Download

'SaaSpocalypse' Debate Misses the Real Issue

AI won't replace SaaS platforms, but vendors must prove their value by delivering measurable business outcomes for insurers.

Partnership

There has been much wild speculation about the future of software-as-a-service (SaaS) applications in light of artificial intelligence (AI), with suggestions that a "SaaSpocalypse" will occur due to AI sweeping away the need for core SaaS-based systems. This is not very helpful for insurance companies weighing up their future IT adoption strategies. It is clear that AI tools will be valuable, but customers need a calm, rational discussion about how best to implement the technology in enterprise IT settings. If SaaS vendors want to prove they still matter, they must demonstrate that SaaS is not just relevant but is the indispensable foundation of every successful IT transformation.

Cloud-native applications are the right foundations to embrace AI

The crucial questions customers need to address when deciding how best to transform their IT systems are not whether they should replace their SaaS applications with AI, but what is the best transformation strategy to minimize disruption and provide measurable returns. Having extensive contextual understanding of the specialist workflows and language that different applications use means SaaS vendors are better placed to ensure AI tools work effectively for insurers. The leading AI vendors understand this point, because they know building enterprise software for mission-critical applications is hard and not something an AI agent can unpick overnight.

However, this does not mean SaaS vendors can rest easy. I would agree with Insight Partners: "Trust, user stickiness, and existing ROI may provide some breathing room for cloud-natives to build new AI features that cement value, but they have to hustle to keep that position."

The key phrase here is "cloud native." Applications built for the cloud are better placed to help customers embrace AI, because this approach requires interoperability between systems. This means AI tools can access a single view of all the data across an organization, which is critical if insurers are to have confidence in the decisions these tools make.

Effective AI adoption requires a deep understanding of the insurance industry

That said, being able to integrate AI more easily is not the main reason customers turn to SaaS vendors instead of going to the AI tools providers. Customers are worried about the squeeze on margins due to current economic conditions, and the unprecedented unpredictability in markets around the world. They only care that their business applications help them address these issues. SaaS vendors must show they understand this driver for IT transformation and be prepared to take risks to prove their commitment.

This requires a dramatic rethink of how value is delivered. The agentic AI model switches the focus to achieving business outcomes rather than simply processing transactions. The technology can proactively interpret objectives, break them into tasks, make decisions within defined boundaries and monitor progress.

This resets how users interact with applications. Consequently, SaaS vendors must leverage their specialist industry knowledge to demonstrate they can deliver outcomes insurers want.

It is somewhat ironic that these vendors, who themselves a mere 15 to 20 years ago were seen as the big upstarts, must prove their worth in the face of new technologies. The winners will be those who seize the opportunity, are ambitious and don't just try to defend their existing position. This is likely to be one of the most exciting chapters in the evolution of enterprise software so buckle up and enjoy the ride!


Mike Ettling

Profile picture for user MikeEttling

Mike Ettling

Mike Ettling is executive chairman of Sapiens.

He is also the executive chairman of SYSPRO, a provider of digital manufacturing technology that was recently acquired by Advent. Previously, he was CEO of Unit4, a global cloud ERP software vendor, formerly owned by Advent, and president of SAP SuccessFactors, a provider of cloud HCM solutions. 

He holds a bachelor’s degree in accounting from the University of the Witwatersrand and a masters in finance from London Business School.  

Best Practices for AI Coding in Insurance

Insurers can safely cut development costs by 75% if they treat AI-generated code like a junior engineer's work.

AI Coding

 

AI Coding Quote

A lot of that anxiety comes from so-called "vibe coding." Vibe coding means using AI to build software by merely describing requirements, without worrying much about structure, testing, or long-term maintenance. Vibe coding can feel reckless (because it is). As if "moving fast and breaking things" at human speed wasn't scary enough; now we can break things with lightspeed automation. Use vibe code in production? No way.

AI coding has clear potential for productivity gain, but only if insurers address two obstacles:

  • Obstacle 1: AI-generated code needs to be safe enough for production use
  • Obstacle 2: AI-generated code must work with existing systems

Without addressing these issues, AI-generated code is limited to standalone prototyping. That does indeed have value, but the real opportunity comes from moving past the two obstacles. 

Over time, we've developed practical ways to make AI-generated code production-ready.

Solving Obstacle 1: Making AI-Generated Code Safe for Production
AI Can Generate Code Image

An LLM can write a research report for me in seconds, but I would never publish the result as is. Code is no different. AI-generated code should be treated like code written by a very junior engineer. It should never go directly into production. Frankly, no code—whether written by a human or AI—should go directly into production!

Here are some best practices for making it safe.

  • Engineer Code Review: Experienced engineers need to review AI-generated code, as they would review a junior engineer's code before putting it into production.
  • Unit Tests: Unit tests are highly valuable, but it's dangerous to let AI generate these. If the AI misunderstands your requirements and generates a bug, it'll probably misunderstand your requirements for the test too. The AI is likely to write unit tests that actually pass only if the bug is there!
  • Integration Tests: Integration tests validate the contracts between components. Most production failures occur here, not inside isolated functions. AI often models these complex interactions too simply: over-mocking dependencies, testing only happy paths, and ignoring real failure modes. Integration tests require judgment about how the systems are intended to collaborate. AI can generate integration tests, but you need to review the tests thoroughly and hand-write any test cases that AI missed.
  • End-to-End Tests: I have some good news: here is a place where AI works great! Use AI to specify tests for overall user-level workflows, which will only pass if the entire system is working as expected.
  • Static and Dynamic Code Analysis: All the tools that review engineers' code quality can also review AI-generated code. There are many popular tools such as SonarQube, CodeQL, and OWASP ZAP. Choose some good ones and use them.
  • Code-Generation Tool Selection: The best AI tool to use is a matter of opinion and will likely change faster than your browser can refresh this page! At the moment, popular code-generation tools include Codex from ChatGPT, Claude Code from Anthropic, and Cursor. Keep watch on the latest developments and be prepared to change when something better comes. Most importantly, only use AI tools approved by your security team.
  • Security Team Collaboration: Every insurer has a security team. All developers should work closely with their security teams to keep up with AI's rapidly evolving capabilities and vulnerabilities.
Solving Obstacle 2: Integrating AI-Generated Code with Other Systems

We've now covered best practices for high-velocity, production-ready AI-generated code. That's great if it's all you need, but in the enterprise, code doesn't operate in isolation. Most insurance software (certainly most insurance core software) wasn't engineered with AI in mind, and that can drastically reduce the benefits of AI for the enterprise.

Here are the things that make software compatible with AI-generated code. They are a must-have list when selecting enterprise software vendors.

  • Look For Modular Design

    Remember when I said AI-generated code should be treated like code written by a very junior engineer? If you give a junior engineer the keys to your whole code base, you can expect an intractable amount of code reviewing before you can ship it. They need guardrails, and so does AI.

    This is why modularity (with well-defined contracts!) is highly important. If the architecture is divided into well-defined plugins, configurations, and integrations, and all the connection points use open standards that are well documented, you can give AI these small components and reasonably review and test each one.

  • Look For Open Languages and Formats

    Code-generating AIs are trained on all mainstream programming languages and file formats. They all know Java, Python, and JavaScript. They also know JSON, CSV, and RESTful APIs. Unfortunately for insurers, a lot of insurance core platforms have invented proprietary languages and file formats, which no LLMs are trained on. Insurers have great difficulty trying to use AI-generated code around these systems.

  • Look For Documentation

    Whether it be APIs, configuration syntax, or system architectures, engineers hate it when they have to ask vendors for information that should be provided in documentation. Whereas humans have the privilege of calling support or emailing other engineers, AI gets stuck.

    Insurers today need to look at their vendors' documentation with a very critical eye. In the past, poor documentation was acceptable when supplemented with weeks of training and continuing access to experts. This model is annoying with human developers, but it totally breaks with AI-generated code.

  • Look For Data-Fluent Systems

    From report generation to business intelligence to data lake integrations, many use cases for AI-generated code deal with data. Your AI-generated software will be no better with data than the enterprise software it relies on.

    • Data fluency for an enterprise software means:
    • Strong APIs for accessing individual records
    • Included data lake for mass queries
    • Webhooks and delta file exports for keeping external systems updated in real time
    • High-speed server responses for all data retrieval operations
    • High uptime so data is always available
    • Real-time data consistency across the system

    If your enterprise platform doesn't have strong APIs, then your AI will struggle to write code to interact with it. If it can't support mass queries, then your AI can't generate reports for you. If the system is slow, the code your AI generates will be slow. If the system has frequent downtime…you get the idea.

