It's a fair question. The barriers to building have come down a lot in the last year, and the benefits of owning your own technology are real. When a convincing demo takes a couple of engineers instead of a sales cycle, the old reasons to buy start to look pretty thin.

I suppose I'm not the most neutral person to ask. I've spent the last decade building and running Clearcover, an AI-native carrier, and now Dearborn Labs, an AI-native software development firm built specifically for insurance; if anyone was going to tell you to build everything, it would probably be me. But when people ask me whether to buy or build, my answer is...yes. We have components of our system that we bought and didn't build, and we're fine with that. We build more than we buy, though, and more every day.

Is "yes" a dodge? Maybe a little. Mostly it's that buy vs. build was never one decision. Every carrier already runs a mix of both, workflow by workflow. And even the build half of my answer comes with an asterisk.

Where should the line sit?

The choices sit along a spectrum, made one workflow at a time, and lately the mix has been shifting toward build. That makes sense to me, and I'd expect it to keep shifting for as long as building keeps getting cheaper.

Buy what's already solved well. If a problem looks roughly the same at every carrier, and someone has spent years getting good at it, there isn't much to gain from building your own version, and the time and money are better spent elsewhere. But that list gets shorter every month.

Build the workflows where being different is the point, like how you triage a new claim or decide which risks to write. Buying those means working the same way as everyone else who bought the same thing, which probably isn't what you were going for.

Where exactly that line falls depends a lot on the carrier and its lines of business, which is why a single company-wide decision on buy vs. build usually ends up wrong somewhere.

The asterisk

What qualifies as building, because you can put a demo together, is usually the smallest portion of the work. AI can change how a claim moves and which decisions get made without a person ever opening the file, but the systems underneath that workflow rarely change, or at least they rarely change quickly. Not everybody is doing a rip-and-replace of their whole company to launch a new AI workflow. A new claims-intake workflow still has to post to your core system and set reserves the way your actuaries expect, and an underwriting workbench still has to read from policy admin and write its decisions back. None of that shows up in a demo.

The interface is mostly a solved problem now, and that's good news. The hard part is the wiring underneath it. Most of the AI conversations I have with carriers wind up being about traditional software engineering. That's the work of fitting whatever the AI application or agent is into the rest of the company, so the systems supporting a reinvented workflow still work properly.

And the wiring is only half of the asterisk.

The other half is running the thing in production until people actually work that way. Most implementation projects are scoped to end at go-live, which is roughly where this half starts. At Clearcover, adoption of the AI tools across our claims team is 92%+. A number like that gets earned after go-live, inside real claims, over however long it takes for the new way of working to become just how the job gets done.

You might reasonably ask whether all of this is an argument for buying something that comes already integrated. For the work that's solved well, it is. But a bought workflow is the vendor's workflow, designed to fit as many carriers as possible, and in the workflows you'd want to run differently from everyone else, that's exactly the problem.

Build realistically.

Before you build a new workflow, ask three questions of it:

1. Foundation — Was this workflow designed around constraints that no longer exist?
2. Fit — Can it work with the systems you already have, without a rip and replace?
3. Engineering — Do your engineers own the integration layer today, and can you test, govern and audit what runs on it?

The first is about not building on top of a bad foundation. A lot of insurance workflows were shaped by limits that made sense at the time, like a handoff that existed because two systems couldn't talk to each other, or a sampling audit that existed because nobody could read every file. If you automate that workflow as it stands, what you get is the old process running a little faster, and you've spent real money to keep a constraint you no longer have.

The second one is more encouraging. You don't have to fully rip out your systems to improve a workflow, so long as you have the software engineering chops to fit the new workflow into the systems you have.

The third is the one I'd push on hardest, mostly because a demo is very good at hiding it. If the answer is no, I'd want to know that before the demo goes in front of the Board rather than after.

If you're asking me whether to buy or build, I'd say build more than you used to. Just go in knowing that the demo is the easy part, and that the wiring, plus the long stretch of running it until people actually work that way, is the job. So where does your line sit this year, and who's doing the wiring?