From Data Silos to Shared Insights: Transforming AI Risk Management

By: Lucy Pilko | August 6, 2025

Lucy Pilko leads AXA XL’s insurance business throughout the US, Canada, and Bermuda. She joined AXA XL in October 2023 from Boston Consulting Group (BCG). She was a Managing Director & Senior Partner in the New York office and led BCG’s North American Insurance Practice. Lucy brings over 20 years of consulting experience with particular focus on growth strategy, operating model and transformation for Insurance carriers and brokerage. She has worked closely with leaders from nearly all of the region’s major carriers and broking houses, helping them solve problems and uncover opportunities. Prior to BCG, Lucy was employed at Marsh McLennan and Bain Capital. Lucy has a BS in Industrial Engineering from Stanford University, and an MBA with High Distinction from the Harvard Business School, where she was recognized as a Baker Scholar.

As artificial intelligence (AI) transforms industries, it brings both remarkable opportunities and significant risks—algorithmic bias, systemic failures in automated decision-making, data breaches, and deepfake fraud, to name a few. And some risks remain off our radar, waiting to emerge.

There are tremendous unknowns with AI. We’re learning as we go. But imagine if we got together and compared notes. Shared the data that we keep close. Collaborated more.

Why AI Risk Management Needs a Collaborative Approach

To manage AI risks effectively, that’s what needs to happen. Insurers, technology companies, and businesses need to change how they work together, enabling us to collectively understand and mitigate the tremendous unknowns of AI, ultimately enhancing innovation while effectively managing risks.

Unlike traditional risks, AI-related risks often arise from proprietary algorithms trained on massive and sometimes undisclosed data sets. This creates a challenge for insurers in underwriting and pricing risks they don’t fully understand. Additionally, organizations using AI face scrutiny from regulators, investors, and customers, who are cautious about potential liabilities and operational disruptions stemming from AI failures.

In this landscape, no single player can address AI risks alone. Data sharing among insurers, tech companies, and businesses is essential for informed underwriting, proactive risk mitigation, and driving innovation.

Enhancing Risk Visibility

AI systems are only as good as the data they process and understanding how these systems behave in the real world requires access to a diverse range of performance metrics, incident reports, and near-miss data.

Consider this hypothetical scenario. A health-tech company uses AI to assist in diagnostic imaging. Over time, the accuracy of the AI model begins to decline — a problem known as model drift. Model drift refers to the phenomenon where the performance of a predictive model degrades over time due to changes in the underlying data distribution. The client doesn’t detect the drift until incorrect diagnostic results trigger a malpractice claim.

Could such a claim be prevented if the company shares de-identified diagnostic and performance data with their insurer? The insurer, who has aggregated similar data from other healthcare clients, detects a broader trend of model drift tied to recent changes in imaging technology. Working with a tech company, they co-develop a tool to monitor AI performance in real-time and flag potential drift early. As a result, clients reduce liability exposure, tech firms improve model stability, and insurers gain tools to assess and mitigate evolving AI risks proactively. Think about it.

Coordinated Defense

Let’s look at another possibility. A global financial services firm is nearly defrauded by a convincing AI-generated voice deepfake impersonating a senior executive. The incident is isolated — until similar attempts are reported by others. How might sharing data help?

Let’s say the company reports the deepfake attempt and shares forensic data with their cyber insurer. The insurer, who insures multiple firms in the sector, sees a pattern and engages a cybersecurity tech partner. Together, they create an early-warning system that uses shared incident data and voiceprint signatures to alert other insured clients of similar threats.

Clients are better prepared to detect and stop emerging AI-based fraud. Insurers reduce exposure to large cyber claims. Tech firms build more responsive security tools using real-world attack data.

Sharing insights can significantly enhance understanding of potential risks, leading to better insurance coverage and risk management solutions.

A New Way of Working

AI risks are dynamic, evolving with model retraining, new data, and software updates. Traditional annual underwriting may not suffice for managing these risks. How will we have to change? Could secure data-sharing frameworks allow insurers to monitor model behavior and system changes, enabling timely adjustments to coverage terms that would aligns risk exposure with insurance protection more effectively.

Could insurers incentivize responsible AI development by offering better terms and conditions to coverage or enhanced limits to companies that adhere to best practices in governance, testing, and transparency?

In turn, might we encourage the creation of safer AI systems and help establish market norms for risk-resilient AI deployment. It’s all possible, if we work closely together.

