The Hidden Hurdles
Agentic AI is set to usher in a new age of insurance.
Agentic systems can reason on a much broader scale than previous AI tools, autonomously planning and executing a variety of tasks all at once. Tools that can assess claims, flag risks, draft responses, and handle customer interactions in real time are understandably appealing to an industry constantly in pursuit of efficiency for both customers and employees.
But while the promise is clear, progress remains uneven.
Many initiatives lose momentum when insurers try to scale beyond early use cases – not for lack of intent or investment, but because of the hidden hurdles that separate experimentation from meaningful impact. What exactly are these hurdles and what separates the successful adopters from the rest of the pack?
Agentic AI in Practice
Agentic AI in insurance refers to systems that can strategise, make decisions, and execute tasks across workflows, with minimal human input.
These capabilities make agentic AI well-suited to the claims process. AI-powered claim management systems can process 70–90% of simple insurance claims, with decisions delivered in minutes, acccording to ScienceSoft Insurance.
Chatbots receive initial loss notifications, categorise claims automatically, and flag cases whose complexities require human intervention. That’s where the buck stops with pre-agentic AI tools. A human would have to take the information and either move forward on their own, or input the information into a different AI tool.
When an AI agent receives a claim, it checks whether the policy covers the reported damage by assessing the claim details, identifying comparable cases, averaging typical settlement values, scanning for fraud markers, and reviewing the customer’s history. If the AI agent sees that everything is in order, it will automatically confirm that the loss is covered.
Human adjusters are freed to spend more time addressing the situations that require a personal touch, rather than wasting time on tedious filing.
A fully optimised AI agent can take the next steps on its own – automatically recommending a settlement, drafting the settlement notice, and routing the claim for final approval.
Why Progress Stalls
Insurers that use agentic AI will unlock key competitive advantages. So, why do so many initiatives slow down or stall altogether?
Data is usually the first hurdle, because agentic AI relies on context, which in turn depends on the data that informs it. Unfortunately, many insurers still work with fragmented or inconsistent data across their systems, which means the agentic AI pulling its knowledge from these sources will be equally inconsistent, leading to outputs that are unreliable and harder to explain.
Technology architecture is another common constraint, as many core insurance platforms are legacy systems, build for a pre-AI world. When AI is layered on top of such outdated systems (rather than built into or around new ones), the resultant AI agents will never reach their full potential.
Governance is another key factor. Autonomous decision-making raises valid questions about accountability, transparency, and auditability. When those questions are not clearly answered, risk teams remain wary and wider rollout can be delayed.
Finally, many initiatives lose momentum due to unclear value-add. Senior leaders want to see tangible outcomes before they greenlight costly new initiatives. When success is not effectively defined, organisational leadership is liable to wave off agentic AI as an interesting experiment, rather than a strategic priority.
Overcoming the Hurdles
The insurers making meaningful progress with agentic AI are building out these capabilities in stages, getting the fundamentals right from the get-go.
The first step is to strengthen the overall digital backbone by breaking down silos, modernising integration, and treating data as a strategic asset. Next is focusing on use cases where value is visible and risk is manageable – areas that benefit from speed and consistency, such as claims triage, document handling, customer communications, and fraud detection.
Early wins will build confidence and prove value, making wider initiatives easier to scale.
Make sure governance is prioritised from the outset. Clear guardrails, transparent decision paths, and well-defined escalation points make it easier for users to trust AI-driven decisions and act on them. When this is combined with metrics that link directly to operational and financial outcomes, agentic AI becomes something leaders can get behind with confidence.
Experimentation to Advantage
Agentic AI is not a switch to flip: it is a progressive process that rewards clarity, discipline, and patience. The roadblocks are real, but they can be overcome.
Once agentic AI is running smoothly, claims move faster, underwriting is more consistent and scalable fraud detection improves through continuous pattern recognition, while customer interactions also improve thanks to systems that understand context, not just scripts. Experienced professionals are then freed up from routine tasks and have more time to focus on oversight, complex cases, and customer relationships – where they add the most value.
Insurers shouldn’t chase autonomy for its own sake. The insurers who thrive in the agentic AI era will be those who scale it thoughtfully and turn intelligent automation into a lasting competitive advantage.