    If the flow of data around your enterprise is too complicated and asynchronous that your own engineers struggle to add new capabilities, then your AI-generated code will struggle too.

  • Look For MCP Servers

    Model Context Protocol (MCP) is currently the most popular standard for connecting AI with other software platforms. A modern enterprise software platform has user interfaces for human interaction, APIs for external software interaction, and MCP servers for external AI interaction.

    This is admittedly off-topic when evaluating enterprise software's ability to integrate well with AI-generated code, but any enterprise software must be considered AI-compatible. With an MCP server, the latest LLMs require no code at all to connect with enterprise platforms. This is why MCP is supported by such enterprise software giants as Salesforce, Snowflake, Atlassian, and HubSpot, among others.

Conclusion

AI-assisted development is here, and it's already too powerful for insurers to ignore. Like any powerful tool, it can be immensely valuable or immensely dangerous. Insurers today can safely realize 75% reductions in their development costs and a 2x increase in their IT velocity if they use best practices and work with AI-compatible platforms.

As AI continues its rapid pace of development, smart platform decisions today will amplify into massive future advantages.


Dan Woods

Profile picture for user DanWoods

Dan Woods

Dan Woods is the founder and CEO of Socotra, an AI-enabled insurance core platform. 

Previously, Woods was an engineer at Palantir, where he composed its first AI functionality, led partnerships and ran several deployments. 

He earned a master’s degree in computer science from Stanford University. 

Legal System Abuse Costs Billions

Legal system abuse costs American families nearly $6,000 annually through rising nuclear verdicts and third-party litigation funding.

Legal System Abuse

If you've ever wondered who pays for those highway billboards promising life-changing payouts for your injury lawsuit, the answer is: We all do.

Those billboards are largely funded by plaintiff firms manipulating the legal system to prioritize profit over justice. And based on the return on investment, there's no reason to believe we won't continue to pay for those garish signs through increasing costs for everyday products and services.

Businesses are deemed by the plaintiffs' bar to have deep pockets of wealth and are able to afford to settle multimillion-dollar and multibillion-dollar lawsuits. However, those costs are passed on to consumers in the form of rising prices for goods and services. The American Tort Reform Association found that lawsuit abuse costs every American $1,424 annually—nearly $6,000 per year for a family of four.

Legal system abuse is not new. Terms like "social inflation" and "nuclear verdict" have been around for years, and both continue to play a role in the broader realm of legal system abuse.

In this article, we will take a deeper dive into what constitutes legal system abuse and explore the rise in nuclear verdicts and the effect of third-party litigation funding in civil suits. We'll explain how policymakers and the insurance industry are working to curb the abuse and, finally, what businesses can do to protect themselves and effect change.

The role of nuclear verdicts

Legal system abuse is defined as the misuse of courts and legal procedures to gain a strategic, financial, or tactical advantage. It has become a growing problem, driving up costs, delaying justice, and eroding public trust in the civil justice system.

One of the defining characteristics of legal system abuse is the rise in nuclear verdicts, which are defined as jury verdicts exceeding $10 million.

Nuclear verdicts often include punitive damages that far exceed a plaintiff's actual economic losses. In addition to the financial effect on defendants, they can raise settlement expectations and contribute to rising insurance premiums.

Nuclear verdicts can also have a significant effect on the public's perception of fair and proportional damages in civil litigation. These awards go beyond what most legal experts consider rational compensation for the harm suffered. They are driven by emotional appeals, aggressive litigation tactics and expanding theories of liability. Here are two examples of nuclear verdicts:

  • Juries have issued massive punitive awards against a multinational pharmaceutical, biotechnology and medical technology company in cases that allege its talc-based products caused cancer. Some verdicts have exceeded $1 billion, and earlier cases produced punitive damages surpassing $4 billion before reductions on appeal.
  • In 2019, an auto-liability case in Nassau County, N.Y., resulted in a jury award of $100 million to the victim's parents for pain and suffering and $900 million in punitive damages against the defendant, a trucking company.

The number of corporate nuclear verdicts rose to 135 in 2024, a 52% increase over 2023 numbers, according to a report, "Corporate Verdicts Go Thermonuclear 2025 Edition," by research firm Marathon Strategies. The total sum of these verdicts reached $31.3 billion, a 116% increase over 2023. "Thermonuclear verdicts" of $100 million or more increased to 49 in 2024, with five of those cases resulting in verdicts greater than $1 billion. A decade ago, these verdicts would have been considered extreme outliers.

Tactics of the plaintiffs' bar

Legal system abuse can occur when attorneys exploit procedural rules, court structures or litigation mechanisms to gain the upper hand in civil litigation.

Common tactics of legal system abuse demonstrate a shift from using the courts as a forum for justice to using them as leverage for financial gain. They include:

  • Frivolous claims and lawsuits: Frivolous litigation is one of the most recognizable forms of legal system abuse. These cases burden courts, increase defense costs and pressure defendants—especially small businesses—to settle rather than endure prolonged litigation.
  • Abuse of class-action procedures: Class actions are intended to efficiently resolve widespread harm, but they can be manipulated. Some filings prioritize attorney fees over meaningful relief for plaintiffs or rely on expansive class definitions to inflate potential damages. This can coerce defendants into large settlements regardless of the underlying merits.
  • Predatory litigation funding: Third-party litigation funding has expanded rapidly, allowing outside investors to finance lawsuits in exchange for a share of the recovery. While sometimes beneficial, it can also encourage speculative or unnecessarily prolonged litigation.
  • Manipulation of courtroom practices, discovery, and evidence: Discovery abuse—such as excessive document requests, refusal to produce information or strategic withholding—can dramatically increase litigation costs. Parties may use discovery not to uncover facts but to pressure opponents into settlement by making the process prohibitively expensive. Jury anchoring is another tactic allowed in some jurisdictions where an attorney introduces a higher numerical value to jurors to rely, or "anchor," on that figure as a reference point when assessing potential damages in a lawsuit. Numerous studies have shown the strong effect this practice has on jury award decisions.

Of these tactics, the recent emergence of third-party litigation funding potentially has the most effect. It has become one of the most influential—and controversial—forces shaping civil litigation. Its rapid growth raises concerns about transparency, fairness and escalating legal costs.

Litigation funding

Third-party litigation funding (TPLF) is the practice in which outside investors finance a lawsuit in exchange for a share of any settlement or judgment. These funders generally have no direct claim in the underlying dispute but use litigation as an investment vehicle. Analysts increasingly identify TPLF as a driver of social inflation, contributing to rising claims costs and larger verdicts because funders profit only when payouts increase.

TPLF operates through several models. In single-case funding, a financier backs one lawsuit, typically one with high potential damages. Portfolio funding involves financing multiple cases at once, spreading risk across a broader set of claims. The key parties include the funder, who supplies capital; the plaintiff, who receives financial support; and the attorneys, who may coordinate with funders on litigation strategy.

While proponents argue that TPLF expands access to justice, critics warn that it can provide incentives for frivolous or overly aggressive litigation. Because funders profit only from large recoveries, they may encourage plaintiffs and attorneys to pursue riskier strategies, prolong litigation or reject reasonable settlements. Reports also highlight that the industry operates with minimal transparency, often without disclosure to courts or opposing parties. This secrecy raises ethical concerns about who is influencing litigation decisions and whether funders exert control over case strategy.

Tort reform

Legislative and policy responses to legal system abuse have grown significantly in recent years, with states adopting new tort reform measures, transparency rules and procedural changes aimed at curbing excessive litigation and rising verdicts. These reforms, however, face political resistance and uneven implementation across jurisdictions.

Efforts to address legal system abuse often begin with limits on punitive damages, which several states have enacted to curb unpredictable and disproportionate jury awards.

A wave of state-level tort reform has also emerged. Georgia, for example, enacted sweeping reforms in 2025 through Senate Bills 68 and 69, introducing new procedural rules, damages limitations and updates to trial practices. Similar initiatives in other states aim to streamline litigation, reduce forum shopping (the strategic practice where a plaintiff chooses to file a lawsuit in a specific court or jurisdiction that is most likely to provide a favorable outcome) and promote fairness in civil proceedings.

Another major development is the push for transparency in third-party litigation funding, requiring greater disclosure of funding providers and financial interests in lawsuits. Between 2023 and 2025, eight state legislatures, including Georgia and Louisiana, have made the contents of TPLF contracts subject to automatic discovery or upon request.

These reforms face significant opposition from trial lawyers and advocacy groups, who argue that caps on damages and procedural restrictions limit access to justice for injured individuals.