Overcoming Data Sharing Hesitancy

Despite its benefits, data sharing faces several hurdles—privacy concerns, competitive sensitivities, and legal restrictions among them. There are ways we can address those challenge to meet these new AI challenges:

Establish clear data governance: Developing industry-wide standards for what data can be shared, how it’s anonymized, and who has access is essential. Frameworks like the EU’s AI Act or the NIST AI Risk Management Framework offer starting points.
Build trust through transparency: Insurers and tech companies should clearly communicate how shared data will be used, how it’s protected, and what the benefits are for all parties. Pilot programs, sandbox environments, and data trusts, a legal and organizational framework that allows data to be shared and managed in a way that protects the interests of those involved, can help demonstrate value and build confidence.
Use enabling technologies: Secure multiparty computation, federated learning, and synthetic data generation are new technologies that let groups share data without revealing private information. These tools could play a key role in how we share risks as artificial intelligence becomes more common.

A Data-driven Future Requires Collective Action

AI offers transformational benefits—but only if its risks are well understood and managed. The future of AI risk management relies on collaboration among insurers, technology companies, and clients, with secure and purposeful data sharing at its core.

Insurers should become proactive partners in AI governance by developing specific policies for AI risks, using loss data to improve model safety, and sharing risk insights with regulators. Meanwhile, technology companies need to integrate risk management into their AI development processes by offering transparency reports, collaborating with insurers on audits, and helping clients adopt responsible AI practices.

We won’t get here overnight. It will require trust, transparency, and shared standards. But the payoff is clear: a safer, more innovative digital economy where the promise of AI can be fully realized without compromising trust or stability.

By working together and leveraging data wisely, we can ensure that AI is not just smart—but also safe. &

Social Determinants of Health: The Missing Piece in Workers’ Compensation

Social and environmental factors can quietly derail recovery; integrating these insights into claims management is key to lowering costs, reducing litigation, and improving outcomes.

By: CorVel | October 1, 2025

Over the past 20 years, the workers’ compensation industry has worked to combat rising medical and claims costs through strategies such as early intervention and early engagement (24/7 nurse triage) and the advancement and acceptance of case management.

Now, we’re at an inflection point. Medical costs continue to rise, and complex cases, such as those involving mental health claims in addition to physical claims, are contributing to higher overall costs. Let’s take a step back from the individual injury or claim and examine cases from a program perspective. It becomes clear that outside factors are influencing the severity and complexity of claims today. These factors are the social determinants of health (SDOH).

The Centers for Disease Control and Prevention defines SDOH as the nonmedical factors that influence health outcomes. They are the conditions in which people are born, grow, work, live, worship, and age. SDOH often include factors like income, education, housing, and access to nutritious food and safe environments. In workers’ compensation, these factors may include an individual’s residence, access to healthcare, their home support system, and financial situation.

SDOH aren’t new, but they’re the next most obvious factor contributing to higher costs and impacting the care, recovery, and return-to-work of injured workers. There’s an opportunity for the industry to reduce the overall cost of claims by first understanding how social factors influence recovery. Then it becomes a process: flag the more severe or complex claims, identify the social determinants at play, and manage them through a claim-centric and holistic approach.

Understand Social Determinants to Guide Recovery

Sarah Scott, Executive Vice President of Product & Corporate Services, CorVel Corporation

When a worker gets hurt, their recovery depends on more than medical treatment. Rural workers may face longer drives to specialists than their urban counterparts. Workers without family support may struggle with their daily tasks during recovery that would otherwise be taken care of. Financial stress adds another layer that could impact healing and recovery if workers are worried about bills piling up.

“Think of it this way,” says Sarah Scott, EVP of Product & Corporate Services at CorVel. “Two workers suffer the same back injury at a manufacturing plant. One lives near a physical therapy clinic and has strong family support. The other lives in a healthcare desert with little support at home. Even with identical treatment plans, their outcomes can look very different because of these external factors.”

In fact, a recent WCRI study of more than 930,000 workers’ compensation claims across 32 states found that degenerative and comorbid conditions consistently correlated with higher costs and longer durations of temporary disability. These findings underscore how non-medical or complicating factors can dramatically shape recovery trajectories.

These longer recoveries and more complex cases account for a small portion of the overall claims, but they’re increasingly responsible for a growing share of the total costs.

