Ryan Smith is the Senior Solutions Advisor for True Insurtech Solutions
Every NCCI Annual Insights Symposium has a center of gravity. AIS 2026 has a name. Ethan Mollick, Associate Professor at The Wharton School and one of the most influential academic voices on practical AI adoption, will take the keynote stage with a session titled From Disruption to Opportunity: Embracing the AI Revolution. AI is the conversation this year. You know it. I know it. The question is whether workers’ compensation can move past the conversation and into application that serves the people on the other end of every claim.
Why Mollick’s Keynote Will Land Differently in Workers’ Comp
Mollick’s frame is cross-industry by design. He studies what happens when knowledge workers across functions, sectors, and skill levels use generative AI in their actual jobs, work he distilled into the 2024 New York Times bestseller Co-Intelligence. His findings are direct and consistent. AI tools, used well, raise the floor of work quality. They flatten learning curves, accelerate output, and shift the value of expertise from production to judgment. He is not a workers’ compensation researcher, and that is precisely why his framing matters.
Workers’ compensation is high-trust, regulated, and claimant-facing. The line operates with thinner data than the consumer-facing industries that have driven most public AI coverage, and it serves people at vulnerable moments in their lives. Translating Mollick’s macro thesis into the workers’ comp context requires more than swapping in industry jargon. It requires a working theory of how AI earns its place in a system that is fundamentally about getting injured workers what they need.
At True, we believe in technology that works in service of people, not in place of them. It is the lens worth applying to every AI adoption decision carriers make this year. Mollick will frame the macro story from the AIS stage. The rest of this article frames the workers’ comp localization.
The Honest State of AI Adoption in Our Industry
The hype cycle on AI in workers’ comp peaked sometime in late 2024. The reality has been slower, more uneven, and more interesting than either the optimists or the skeptics predicted.
The Workers Compensation Research Institute’s 2025 study, Artificial Intelligence in Workers’ Compensation: An Overview of Promises and Challenges, interviewed 34 leaders across 20 organizations and found that stakeholders believe 70 to 80 percent of workers’ compensation claims could be highly automated using AI. That headline number is a forecast of where the industry could be. The gap between potential and current practice is the single most important thing to understand about AI in workers’ comp going into AIS 2026.
Most carriers and administrators are running AI somewhere in their organization right now. Few are running it everywhere. The early adopters have moved past pilots into production on specific use cases, while the broader market is still working through governance frameworks, vendor evaluations, and the harder question of which problems are worth solving with AI in the first place. True’s 2025 State of AI in Workers’ Comp Report catalogs where that adoption is concentrating and where it is stalling.
That mixed picture is a feature, not a bug. Workers’ compensation has historically been slow to adopt new technology, and that caution has served claimants well in some respects. The same caution becomes a liability when the technology has matured to the point that it can demonstrably improve outcomes and the industry’s reluctance is mostly cultural.
Where AI Is Already Paying Off
Three application areas have moved past proof of concept into real production value for workers’ compensation carriers. Each one maps to a specific claimant or underwriting moment that AI is making demonstrably better.
Claims communications and the claimant voice
The first place AI has earned its keep in workers’ comp is in claims communications. Adjusters write more letters, emails, and explanations than most people outside the industry realize, and the quality of those communications shapes how claimants experience their recovery. A confused, defensive, or off-tone communication can extend a claim’s emotional duration well past its medical one. A clear, warm, accurate communication can do the opposite.
Generative AI applied carefully to this layer returns adjusters’ attention to the claims that need human judgment instead of consuming it on the routine correspondence that does not. The early operators in this space, including Vera by True, are the ones treating tone, accuracy, and claimant context as design constraints rather than afterthoughts. The carriers that get this layer right see measurable shifts in claimant satisfaction, in adjuster retention, and in the speed at which routine claims close.
Analytics-led underwriting and decision support
The second area is decision support. Underwriting and claims both involve a near-constant stream of decisions made under partial information, and the cost of small judgment errors compounds across a book. AI does not replace those judgments. It surfaces the patterns that experienced underwriters and claims professionals would identify if they had the time to review every relevant data point on every account.
The insurers getting the most out of analytics-led AI are the ones investing in tools that present leading indicators in established workflows underwriters and claims leaders already use. Dashboards that need a translator are not decision support. They are reporting. The difference matters, and it is one of the cleanest dividing lines between AI investments that compound and AI investments that stall.