Potential solutions to the problem of legal system abuse include federal legislation to create uniform standards for litigation funding disclosure, punitive damages and class-action procedures. Congress is currently examining legislation that would introduce transparency when TPLF agreements are present in a suit before a federal court or bar foreign actors from participating in federal TPLF arrangements. Nationwide rules could reduce inconsistencies and prevent forum shopping.

The role of insurance

The insurance industry has become one of the most active voices in confronting legal system abuse, responding with risk-management strategies, advocacy efforts and collaborative initiatives aimed at reducing inflated claims costs and restoring balance to the civil justice system.

Insurers have adopted risk-management practices to stem the rising costs associated with legal system abuse. These practices include enhanced claims monitoring, early case assessment and the use of analytics to identify patterns of excessive litigation. Industry research shows that legal system abuse has significantly inflated claims payouts across multiple lines of insurance, with commercial auto insurers alone paying $20 billion more than expected between 2010 and 2019 due to litigation.

Many insurers advocate for tort reform, supporting legislative efforts to curb abusive litigation tactics and reduce the frequency and severity of nuclear verdicts. Insurance industry groups such as the American Property Casualty Insurance Association and the Insurance Information Institute have led the charge.

What can you do?

Putting an end to legal system abuse will require coordinated efforts from individuals, businesses, policymakers and the broader public. With tort costs reaching hundreds of billions annually and driving up prices for consumers, every stakeholder has a role to play in promoting fairness and transparency. And consumers and businesses have the most influential voices.

Companies can reduce exposure to abusive litigation by strengthening compliance programs, improving documentation and using early dispute-resolution strategies. Businesses can also actively support state and national efforts to curb abusive tactics.

Working together, insurers, consumers, businesses and policymakers can help restore balance to the civil justice system, reduce unnecessary costs and ensure that the legal process works in the service of justice instead of profit.

The Disclosure Problem No Compliance Framework Catches

Leaders skilled at calibrating disclosure for different stakeholders can inadvertently omit critical information when building support for high-stakes initiatives.

Art

A senior leader has been in their role for 15 years. They know how the steering committee reads bad news. They know that if they lead with the complication, the conversation shifts from "how do we solve this?" to "should we proceed at all?," and this initiative is too important, too close, to lose to a room that hasn't seen what they've seen.

So, they frame the complication as a minor implementation consideration. It's real. It's manageable. They're protecting the initiative. They're advocating for the people this change is meant to serve.

Six months later, that minor implementation consideration is the reason three functions can't use the system.

The leader didn't lie. They translated. And the translation left something out.

The Skill That Made You Good at This Job

Every effective leader in a complex organization learns, early and through experience, that the same initiative requires different conversations with different stakeholders. The CFO needs a business case. The front-line manager needs to know what changes about how they do their job on Monday. The steering committee needs enough confidence to approve. The actuary needs the model assumptions. The board needs strategic context, not operational detail. And each stakeholder comes to the table with different expertise, different risk tolerance, and different decision authority.

This is not manipulation. It is sophisticated communication. I call it calibrated disclosure: the practiced judgment of what to say, to whom, in what form, at what moment. In insurance and financial services, calibrated disclosure is more than a skill. It is a professional requirement.

Part of an initiative leader's role in this sector is building the confidence required to move change forward in an environment where every proposed modification will be examined under a microscope by internal functions, by regulators, and in some cases by rating agencies.

And then there is the problem.

The same discipline that makes the briefing cleaner makes any omission invisible. The skill that builds confidence can, without anyone deciding it should, begin to protect the initiative rather than serve it. This shift does not announce itself. It operates below the level of ethical self-examination. You don't decide to calibrate away a risk. You calibrate because you've always calibrated. Because it works.

Where the Coalition Compromise Lives

This year, I developed a framework for change leaders on the ethical traps that become available under pressure. The one most relevant to this industry is what I call the "coalition compromise": the trap that activates when building stakeholder support begins to require managing information differently across audiences, not just for clarity but for outcomes.

Three structural features of regulated industries make a coalition compromise more available, more normalized, and harder to catch.

Disclosure is a professional discipline. In insurance, calibration is not optional. It is governed. Actuarial standards specify how risk models are presented to different audiences. Regulatory filings require precise framing for specific agencies. Leaders who have navigated all of this for years develop a deep, practiced confidence in their own judgment about what belongs in which conversation. There is an additional psychological dimension worth naming: When the calibration feels like it is following the rules, when a leader genuinely believes they are operating within the standards set by compliance and the regulators, the ethical self-check rarely fires. The leader is doing the right thing. They have the documentation. They have the process. And they are wrong in a way the process cannot catch.

Stakeholder complexity is structurally high. A single AI initiative at a large carrier may require navigating underwriting, claims, actuarial, compliance, legal, IT, distribution, and the board, each with different languages, expertise, risk tolerances, and decision criteria. In markets with distributed regulatory oversight, the complexity multiplies: A product requiring filing across dozens of jurisdictions generates separate stakeholder conversations at each level. The temptation to frame the initiative differently for each audience is not laziness. In the short term, it works.

The initiative window is often compressed. Regulatory timelines, competitive pressure, and board cycles create genuine urgency. Urgency compresses the ethical review. The question — "Am I framing this differently because it's clearer, or because it speeds movement?" — doesn't get asked when the steering committee meets in four days and the filing deadline is the following week. The coalition compromise doesn't require a decision to compromise. It requires only the absence of time to examine what you're doing.

The Line That Moves

The coalition compromise doesn't arrive as a single choice. It moves through stages. What makes it difficult to catch in regulated environments is that the early stages are explicitly trained.

Stage 1 — Translation. You simplify a complex finding for an audience that doesn't need the technical detail. A valid actuarial concern becomes an implementation consideration.

Stage 2 — Emphasis. You lead with the upside and position the complication toward the end of the document. The information is present. The weighting is a choice.

Stage 3 — Selective inclusion. The complication is in the record, available to anyone who asks. Steering committees are time-compressed environments where significant decisions get made amid competing "day job" responsibilities. Key members may not be present. Questions don't always get asked. The meeting moves on.

Stage 4 — Omission. The finding doesn't appear in the materials for this audience. It may have been raised briefly in a prior meeting where it wasn't picked up. In the room where the decision is made, it isn't there.

What makes this particularly difficult in regulated environments: Stages 1 and 2 are competencies organizations develop deliberately. The line between Stage 2 and Stage 3 is invisible in the moment. Stage 4 is reached without a single decision that felt like a decision.

What No Compliance Framework Catches

Compliance frameworks, including the three-lines-of-defense model that governs risk oversight across most large carriers, are built for decisions: for specific acts that can be governed, audited, and reviewed against a documented standard. The coalition compromise operates below the level of decision. It lives in framing choices, in the weighting of a slide, in which objections are raised before the meeting, and which are handled in the hallway afterward. None of those frameworks are designed to catch it.

There is a second dimension that rarely gets discussed: organizational contagion. When a transformation leader calibrates information selectively, they do not do it in isolation. The briefing gets built with a team. Materials get shaped by people who observe what their leader includes and what they don't, what gets said in the room and what gets managed around the edges. Those observations form behavioral norms: what is acceptable here, how we handle complications when the stakes are high.

The coalition compromise, left unexamined, doesn't stay in the steering committee room. It permeates. Product requirements to IT get shaped by the same logic. Customer-facing materials, sales training, and claims handling documentation are each an opportunity for the same framing choices to compound. What started as one leader's judgment call becomes the organization's operating standard. 

I have seen this dynamic inside a large financial services organization where siloed functions, low inter-unit trust, and a prevailing culture of risk aversion made it easy for misunderstandings to surface at the point of delivery to the customer, and for finger-pointing to follow when issues finally became visible.

What would catch the problem isn't a framework. It's a question built as a habit before every steering committee, every board presentation, every stakeholder briefing:

What does this audience not know that they would want to know if they did?

That question is not a compliance requirement. It is a capability. And most organizations have never built it deliberately.

Building the Discipline Before You Need It

The coalition compromise is most likely to emerge precisely when the initiative is most important, the pressure is highest, and the leader's conviction is strongest. Those are not the conditions under which most people do their best ethical reasoning. Which is why the work must happen before you're in that room.

Three disciplines, developed deliberately:

Name your translation standard before the briefing. Before preparing materials for a specific audience, define in writing what "translation" means versus what "omission" means for this conversation. What is the threshold — whether the finding, the risk, the complication — below which simplification is legitimate and above which it must be present regardless of how it lands? The act of writing forces the distinction. Calibrated disclosure that has been examined is a professional skill. Calibrated disclosure that hasn't is a liability waiting to surface.