Detect High-Risk Cases Early to Reduce Complexity

“The earlier we identify social and environmental barriers, the more likely we are to keep a claim on track,” says Scott. “Early recognition gives us time to act before small challenges become costly delays.” That’s why leading organizations are training intake nurses to look beyond symptoms and capture broader context about the injured worker’s living situation, support network, and potential barriers.

At CorVel, structured nurse triage protocols document not only the injury itself but also cues related to psychosocial and environmental risk. These details are codified into the claim file and automatically feed into CorVel’s system, Care^MC, where predictive workflows can flag cases for additional support.

For claims that may become complex, the first month is pivotal. High-touch involvement during this window, through outreach with providers, injured workers, and employers, can dramatically improve the likelihood of a timely recovery. During this period, CorVel deploys registered nurse Care Advocates to establish evidence-based treatment plans, coordinate optimal care, and initiate return-to-work planning. This proactive, early focus helps reduce costly delays, mitigate litigation risk, and improve closure rates.

Scenario: An injured worker reports a knee injury at the time of intake. During the triage conversation, the nurse notes the individual lives far from the nearest clinic, is the primary caregiver for a parent, and is anxious about lost wages and transportation challenges. Because these factors were documented early, the claim was flagged for additional support. A nurse case manager collaborates with providers to develop a treatment plan and consults with the employer on modified duty options. Transportation and virtual behavioral health check-ins are arranged to reduce barriers, while predictive modeling monitors recovery and flags risks. By addressing these concerns up front, the claim stays on track and avoids costly delays or disputes.

These conversations and early interventions capture details that automated systems miss. People are often more inclined to reveal information about themselves in conversation than on a form. By blending structured protocols with human interaction, CorVel is able to identify risks earlier and align interventions accordingly.

Apply Data and Technology to Optimize Claims Management

Human involvement is undeniably essential to the recovery and return-to-work of injured workers; however, technology and automation also play a significant role in this process. Modern claims platforms don’t just process injury codes; they analyze dozens of data points, including age, location, past medical history, and environmental risks. Advanced systems utilize these insights to flag claims that require special attention, triggering targeted interventions before costly complications arise.

“Technology can’t replace human judgment, but it can sharpen it,” says Scott. “By combining predictive analytics with the insights of experienced claims professionals, we can spot risks earlier, act faster, and keep recoveries on track in ways that wouldn’t be possible with either alone.”

CorVel’s predictive modeling tools continuously recalculate dynamic risk scores using medical, historical, and psychosocial data points. This enables claims teams to respond in real-time as new information is added, whether that involves initiating a behavioral health escalation, aligning with an evidence-based treatment benchmark, or coordinating transportation services.

For example, if a high pain score combines with a lack of transportation access, Care^MC Edge generates real-time alerts for adjusters and case managers. Diagnosis Profile tools then project lost workdays based on diagnosis, physical job demands, and comorbidities, ensuring recovery plans stay realistic and measurable.

When humans and technology work together, social determinants become manageable factors instead of invisible barriers.

Commit to a Path Forward to Manage Costs

As the cost of care and wages continue to rise, 2025 brings outsized risk for both insurers and employers. More complex and severe claims require a holistic, data-driven approach that considers the whole person rather than just the injury. Legislative shifts, such as those related to mental health coverage, fee schedule updates, and pharmacy controls, add yet another layer of complexity.

The next frontier of SDOH integration across the industry will likely involve AI and machine learning. These tools can analyze unstructured data, such as case notes or provider communications, and enrich risk models with external datasets, including housing or transportation indexes. CorVel is actively evaluating how these innovations can enhance our integrated approach, with the goal of identifying risks earlier and responding more effectively to support improved outcomes.

“When it comes to addressing social determinants of health in workers’ compensation, the evidence is clear,” says Scott. “Programs that blend technology, early human interaction, and targeted resources not only improve worker well-being, but also help organizations manage costs more predictably, reduce litigation exposure, and meet compliance requirements.”

To learn more, visit: https://www.corvel.com/.

This article was produced by the R&I Brand Studio, a unit of the advertising department of Risk & Insurance, in collaboration with CorVel. The editorial staff of Risk & Insurance had no role in its preparation.

CorVel is a national provider of risk management solutions for employers, third party administrators, insurance companies and government agencies seeking to control costs and promote positive outcomes.

From Data Silos to Shared Insights: Transforming AI Risk Management