Safety and prevention: AI in service of people
The third area is the one that most directly answers Mollick’s framing. AI applied to workplace safety and risk prevention has produced measurable injury reduction. True’s State of AI Report documents that companies implementing AI-driven safety programs have seen workplace injuries decline by roughly 25 percent in high-risk industries, drawing on NCCI’s research into the application of AI to workplace risk. Predictive analytics paired with sensor and wearable data is identifying high-risk activities and prompting intervention before incidents occur. Used well, AI can streamline operations and help prevent the injuries that create the claims that drive the conversations Mollick will frame at AIS 2026.
Read Our State of AI in Workers' Comp Report
Where the Industry Is Still Figuring It Out
Honesty matters more than enthusiasm in this section. There are real barriers to AI adoption in workers’ compensation, and the carriers that will lead this conversation in 2027 and beyond are the ones treating those barriers as practical problems with practical answers rather than reasons to wait.
Three challenges keep surfacing in conversations across the industry.
The first is governance. Workers’ compensation operates under state regulatory variation, federal data and privacy frameworks, and bureau-specific reporting requirements. AI introduces new questions about model transparency, bias monitoring, and audit trails that existing compliance functions were not built to answer. The WCRI study captured the same dynamic, with its authors noting that AI’s value depends on how transparent it is, how well it is governed, and whether it genuinely supports the mission of helping injured workers recover. Solving the governance problem is not optional for carriers operating across multiple states, and the carriers solving it well are the ones building governance into their AI roadmap from day one rather than retrofitting it after a model is in production.
The second is data quality. AI is only as good as the data it learns from, and most carriers’ data is scattered across legacy policy systems, claims platforms, billing engines, and document repositories that were never designed to talk to one another. The result is that many AI projects in our industry stall not because the model fails, but because the data underneath it is too fragmented, too inconsistent, or too dirty to support production use. Quietly, this is the limiter on most AI ambitions in workers’ comp.
The third is change management. McKinsey’s 2025 report, Superagency in the Workplace, found that 48 percent of US employees would use generative AI tools more often if they received formal training, and 45 percent would use them more frequently if those tools were integrated into their daily workflows. That gap is the difference between an AI rollout that compounds and one that stalls. Tools without training do not produce outcomes. They produce skepticism.
What “Embracing the AI Revolution” Looks Like for Different Players
Mollick’s framing is intentionally portable, and the workers’ comp adaptation needs to be too. The right starting point depends on the constraints and opportunities of the operator.
Carriers should start with the use case that touches the most claimants and underwriters most often. For most insurers, that means claims communications and analytics-led decision support, in that order. Both produce visible results within a quarter and build the organizational confidence required for broader AI investment.
Self-insured groups have the advantage of focused books and faster decision cycles. The right first move is often analytics-led, because the data set is concentrated enough to produce reliable insights and the operating leadership can act on them without the layered governance review a multi-state carrier requires.
Captives benefit most from the experience layer. The parent company already has a view of risk, and the captive’s value is in the precision of its claimant and policyholder experience. AI experience engines are a natural starting point.
MGAs operate at the edge of carrier and broker workflows, and AI’s clearest payoff for them is in the speed and consistency of underwriting communications and decision support. The MGAs that lead the next cycle will be the ones that pair their underwriting expertise with AI tooling that scales the consistency of that expertise across every account they touch.
TPAs sit closer to the claimant than any other entity in our industry, and the AI investment that pays the fastest is in the tools that improve the claimant experience without adding burden to adjusters. Communications AI, paired with predictive triage, is the highest-leverage starting point.
From Disruption to Opportunity
Mollick’s keynote title is well chosen. Industries that stay passive get disrupted, while those that make deliberate choices about how AI serves their customers find opportunity instead. Workers’ compensation has the chance to be the first regulated insurance line that embraces AI in a way that visibly improves the claimant experience, not just the operator’s margin. That outcome is not automatic. It depends on insurers treating AI adoption as an exercise in service rather than substitution.
To go deeper on the data behind this years’ AIS conversations, keep an eye on this space for our full post-conference analysis. To talk through what it means for your organization, schedule a follow-up call with Ryan Smith.