I have worked with leadership teams who approach this not as a compliance exercise but as a personal commitment, drafting explicit principles about what they will and will not do when the pressure is high, before the pressure arrives. The ones who find it most useful are not the ones who already have strong ethical instincts. They are the ones who understand that under sufficient pressure, strong instincts alone are not enough. You're not writing rules for other people. You're building your own code.

Apply the adversarial read before sending. Before any significant briefing goes out, ask: if someone who actively opposed this initiative reviewed this document, what would they say is missing? If the answer is something that would change the audience's decision if they knew it, it belongs in the document. This is not about playing defense. It is about the difference between advocacy and selective presentation.

Build one truth-telling relationship. The coalition compromise flourishes in the space between what a leader knows and what the room is allowed to see. One relationship, built specifically for this purpose, changes that structure: a peer, a trusted direct report, a board member whose explicit role is to tell you what you don't want to hear before you've told the room something incomplete. This relationship is worth more than any compliance protocol, because it operates at the level where the coalition compromise actually lives — in the judgment calls made before the materials are finished.

The Capability the Industry Has — and the One It Needs

Insurance has built deep institutional capability around disclosure. That capability is a genuine competitive advantage, and the precise environment in which the Coalition Compromise becomes structurally available, culturally normalized, and professionally indistinguishable from good judgment.

Until the moment it isn't.

The leaders who navigate this well over time are not more ethical. They are more deliberate. They have built the habit of asking, before the brief is finalized, before the room fills up, what this audience doesn't know that it would want to know.

If the answer is "nothing," then proceed.

If the answer takes more than a moment to arrive, that pause is worth honoring. It is the one place a framework can't go that a leader can.


Amy Radin

Profile picture for user AmyRadin

Amy Radin

Amy Radin is a strategic advisor, keynote speaker, and Columbia University lecturer focused on why transformation succeeds or stalls in large, complex organizations. 

Drawing on senior leadership roles at Citi, American Express, and AXA, including one of the world’s first corporate chief innovation officer roles, she helps leaders build the capabilities required to absorb, scale, and sustain change.

Learn more at amyradin.com.

 

How Health Plans Can Control Pharmacy Risk

Rising specialty drug costs and regulatory pressures are pushing health plans toward marketplace-based pharmacy benefit models.

Expensive Pharmaceuticals

For health plan officers and financial leaders, pharmacy benefits have crossed a threshold. U.S. prescription drug spending surged to $915 billion in 2025 and is projected to exceed $1 trillion in 2026 — one of the fastest growth rates in two decades. Pharmacy benefits is no longer a cost category — it has become a material driver of financial volatility.

The forces driving this exposure are converging. Specialty medications now account for roughly half of total drug spending, with many therapies exceeding $100,000 annually per patient, and some cell and gene therapies reaching into the millions. Overall healthcare spending is projected to grow 10% or more, with pharmacy costs among the primary drivers. At the same time, the passage of the Consolidated Appropriations Act of 2026 (CAA 2026) has expanded transparency, reporting and fiduciary requirements for health plans — adding regulatory and operational complexity on top of financial exposure.

In one-on-one conversations, I am hearing an acute reaction to mounting pressure being felt across the payer ecosystem. Plan sponsors are increasingly focused on how to maintain cost control while also meeting expanding fiduciary and transparency obligations. Health plans and TPAs are actively exploring more proactive approaches to medical and specialty pharmacy management as usage and cost volatility increase. At the same time, PBMs are working to redefine their role as traditional operating models face greater scrutiny and disruption.

The legacy PBM model was not designed to manage challenges at this scale. It was designed to aggregate volume. The result is a system that is increasingly difficult to measure, interpret, and control, precisely when greater visibility, accountability, and agility are most needed.

The Legacy PBM Model Concentrates Risk in the Wrong Places

The legacy PBM structure systematically limits visibility. Opaque pricing methodologies, hidden rebate structures, and aggregate-only reporting make it difficult for health plans to determine true net cost, validate savings claims, or identify the actual drivers of spending early enough to intervene effectively. Addressing this requires a more transparent operating framework built around integrated data visibility, component-level financial reporting, and greater accountability across pharmacy benefit stakeholders.

Flexibility is equally constrained. Bundled PBM arrangements limit a health plan's ability to adjust benefit design, introduce new strategies, or respond quickly when specialty usage shifts unexpectedly. Fragmented vendors and siloed data eliminate the possibility of a single, reliable source of truth, which is essential for effective financial and operational oversight. More adaptable, modular benefit structures allow plans to introduce targeted specialty management strategies, optimize site-of-care programs, and implement new financial controls without overhauling the entire pharmacy benefit model.

Market concentration compounds the challenge further. The top three PBMs control roughly 80% of U.S. prescription claims, limiting competitive alternatives and slowing adoption of models better suited to the current environment. The practical consequence is a widening gap between what health plans are accountable for and what their current structure actually allows them to manage.

To close these gaps, stakeholders are looking for new models that provide access to broader contracting scale, specialty networks, and integrated management capabilities to maintain greater control over benefit strategy, vendor selection, analytics, and member experience.

A Marketplace Model Approach to Pharmacy Benefits Strategy

Managing pharmacy benefit risk requires the same discipline applied to any significant financial exposure: visibility into the drivers, flexibility to respond, and the ability to make decisions based on empirical data rather than aggregated summaries.

A marketplace-based approach to pharmacy benefits addresses each of these directly. Rather than relying on a single bundled arrangement, health plans can evaluate pharmacy benefit components individually — assessing performance, replacing underperforming elements and aligning each component with specific organizational priorities and financial objectives. The result transforms a monolithic, difficult-to-audit arrangement into a set of discrete, measurable risk factors that can be actively managed and adjusted as market conditions evolve.

The implications are real:

Concentration risk is reduced. Dependence on a single PBM relationship — with its inherent opacity and limited leverage — is replaced by a diversified model in which no single vendor controls the full picture of pharmacy economics. A marketplace model broadens choice and competition, enabling organizations to manage risk based on the profile of their covered lives, not that of the masses.

Regulatory and fiduciary exposure are addressed directly. A marketplace model built on transparent, component-level reporting supports the disclosure, audit trail, and accountability requirements that CAA 2026 and ERISA-aligned fiduciary standards now demand. Plans can demonstrate not just what they spent, but why, and what oversight was applied. Furthermore, real-time access to integrated data enhances the power of predictive modeling and the potential for proactive risk mitigation.

Specialty drug volatility becomes more manageable. A small number of high-cost therapies now account for a disproportionate and increasingly unpredictable share of total spending. A marketplace model makes it possible to coordinate key cost savings components — site-of-care optimization, stronger clinical oversight, and more targeted financial management strategies — within a unified framework designed to intervene where costs are most concentrated. For example, plans may redirect eligible specialty infusions from high-cost hospital outpatient settings to lower-cost ambulatory infusion centers, implement enhanced monitoring protocols for emerging high-cost therapies, or better align pharmacy and medical benefit management strategies to more effectively identify, track and control specialty spending across the continuum of care.

Operational risk from rigidity is eliminated. Currently, when the prescription drug market changes — new therapies enter the market, GLP-1 usage accelerates, or a new cell and gene therapy creates an unanticipated claim event — payers can be slow to react, often having to renegotiate a bundled contract or wait for a vendor partner to build a new capability. A marketplace model enables organizations to act in real time, rather than reactively, as these dynamics unfold. This flexibility enables organizations to rapidly deploy targeted utilization management programs, introduce condition-specific clinical management strategies, expand specialty network access, or adjust financial controls in response to changing market conditions.

From Reactive to Proactive: Data as a Strategic Management Tool

Effective pharmacy benefit management is anticipatory, not retrospective. The legacy pharmacy benefits model is built around periodic, aggregated reporting — a structure designed for billing, not strategic decision-making. A marketplace approach changes the information architecture fundamentally.

With integrated, component-level data and real-time reporting, health plans gain the ability to identify cost drivers earlier, compare vendor performance against benchmarks, and make empirically grounded decisions rather than relying on assumptions embedded in aggregate summaries. This shifts pharmacy benefit management from a reactive posture — responding to spending that has already occurred — to a proactive one, where risk is identified and addressed before it becomes a financial event. Unified analytics also help align pharmacy and medical benefit insights, improving visibility into total specialty care costs and supporting more coordinated enterprise-wide decision-making.

