The emerging role for contextual AI in insurance

<p><b>By bringing together multiple streams of customer data, contextual AI could unlock exciting new results.</b></p>

<p><i>This article was originally published in </i><a href="https://www.carriermanagement.com/" target="_blank" rel="noopener noreferrer"><i>Carrier Management</i></a><i> in February 2026.</i></p> <p>The wave of excitement about artificial intelligence is breaking across virtually every industry. We are discovering in the insurance industry that the current generation of tools may have the ability to reshape work—especially knowledge work and content creation—should regulatory, organizational and leadership dynamics align.</p> <p>This is in stark contrast to the AI and machine learning interventions that have been rooted in the risk and insurance industry for well over a decade.</p> <p>Based on first-generation experiments, most insurers now believe the truth is out there. (Apologies to fans of <i>The X Files</i>.) On verticalized topics like underwriting, claims propensity and fraud, members of the insurance C-suite are gaining confidence that the latest AI can deliver beyond straight speed to support better decision-making about how to proceed.</p> <p>But as usual, there’s a catch. Tapping this potential will require a rethink of how the industry works with its value chain, roles and human caretakers while protecting the integrity of the business execution. Supporting nuanced and high-risk decisioning requires that insurers execute a thoughtful shift into&nbsp;contextual AI.</p> <h4>Outlining the shift</h4> <p>Contextual AI and context engineering are entering the zeitgeist because a consensus is forming around where AI still falls short of human decision-making.</p> <p>This includes issues like persistence, where the machine needs to remember related conversations that preceded the current one, and transparency, which is the ability to retrace a decision back to its roots. Insurance regulators are particularly keen on this.</p> <p>Most critically, we’ve come to realize that human decisions are rarely as black and white as they appear. True understanding comes from considering a range of inputs beyond the obvious. Humans think of this as intuition, and machines have historically been unable to replicate or even approximate it.</p> <p>In an underwriting context, core decision factors might include property values, risk mitigation tools, the coverage requested and information about the property owner. Additional context can be derived from things like the property owner’s other insurance policies, their communication and prior service requests, social media linked to the property or the owner, and so on.</p> <p>Many of these data points are already captured somewhere within an insurer’s four walls, or are easily accessible. But most insurers have not engineered systems to aggregate, manage and extract value from this helpful contextual data. And without it, AI’s ability to impact decision quality, speed and cost will be limited.</p> <h4>Making it happen</h4> <p>When it comes to AI and insurance, we do not know (yet) what we do not know. Things are developing quickly, and sometimes in unexpected ways. Ironically, this means that creating and executing a plan to support contextual AI requires—you guessed it—intuitive human decision-making and planning. This means emphasis on structure and process that can accommodate future shifts in thinking and mind the guardrails of both regulation and consumer protection.</p> <p>In broad strokes, we believe the path to contextual success will include:</p> <ul> <li>Building a core team that blends both AI skills and insurance knowledge. If your team structure sounds like the start of a joke (“An IT architect, a claims manager and an actuary walk into a bar …”), you’re probably on the right track.</li> <li>Integrating data across fragmented core systems (e.g., underwriting, policy admin, claims, CRM), with unified real-time access as a central construct.</li> <li>Selecting additional data to inform a context layer, from both public and internal sources.</li> <li>Heightened security awareness, with threats like prompt injection addressed by role-based filters and audit trails.</li> </ul> <p>All this must be taken into context (pun intended), positioned and engineered in alignment with governance that follows guidance such as the NAIC Model Bulletin. For example, Amazon Bedrock, from AWS (<a href="https://www.cognizant.com/us/en/about-cognizant/partners/aws">a Cognizant partner</a>), helps insurers align with NAIC principles—specifically focusing on transparency, fairness and accountability—creating a safe, scalable foundation for agentic solutions, whether applied to underwriting, claims, service, billing, fraud or agency operations.</p> <h4>Eye on the prize</h4> <p>There is no magic wand that will make contextual AI a reality. Resources and commitment are required. A number of insurers understand the value and are already engineering for context in search of a powerful competitive advantage. What will they, and others who pursue contextual AI, get in return for this investment? What business value and KPIs will define success?</p> <p>Engineering a solution that unlocks business value will require a series of non-negotiable design principles, including but not limited to:</p> <ul> <li>Modeling agent authority and escalation paths.</li> <li>Establishing and monitoring acceptable error thresholds.</li> <li>Ensuring explainability and regulatory compliance.</li> <li>Managing model versioning, drift detection and continuous validation.</li> <li>Designing with human‑in‑the‑loop processes.</li> <li>Defining holistic integration across systems, data and operations.</li> </ul> <p>For most efforts, the chain of benefits will start with<b>&nbsp;</b>improved decision quality.&nbsp;In pricing, this means addressing customer needs more dynamically and reliably. In underwriting, it means decisions that reflect complexity better, with more clean cases (submissions with all required information and no data inconsistences) teed up for underwriter review than ever before. In claims, it means faster and more accurate payments, with improved fraud detection.</p> <p>We believe contextual AI may ultimately power&nbsp;over-the-horizon insights, highlighting opportunities that are invisible today. For example, the industry has been chasing upsell and cross-sell for decades with mixed success. By bringing together multiple streams of data related to customer wants and needs and their behavior outside of current insurance transactions, contextual AI may help unlock new top- and bottom-line insurer results critical in today’s soft markets.</p> <p>For channel partners and customers, the anticipated benefit will be&nbsp;improved process speed and convenience, plus&nbsp;tailored services&nbsp;that feel more thoughtful and relevant. This is a case in which sound technology use can actually make processes feel more human and connected.</p> <p>Insurers are right to be cautious in their pursuit of all things AI, particularly as regulators have yet to supply definitive guidance on how and when the technology may be applied. But the AI genie is not going back in the bottle. Creation of a thoughtful, contextual AI approach to help maximize the value and drive the industry forward is sensible.&nbsp;</p>