AI in Insurance Use Cases and Challenges: Insights from LIMRA ETSS 2026

AI adoption in insurance is no longer limited by access to technology. It is limited by how effectively that technology is applied within real workflows.

At LIMRA ETSS 2026, discussions across carriers, platforms, and partners pointed to a clear shift. The industry is moving beyond experimentation, but measurable impact remains uneven. While AI is being deployed across communication, data processing, and decision support, most implementations are still constrained by data availability, integration gaps, and unclear ownership of outcomes.

This is happening in an environment where operational pressure continues to increase. A significant portion of work across benefits and insurance remains communication-heavy, with estimates suggesting that nearly 64–68% of work time is spent on communication and information processing. At the same time, employees expect faster, more personalized interactions, while employers demand clearer ROI from technology investments.

Where AI in Insurance Is Actually Working

AI adoption is delivering the most consistent value in areas where work is repetitive, communication-heavy, and operationally intensive. These are workflows where outputs can be structured, validated, and improved without introducing decision risk.

Across LIMRA ETSS 2026 discussions, AI is not replacing core insurance decisions. It is improving how work gets executed around them.

This is where adoption is already visible:

• Communication workflows: AI is being used to draft emails, summarize conversations, generate documentation, and streamline coordination across teams.
• Data consolidation and processing: Insurance workflows often require combining inputs from multiple systems. AI is applied to structure and organize this data, reducing manual effort across enrollment, case setup, and servicing.
• Operational support and internal productivity: Use cases such as case summaries, reporting, and internal documentation improve consistency and reduce turnaround time.
• Pre-decision support layers: AI assists by organizing inputs, identifying gaps, and suggesting next steps before decisions are made.

These use cases operate within clearly defined workflows, where outputs can be reviewed before being finalized. This is where AI delivers value today.

Where AI Breaks Down for Insurance

AI is often positioned as a decision-making layer. In practice, this is where it is least reliable. These intelligent systems generate outputs based on probability, not certainty. They produce the most likely response given available data, which does not always translate to accuracy in real-world scenarios. This becomes a constraint in insurance workflows that require consistency, explainability, and accountability.

These limitations are most visible in areas that depend on judgment and context:

• Interpreting incomplete or nuanced data: Insurance decisions often rely on context that is not fully captured in structured inputs.
• Handling exceptions and edge cases: AI performs well on common scenarios but is less reliable when workflows deviate from expected patterns.
• Supporting accountable decisions: Underwriting and eligibility decisions carry financial and regulatory implications, requiring controlled and auditable outcomes.

AI outputs are often presented as complete and well-formed, which can lead to over-reliance if not validated. In practice, consistent verification remains limited, increasing the risk of downstream errors.

For this reason, AI is not replacing decision-making in insurance. It is being applied around it, with human oversight remaining essential in high-stakes workflows.

Why AI in Voluntary Benefits Is Constrained by Data and Workflow

Applying AI in benefits and enrollment workflows introduces constraints shaped by data, time, and user behavior.

The first constraint is data availability. Personalization depends on detailed inputs, yet willingness to share sensitive data remains limited. For example, only ~30% of users are willing to share claims data, restricting the effectiveness of AI-driven recommendations.

The second constraint is workflow placement. Decision support tools are often introduced outside the core enrollment flow, requiring additional inputs or steps. With average enrollment interactions at ~17 minutes, this added friction can directly impact completion rates and participation.

The third constraint is trust and usability. AI-driven recommendations must be clear and explainable. In practice, long questionnaires, opaque logic, and disconnected experiences often lead to drop-offs rather than engagement. AI is most effective when it reduces friction within enrollment. It is least effective when it introduces new steps or operates outside the primary workflow.

How AI Works Within Insurance Workflows

AI delivers the most value when it is embedded within workflows rather than introduced as a separate layer.

Standalone tools and parallel processes often create fragmentation, requiring users to switch systems or re-enter data. This limits adoption and reduces impact.

The shift is toward integrating AI directly into workflows:

• Data is processed in-line, as it is captured
• Outputs are tied to actions, enabling immediate next steps
• Workflows remain continuous, without additional system dependencies

As infrastructure moves toward real-time, event-driven models, AI becomes part of how systems operate, not an external layer.

Effectiveness depends on alignment. When embedded, AI reduces friction and improves consistency. When separate, it introduces complexity.

Where to Apply AI First in Insurance

The most effective starting points for AI adoption are operational workflows where effort is high, inputs are structured, and outcomes can be validated.

This includes:

• Communication-heavy processes
• Data consolidation across systems
• Internal documentation and reporting
• Pre-decision support layers

These areas sit adjacent to decision-making, allowing AI to improve execution without introducing risk.

Starting with fully automated decision-making often adds complexity. A more effective approach is to begin with execution layers, then expand as data, workflows, and governance mature.

Execution Will Define AI Value in Insurance

What is emerging from LIMRA ETSS 2026 is a shift in how AI is being evaluated. The focus is moving away from capability and toward measurable impact within day-to-day operations.

In practice, this is changing how organizations approach adoption. Rather than starting with broad transformation goals, teams are focusing on specific points of friction across enrollment, servicing, and internal workflows, where effort is high and outcomes are easier to track.

This creates a more practical path forward:

• Start with workflows that are communication-heavy or operationally intensive
• Prioritize use cases where data is already structured and accessible
• Measure impact through improvements in turnaround time, completion rates, and error reduction
• Expand gradually into more complex workflows as systems and governance mature

This approach allows AI to scale with the system, rather than forcing the system to adapt to it. The shift is subtle but important. Progress is no longer defined by how much AI is deployed, but by how effectively it improves execution across the lifecycle.