This capability is especially critical as the pace of change in the drug market accelerates. Traditional tools — formularies, rebates, utilization controls — remain important, but they are no longer sufficient as standalone management mechanisms. Precise, real-time visibility into cost drivers and direct operational control over how those costs are managed have become baseline requirements for any health plan operating in today's pharmacy environment.

A Sustainable Framework Requires Structural Change

The structure of pharmacy benefits will continue to evolve as drug acquisition costs remain elevated, new high-cost therapies enter the market, and regulatory expectations increase. Health plans that continue to rely on the legacy model are operating within a framework that often limits visibility and flexibility at the exact moment both are most needed.

Organizations with greater expertise, visibility, and flexibility will be better positioned to adapt to ongoing market disruption, manage specialty cost pressures more effectively, and meet rising stakeholder expectations. Those that fail to evolve risk operating within a pharmacy benefit framework that is increasingly misaligned with the financial, regulatory, and operational realities of today's healthcare market.

An adaptable, marketplace-driven pharmacy benefit model enables health plans and payers to adapt to these realities without major structural disruption or service interruption. It replaces opacity with transparency, rigidity with modularity, and assumption-based management with data-driven oversight.

Executing this kind of transition requires more than a vendor swap — it requires a new paradigm and a strategic partner with the clinical, financial, and operational depth to navigate this environment. Collaborating with specialized pharmacy benefit partners — those with proven capabilities across specialty management, analytics, rebate optimization, and benefit design — can help plans build a more responsive and sustainable pharmacy benefit strategy.

AI's Insights Into Customer Sentiment

AI-powered tools now capture real-time behavioral signals from claims interactions, transforming employee experience data into measurable business performance.

AI in the Workplace

Although insurance is a data-first business, it has been traditionally tough to quantify the human side of operations: the quality of a claims adjuster's interaction with a policyholder in crisis, the patterns that predict when a top-performing call center agent is about to disengage, the signals that distinguish a renewal conversation that's going well from one that's creating retention risk.

AI is quickly closing this gap. A new generation of AI-powered tools is giving insurance organizations real-time visibility into behaviors that determine outcomes. The carriers moving fastest on this are redesigning how employee experience data flows through the organization and, in doing so, creating a direct line between workforce intelligence and business performance.

The Data Gap in Insurance Operations

A field adjuster handling catastrophe response may conduct 12 to 15 policyholder visits in a single day, and each one can determine whether a customer walks away feeling supported or abandoned. A call center agent in a high-volume property claims environment may handle 80 inbound contacts in a shift, each requiring rapid judgment calls about coverage, tone, and escalation.

Historically, the data generated by those interactions has been gathered by call recordings, satisfaction surveys, and observation/field rides. However, these are challenging to scale and to measure which behaviors drive impact across the workforce.

The tools available today — AI-powered conversation analysis, real-time behavioral signal detection, field mobility platforms, IoT-enabled workspace intelligence — connect not just data, but patterns, behaviors, and activities that can predict and redirect outcomes.

What the Tools Actually Do

The technology stack for frontline workforce intelligence in insurance has matured significantly. There are four primary capture mechanisms now deployed at scale across financial services and insurance operations.

  1. Real-Time Conversation Analysis

    Ambient voice AI and natural language processing engines transcribe, classify, and analyze spoken interactions as they happen. These systems go beyond transcription by detecting sentiment in both the employee and the customer, flagging compliance-relevant language, identifying missed resolution opportunities, and surfacing coaching prompts — during the call or conversation.

    A system that detects elevated customer distress during a first notice of loss call can prompt the adjuster with an empathetic acknowledgment script, which can enhance the interaction in real time.

  2. Behavioral Signal Detection Beyond Words

    The most advanced conversation AI systems capture language, analyze voice and behavioral signals (think pace, hesitation, tonal shifts, speech rate under stress), and build a real-time picture of both the employee's and customer's emotional trajectory.

    These signals, aggregated across thousands of interactions, also reveal patterns that pure text analyses miss. Which claims generate the most agent stress? Which coverage explanations correlate with downstream complaints? Which interaction profiles predict a renewal at risk? The answers sit in the behavioral data that exist in insurance operations but now can be systematically mined.

  3. Field Intelligence for Road-Based Adjusters

    The data capture challenge for field adjusters is fundamentally different from the call center environment. Their day consists of unscripted property assessments, policyholder conversations, and documentation tasks conducted in varying conditions under time pressure.

    Mobile-first field platforms now give adjusters AI-assisted documentation tools. These include photo capture with damage classification, voice-to-text for on-site notes, and real-time coverage guidance based on policy type and loss description. These can reduce administrative burden and improve documentation quality simultaneously. Importantly, they also generate experience data: how long assessments take, where documentation quality varies, which loss types generate the most adjuster rework, and why.

  4. Workspace and Interaction Pattern Intelligence

    In fixed-location operations like service centers, claims hubs, and regional offices, occupancy sensors, spatial analytics, and interaction pattern data provide a layer of intelligence that traditional workforce management tools cannot. They identify whether certain workspace configurations are creating service bottlenecks, whether team placement is affecting collaboration patterns, and whether high-performing agents cluster in specific environmental conditions that could be replicated broadly.

    For a carrier managing a large claims service center, this kind of environmental intelligence can surface insights like which team with the highest attrition sits in a zone with the highest ambient noise and the lowest natural light, and which agents who access shared knowledge bases most frequently produce the fewest coverage disputes.

What Better Looks Like For Employee and Customer Experiences

Claims Quality and Cycle Time

When adjusters receive real-time documentation guidance and coverage support in the field, documentation quality improves and rework decreases. With in-the-moment coaching on empathy and resolution framing, first-call resolution rates increase, and post-interaction complaint rates fall.

Agent Development and Retention

The traditional model for developing insurance frontline talent is episodic — via classroom training, annual reviews, and occasional supervisor observation. An agent receiving specific, behavioral, in-context feedback across hundreds of interactions develops faster and retains more than one who receives a quarterly performance review.

Customer Experience Outcomes

The EX-CX link in insurance is not theoretical. Adjusters who are well-supported, well-coached, and experience manageable cognitive load produce different customer interactions than those who are overwhelmed and under-resourced. Employee experience quality is a leading indicator of customer experience quality, not a lagging one.

Reducing Burnout and Retention

The employee experience dimension here is direct. Adjusters consistently report that administrative burden — and not the complexity of the work itself — is the primary driver of burnout. Tools that absorb that burden and do so in a way that improves, rather than complicates, the field workflow produce measurable improvements in adjuster retention and output quality simultaneously.

What Deployment Actually Requires

What determines whether a carrier captures this opportunity is how the implementation is designed.

Three elements separate the deployments that deliver from the ones that stall.

  1. Tools must be introduced with clear communication about their purpose. Define specifically what the AI is optimizing for, how data will be used, and what employees can expect from the experience.
  2. Managers must be equipped to use the data the tools generate in ways that support development. The most valuable output of an AI workforce intelligence platform is the conversation it enables between a team lead and an agent.
  3. Metrics used to evaluate the deployment must include employee experience indicators alongside operational ones. If the only measures of success are call handle time and claims cycle time, the deployment will be optimized for those measures, and the experience improvements that make the gains sustainable will be invisible until attrition numbers force the conversation a year later.
The Window Is Now Open

The tools are available. Technology has been proven in adjacent industries. The workforce data that has been sitting uncaptured in every claims call, every field visit, and every agent interaction for decades is finally capturable at scale — and the infrastructure to turn it into real-time organizational intelligence is no longer experimental.

The question for insurance leaders is not whether this shift is coming. It is whether their organizations will be the ones that shape it.


Tamar Cohen

Profile picture for user TamarCohen

Tamar Cohen

Tamar Cohen is the founder of HaloEffect, a people strategy firm built on the belief that traditional “employee engagement” models no longer work in an AI-driven world. 

She has spent more than two decades leading global EX and CX programs across financial services, healthcare, and technology. She is a frequent keynote speaker and executive facilitator. 

The Emerging Threat to Cybersecurity

Liz Kim, president, US, at BOXX Insurance, says invoice manipulation has developed into a major threat -- and explains how not to fall victim. 

An Interview with Liz Kim

Paul Carroll

To start us off, what is your overall outlook for the cyber insurance industry?

Liz Kim

If only we could predict where it's going to go, right? I've been in the cyber insurance industry for many years and I've worked across a lot of different roles. I've been in it as a lawyer, as head of claims for a major insurer, in underwriting, in product development, and as a broker. We’re in a soft market now, but, as with all insurance, the cyber market is cyclical.

I think cyber is more cyclical than other lines, for a couple of reasons. First, although there are plenty of disaster scenarios that people talk about in the industry — things like a worldwide AWS outage — we haven't yet really had a true disaster, which would tighten capacity and increase prices. Second, we have new entrants coming into cyber all the time. Because of that constant influx, a lot of their value proposition comes down to nothing more than having the lowest prices. That dynamic drives pricing down more than you'd typically see in other lines of business.

That said, companies that maintain underwriting discipline and pair their insurance with meaningful services or technology solutions to reduce digital risks are better positioned to hold pricing than those that are purely insurance plays.

My overall outlook? It's always positive — because if it wasn't, I wouldn't still be in cyber after all these years. With the kinds of innovations we're seeing across the industry, it's always going to be an area that drives the market forward.

Paul Carroll

How has the cyber threat landscape evolved over the past year or so, particularly with the rise of AI capabilities being leveraged by bad actors?

Liz Kim

The biggest concern I'm hearing right now with respect to AI, in particular, is deepfakes. It's not necessarily a claims-oriented concern in the way ransomware or even extortion is — it's more of an existential threat, which is how most people seem to view it. But deepfakes do connect to insurance, in that some cyber insurers have AI-related exclusions that could potentially leave a deepfake-related claim uncovered. I don't think we've landed on a perfect solution there yet.

We're also seeing a rise in invoice fraud, especially through business email compromise and invoice manipulation. Deepfakes and other trickier social engineering techniques that are leveraging AI certainly make those types of scams even easier to pull off.

Paul Carroll

Tell me a bit more about invoice manipulation, if you would. It’s been around for a while, but I find it interesting that it’s picking up.

Liz Kim

What you see is threat actors creating a fake invoice from a vendor that the insured already does business with. They'll know somehow that you work with that vendor — maybe through social engineering, or maybe they've actually gained access to your systems. From there, they'll send you a fraudulent invoice using one of two main methods.

The first is domain spoofing, where the email domain looks very, very similar to the legitimate one — maybe just a misspelling, or .com versus .ca or .co. They're counting on the recipient not noticing.

The second is BEC, or business email compromise, where the hacker has actually compromised the email account of the other party and is sending invoices directly from it to legitimate contacts.

Either way, there's generally a payment redirection involved — like, "Oh, I've changed my banking information." That's a key red flag. There's no tech hack to prevent this; it's really a process issue. Accounting teams need to have protocols around anything involving a change to payment details. Verify it through a phone call to a known contact using confirmed contact information. And you need to be the one calling them, not the other way around.

You can even see cases where someone on the accounts payable team receives a fake email thread or a WhatsApp message from their CEO or CFO saying, "Pay this invoice."

What makes it convincing is that the thieves have done their homework: they know the CEO is connected to that person professionally on LinkedIn, or they've worked together, or they're tagged in a post. The thieves replicate real, believable scenarios to trick employees.

Junior employees often don't want to question something that looks like it's coming from the CEO. But that's exactly the point: You cannot bypass your internal validation protocols at any level, no matter how legitimate something looks. Because threat actors are trying to trick you.

Paul Carroll

I gather AI is also increasing the volume of cyber threats like phishing. What is the industry doing to counter that?

Liz Kim

Yes, the volume is going up. AI is an enhancement to all of the tactics that bad actors are already using.

A phishing email from three, four, or five years ago is not the same as a phishing email today. Before, there were very obvious markers — the spelling mistakes, the type of email. Nowadays, with advanced social engineering, they can make an email seem so much more legitimate. It's hyper-personalized.

At the same time, the prevention aspects are improving, too. The market is providing more education around building digital resilience and awareness of how phishing scams work: Don't click on things from someone you don't know, and don't click on something you weren't expecting — even from someone you do business with every day.

As an industry, we're not seeing the increase in claims volume that you would expect given the sophistication of AI, and that's because there's much more emphasis on education and prevention.

Paul Carroll

I've been tracking hackers since probably the late 1980s, back when friendly Nigerian princes used to offer me a lot of money. Now I'm at the point where I may report emails as phishing scams that turn out to be legitimate — I'm suspicious of everything these days.

Beyond education, what are you and others in the industry doing to prevent cyber threats from succeeding?

Liz Kim

We have our in-house technology experts — the BOXX Hackbusters team — who are on call 24/7. In addition, each commercial policy is bundled with Cyberboxx Assist, a suite of tools and services designed to help individuals and businesses predict, prevent, and respond to cyber threats. We also offer a virtual CISO [chief information security officer]. These services are focused on the SME space, because many of our insureds don't have any in-house technology expertise, and we can help them bridge that gap.

By offering services like our virtual CISO, it does two things. One, it raises awareness among the management team that cyber risk is a real issue that can impact their organization. Two, it allows our insureds to get expert-level cybersecurity advice. Our vCISO will work with them on a plan — and offer them advice and resources on how to execute it.

In terms of insurance offerings, BOXX introduced a tech E&O policy earlier this year. It has a really strong cyber focus and offers something very specific for technology companies: breach of contract coverage. To get a little lawyerly, the liability that tech companies face isn't a breach of industry standard, which is what you normally have for more traditional professionals like accountants, lawyers, doctors, and so on. Breach of industry standard is the coverage traditional professional errors and omissions insurance provides. But in the technology space, liability is driven by what companies put into their contracts. So the importance of having affirmative coverage for contracts really cannot be overstated when it comes to Tech E&O.

Paul Carroll

There seems to be growing emphasis not just on an organization's own cybersecurity, but on the security of all the vendors and business partners they interact with, because those relationships can serve as entry points for attackers. How is that third-party risk being addressed in the insurance market?

Liz Kim

We can't physically address the security of our insureds' vendors or business partners. So we have to essentially push our insureds to address the security of their own vendors and business partners through education and awareness.

That's critical, because as an industry, we're seeing more claims come in through those third-party relationships than through direct attacks on the insureds themselves.

Paul Carroll

Looking ahead two or three years, where do you think the cyber landscape is headed — and does it ever really change, or is it always going to be that back-and-forth battle between attackers and defenders?

Liz Kim

Even though cyber is an area that changes a lot, it also stays the same in a lot of ways. The role of cyber insurers — especially technology-focused ones like BOXX — is to stay ahead of what the bad guys are doing. That's why things like monitoring the dark web, which is still a relatively new practice for the industry, are so important.

The focus will shift, sure. We've seen it move from ransomware to extortion to business email compromise and invoice manipulation. But ultimately what we're doing remains the same: predicting, preventing, and insuring against negative cyber and technology-related events.

Paul Carroll

Are you seeing much change in the geographic origin of cyber threats? For a while, North Korea, China, and Eastern Europe were significant sources — has that landscape shifted?

Liz Kim

The industry is seeing what appears to be a decrease in volume from Russia and Ukraine, simply because their energies are focused on each other rather than on the rest of the world.

That said, I don't think we're seeing any decrease from the other traditional sources, such as China and North Korea.

Paul Carroll

This has been very helpful. I hope people get the message, especially about the vigilance needed to head off invoice manipulation.

Any final message?

Liz Kim

I’ll just note that I’m excited to have joined BOXX. Just in the few weeks I’ve been here, I’ve found the people to be amazing. The other thing that really drew me to BOXX — and that I've come to appreciate even more since joining — is that we have all the excitement and focus of a startup and we’re channeling that energy into growth after officially becoming a part of Zurich Insurance, a 150-year-old, well-established, and well-respected insurance company.

Paul Carroll

Thanks, Liz.

 

About Liz Kim

Elizabeth (“Liz”) Kim is the President, US at BOXX Insurance, where she leads the company’s U.S. strategy, operations and market expansion. With nearly 30 years of experience in the insurance industry, she brings deep expertise across underwriting, claims, product development, reinsurance and legal analysis. 

Before joining BOXX, Liz served as a Cyber Reinsurance Broker at Gallagher Re, structuring reinsurance solutions for cyber insurers, InsurTechs, and MGAs. She also held senior leadership roles at Hiscox in underwriting and product development after spending more than a decade as a litigator and claims leader managing complex technology, media, cyber, and professional liability matters. Her broad background gives her a practical, well rounded perspective on emerging risks and the operational needs of insurers and insureds. 

Liz holds a J.D. magna cum laude from Seattle University School of Law, an M.S.W. with honors from California State University Sacramento, and a B.A. in Sociology with a Minor in Women’s Studies from the University of California Davis. She is also committed to mentorship and community service and has volunteered with San Francisco’s Volunteer Legal Services Program, the Asian American Bar Association/Asian Pacific Islander Legal Outreach Pro Bono Clinic, and the ACLU of Washington. Liz also served on the Board of Directors of the Korean American Bar Association of Northern California and continues to serve on the Board of its affiliated Foundation, where she has chaired the Scholarship Committee and mentors law students.[


Insurance Thought Leadership

Profile picture for user Insurance Thought Leadership

Insurance Thought Leadership

Insurance Thought Leadership (ITL) delivers engaging, informative articles from our global network of thought leaders and decision makers. Their insights are transforming the insurance and risk management marketplace through knowledge sharing, big ideas on a wide variety of topics, and lessons learned through real-life applications of innovative technology.

We also connect our network of authors and readers in ways that help them uncover opportunities and that lead to innovation and strategic advantage.

What Insurance Leaders Get Wrong About Culture

Insurance leaders often mistake compliance for culture. Here are five crucial steps that will let you build the sort of agile culture that AI requires.

Leaders

When people think about a healthy culture in most insurance organizations, they usually say the same things: People follow the process, leaders avoid trouble, and long-tenured employees are promoted. Leaders track metrics like compliance, regrettable attrition and engagement scores to determine success.

But this definition is incomplete because it hides what's beneath the surface: behavioral norms fragmenting across business units and new hires leaving within 18 months because the operating environment is significantly different from what was described during recruitment. The people who stay do so because of inertia, not because the work feels meaningful.

Behavior design matters now more than ever. Cultures built on adherence and compliance are based on a set of rules that are assumed to stay static. But when disruptions occur, the organization is not prepared, and confusion and bad decisions are made.

Culture Redefined as Adherence

In high-compliance industries, culture and process compliance tend to conflate. When regulatory requirements, audit cycles, and procedural accuracy define how work gets evaluated, organizations start measuring culture through proxies like tenure, compliance rates, and survey scores. These operational measures are important, but they describe stability rather than resiliency.

The result is a culture that performs well on paper and fragments in times of instability. This distinction is becoming costly in ways that are hard to ignore. Gallup's 2026 State of the Global Workplace report found that only 20% of employees worldwide are engaged at work (the lowest scores ever), with disengagement costing the global economy an estimated $8.9 trillion annually in lost productivity.

AI Is Making the Problem Visible

When AI shifts what the work actually requires, carriers that have spent years measuring culture through stability metrics cannot answer basic questions: what does good behavior look like to support and sustain change in the organization and the industry overall? How do we sustain and manage our teams through all the changes? How do we ensure confusion isn't jeopardizing team performance?

We need to start moving away from dependence on easy-to-measure metrics, most of which are lagging, and tell us how people worked in the past, during a moment in time. What companies need to start thinking about is how to ensure their teams can learn and adapt to new technologies, new processes, and new requirements. How will their employees learn, develop, and grow in the new AI world? Answering these questions requires the much harder work of defining culture at the behavioral level, specifically enough that people can act on it even when circumstances shift.

Five Steps to Building a Culture That Is Specific Enough to Work

Behaviors don't just change by mandate. They require deep thought into understanding what outcomes the company is looking to create, and then designing the behaviors required to achieve those outcomes. Getting clear and actionable provides transparency and, finally, building recognition and accountability systems around those specific behaviors guides future alignment. Here is how to do it:

Step 1: Get Clear on the Problem and the Desired Outcome

Before naming behaviors, leaders need to name the gap. Is the issue that collaboration is breaking down at the team level? Does accountability diffuse when decisions cross business units? Are new hires not connecting to the real operating and cultural norms fast enough? The behavior work only holds if it is designed to close a specific gap that is clearly articulated. This is where value statements can't hold up.

Step 2: Get Clear on the Dos and Don'ts

Get team members into a room and via digital whiteboard (like Miro or FigJam) or sticky notes, ask them to independently write down:

  • Green Behaviors (Dos): Actions you want to see more of (e.g., "turn on cameras during internal video calls," "respond to messages within 24 hours").
  • Red Behaviors (Don'ts): Actions to avoid (e.g., "canceling 1-on-1s without 24 hours notice," "speaking over colleagues in meetings").

Step 3: Write a One-Sentence Definition

Condense the list into clear, actionable statements and avoid buzzwords as much as possible. If the definition requires interpretation, it is not finished yet. Build a simple test: read it to someone two levels below you and ask them what they would do Monday morning. If the answer matches what you had in mind, the definition works.

Step 4: Apply the Street Test

The street test asks: Would someone off the street be able to easily recognize the stated behaviors in play? Behaviors are exhibited and reflected in the day-to-day actions of employees. It should be in the feedback of customers or vendors. Behaviors that fail the street test are aspirational language that do not change anything.

Step 5: Model It and Reward It

Defined behaviors only become cultural norms when leaders model them visibly and when the recognition system tells the stories of the people doing them. In insurance, where performance management tends to be tightly tied to technical output metrics, this requires a deliberate decision to build behavioral recognition into performance conversations, team meetings, and promotion criteria. What gets recognized gets repeated.

The Real Competitive Risk

Culture defined as adherence will hold together as long as the rules stay stable. The insurance industry is entering a period where the rules, the work, the tools, and the talent expectations are all shifting simultaneously. Carriers that have not done the behavioral definition work have a structural gap that a survey will not capture.

The organizations that will retain the best people and navigate transformation most effectively will be the ones where culture is specific enough to guide behavior when the process is no longer clear. That requires doing the unglamorous work of naming what good actually looks like, one behavior at a time.

Engagement scores will tell you how people feel today. Behavioral culture design determines how they act tomorrow.


Tamar Cohen

Profile picture for user TamarCohen

Tamar Cohen

Tamar Cohen is the founder of HaloEffect, a people strategy firm built on the belief that traditional “employee engagement” models no longer work in an AI-driven world. 

She has spent more than two decades leading global EX and CX programs across financial services, healthcare, and technology. She is a frequent keynote speaker and executive facilitator. 

3 Key Questions for Boards on AI

Boards approving AI deployments often miss critical questions about data foundations, decision documentation, and pipeline positioning.

Questions about AI

In the past two years, I have sat in on more board-level reviews of AI deployment proposals than in the previous 10 combined. The dollar amounts have gone up, as have the strategic stakes. What has not gone up enough is the quality of the questions board members ask before approving these proposals.

Most board reviews follow a predictable script. Management presents a use case, a vendor, a timeline, and an ROI model. Board members ask about the vendor's track record, the integration risk, and the projected savings. The proposal gets approved or sent back for revisions on the basis of those answers.

Those questions are not wrong. They are insufficient.

Across 20 years of building insurance software and watching deployments succeed or fail, I've seen three other questions matter more for the long-term success of an AI decision-making deployment. Most board members are not asking them yet. The ones who do are the ones whose organizations end up with AI in production rather than stuck in pilot.

Question 1: What is the data foundation we are building this on?

Every AI deployment runs on data the organization already has. The proposal will say so. What the proposal usually does not say is whether that data is in the shape the model needs.

In my experience working with carriers, the data foundation question surfaces three different problems depending on the organization. Sometimes the data exists but is fragmented across systems of record that cannot be reconciled to a single customer or transaction. Sometimes the data is reconciled but tagged inconsistently across business units. Sometimes the data is consistent but does not include the variables the model actually needs to make a defensible decision.

Each of these is a deployment-stopping problem. None of them is visible from the vendor demo. All of them surface in the first 60 days of integration when the model starts producing outputs that the business cannot use.

When my team published research from 20 US insurance leaders earlier this month, data quality and fragmented systems came in as the single biggest blocker to AI deployment. ROI uncertainty came second. Internal IT capacity came third. The order surprised some readers. It did not surprise me.

Board members who ask "what is our data foundation" are not slowing the project. They are giving it a chance to succeed.

Question 2: How do we document an adverse decision?

Most AI deployments at the decision-making layer touch outcomes that affect customers. In insurance, that means claim denials, payout reductions, and coverage decisions. In banking, that means lending denials and credit limit reductions.

For each of those, the regulatory environment in 2026 expects you to be able to document why the AI system reached the decision it did, what input features it considered, what model version was active at the time, and what human review was involved. The expectation is not theoretical. The NAIC Model Bulletin on AI in insurance, adopted in 24 U.S. states as of early 2026, makes this explicit. The EU AI Act makes this explicit. Federal regulators in banking and healthcare are moving the same direction.

In recent research, only three of 14 carriers with AI in production were confident they could produce a complete decision audit trail within five business days if a state regulator asked. The other 11 were less confident. Some far less.

The board question is not whether the AI deployment will be regulated. It will be. The board question is whether the deployment as proposed has the documentation discipline built in, or whether it will require an expensive retrofit when a regulator first asks.

Question 3: Does the AI touch the actual decision?

This is the question most likely to change the structure of the deployment itself.

In my experience, AI deployments succeed faster and create more measurable value when they sit on the edges of a decision-making pipeline rather than at the decision itself. The reasons are operational and regulatory. Capabilities that handle structured but tedious work, such as document interpretation, intake triage, and fraud signal generation, scale faster because the data they consume is more structured, the output they produce is easier to measure, and the regulatory scrutiny they attract is lower.

Capabilities that touch the decision itself, such as reserve recommendation in claims, credit decisioning in lending, and care recommendation in healthcare, scale more slowly. The data is messier. The measurement is harder. The regulatory exposure is higher.

This does not mean decision-making AI is wrong. It means decision-making AI should generally not be the first deployment. The board approving a first AI deployment that sits at the decision itself, before the organization has built the data foundation and the documentation discipline on lower-stakes capabilities, is approving the most complex deployment under the worst conditions.

In that same research, of the 20 U.S. insurance leaders surveyed, just three described their organizations as running AI-assisted decision making. Seventeen still relied on humans for every meaningful decision. The 13 with any AI in production had deployed it on the edges of the pipeline, not at the decision.

A different conversation

When boards ask these three questions, the conversation about an AI deployment changes. The vendor selection becomes secondary. The integration timeline becomes secondary. The ROI model becomes secondary, because all three depend on getting the data foundation, the documentation discipline, and the pipeline position right first.

The boards that ask these questions get more AI deployments into production. The boards that do not ,get more pilots that never move beyond pilot.


Piotr Biedacha

Profile picture for user PiotrBiedacha

Piotr Biedacha

Piotr Biedacha is the CEO and head of delivery at Decerto. 

A graduate of software engineering and postgraduate management studies, he has been working in the insurance industry for over 20 years. 


 

Trends in AI-Powered Underwriting in P&C

AI is transforming P&C underwriting with faster decisions and precision pricing, but data quality and regulatory hurdles remain critical obstacles.

Underwriting

Underwriting has always been at the heart of property and casualty (P&C) insurance, helping insurers assess risks, determine premiums, and make informed policy decisions. Traditionally, underwriting relied heavily on manual assessments, historical data, and rule-based decision making. However, the insurance industry is now experiencing a major shift with the rise of artificial intelligence (AI)-powered underwriting.

As insurers face increasing customer expectations, growing competition, evolving risks, and pressure to improve profitability, AI is becoming a strategic enabler for smarter underwriting decisions. By leveraging predictive analytics, machine learning, real-time data, and automation, insurers can improve risk assessment accuracy while accelerating policy approvals.

As we move into 2026, AI-powered underwriting is expected to play an even larger role in transforming P&C insurance operations. But what benefits does it offer, what challenges remain, and what emerging trends should insurers watch?

1. What is AI-Powered Underwriting in P&C Insurance?

AI-powered underwriting refers to the use of artificial intelligence, machine learning, predictive analytics, and automation technologies to evaluate insurance risks and improve underwriting decisions.

In traditional underwriting, professionals assess applications manually using limited structured data such as claim history, demographics, and historical risk factors. This process can be time-consuming and often lacks personalization.

AI-powered underwriting, however, analyzes vast amounts of structured and unstructured data in real time, enabling insurers to make faster and more accurate decisions.

Examples of data sources used in AI underwriting include:

  • Historical claims data
  • Geospatial and weather information
  • Vehicle telematics data
  • IoT-enabled property monitoring systems
  • Social and behavioral insights
  • Third-party risk databases

By analyzing these datasets, AI helps insurers better predict risks and price policies more effectively.

For P&C insurers, this means moving beyond traditional risk evaluation toward intelligent and dynamic underwriting models.

2. Key Benefits of AI-Powered Underwriting in P&C Insurance

AI is helping insurers modernize underwriting processes while improving operational efficiency and customer satisfaction.

Faster Risk Assessment and Policy Approvals

Traditional underwriting can take days or even weeks, especially for complex cases. AI significantly accelerates risk evaluations by analyzing data instantly and automating repetitive tasks.

This allows insurers to issue policies faster and reduce turnaround times, improving customer experience.

Improved Risk Accuracy

AI models continuously learn from historical and real-time data, helping insurers make more precise underwriting decisions.

For example, insurers can better predict property risks using weather analytics or assess driving behavior through telematics data in auto insurance.

Personalized Insurance Pricing

AI enables more customized policy pricing based on individual risk profiles instead of broad risk categories.

This creates fairer premiums and improves customer retention by offering tailored coverage options.

Reduced Underwriting Costs

Automation minimizes administrative workloads, enabling underwriters to focus on high-value and complex risk evaluations.

Insurers can improve productivity while reducing operational expenses.

Better Fraud Prevention

AI can identify suspicious application patterns and anomalies early in the underwriting process, reducing fraud exposure before policy issuance.

3. Challenges of Implementing AI in Insurance Underwriting

Despite its advantages, AI adoption in underwriting comes with several challenges that insurers must address.

Data Quality and Availability

AI systems depend on high-quality, accurate data. Incomplete, inconsistent, or outdated information can lead to poor underwriting decisions.

Insurers often struggle with fragmented legacy systems and siloed datasets.

Regulatory Compliance and Transparency

Insurance is a highly regulated industry, and underwriting decisions must remain explainable.

Regulators increasingly expect insurers to demonstrate how AI-driven decisions are made, especially when policy approvals or pricing decisions are involved.

Bias and Ethical Risks

If AI models are trained on biased historical data, they may unintentionally discriminate against certain customer groups.

Responsible AI governance and fairness monitoring are becoming essential priorities for insurers.

Legacy Technology Limitations

Many P&C insurers still operate on older core systems, making AI integration difficult and expensive.

Successful AI implementation often requires modernization of policy administration systems and cloud-based infrastructure.

Human Resistance to Change

Some underwriting professionals may view AI as a replacement threat rather than an enhancement tool. Effective change management and employee training are necessary to ensure successful adoption.

4. Emerging AI Underwriting Trends in P&C Insurance for 2026

The future of underwriting is evolving rapidly, and several AI-driven trends are expected to shape P&C insurance in 2026.

Predictive Underwriting Models

Insurers are increasingly using predictive analytics to anticipate risks before losses occur.

For example, climate risk modeling can help insurers assess wildfire or flood risks more accurately.

Real-Time Underwriting

AI is enabling real-time policy approvals by instantly evaluating customer data and risk signals.

This is especially valuable for digital-first insurance experiences.

Generative AI for Underwriters

Generative AI is expected to support underwriting teams by summarizing risk reports, drafting recommendations, and simplifying documentation.

Instead of replacing underwriters, AI will act as an intelligent assistant.

Hyper-Personalized Policies

Using behavioral and IoT data, insurers can create highly personalized policies tailored to customer lifestyles and risk behaviors.

Embedded AI Decision Support

AI-powered underwriting tools will increasingly integrate with platforms such as Guidewire to provide smarter risk scoring and decision support.

These trends are expected to improve underwriting precision while enhancing profitability for insurers.

5. Why Human Expertise Still Matters in AI-Powered Underwriting

While AI offers speed and intelligence, human expertise remains essential in underwriting.

Complex commercial risks, high-value policies, and unusual scenarios often require professional judgment that AI cannot fully replicate.

Experienced underwriters provide:

  • Strategic decision making
  • Risk interpretation
  • Regulatory oversight
  • Customer relationship management
  • Exception handling for complex cases

The future of underwriting is likely to be collaborative, where AI supports underwriters rather than replacing them.

Insurers investing in AI-driven underwriting must also ensure seamless technology integration, platform reliability, and strong testing frameworks. This is where insurance technology partners like FECUND Software Services can help through Guidewire expertise, QA testing, cloud modernization, and digital transformation support tailored to the insurance industry.

Conclusion

AI-powered underwriting is rapidly transforming P&C insurance by enabling faster decisions, more accurate risk assessments, personalized pricing, and improved operational efficiency.

However, challenges related to compliance, bias, data quality, and legacy systems still require careful attention. As insurers move into 2026, the most successful organizations will likely combine AI-powered intelligence with human underwriting expertise to create smarter, fairer, and more efficient underwriting processes.

The future of underwriting is not simply automated—it is intelligent, collaborative, and data-driven. Insurers that embrace AI strategically today will be better positioned to compete in tomorrow's increasingly digital insurance landscape.