Table of Contents
- What Is Explainable AI in Insurance?
- Why Explainable AI Has Become a Compliance Requirement in 2026
- Every Communication Is an Accountability Moment
- Reason Codes Are Not Customer Explanations
- Four Places AI-Driven Decisions Create Communication Risk
- Template Sprawl Makes Explainability Harder
- How Insurers Can Operationalize Explainable AI Across Communications
- How MHC NorthStar CCM Supports Defensible Communications
- The Bottom Line
- Key Takeaways
- FAQ: Explainable AI in Insurance
WEBINAR
Watch the full webinar, Closing the AI Explainability Gap in Insurance. Andy Young, Rao Tadepalli, and Emily Washington share practical perspective on how insurers can bring greater visibility, consistency, and control to the last mile of AI explainability.
Explainable AI in Insurance: From Model Governance to Defensible Communications
Sharon Malloch 
June 23rd, 2026
In Brief
Explainable AI in insurance means insurers can show how AI-supported decisions were made, governed, and communicated to policyholders. As AI accountability expectations rise, the biggest explainability gap is not only in the model. It is in the customer notices, claim denials, policy documents, and disclosures that result from those decisions. Closing that gap requires both AI governance and communications infrastructure built for compliance, traceability, and scale.
Table of Contents
- What Is Explainable AI in Insurance?
- Why Explainable AI Has Become a Compliance Requirement in 2026
- Every Communication Is an Accountability Moment
- Reason Codes Are Not Customer Explanations
- Four Places AI-Driven Decisions Create Communication Risk
- Template Sprawl Makes Explainability Harder
- How Insurers Can Operationalize Explainable AI Across Communications
- How MHC NorthStar CCM Supports Defensible Communications
- The Bottom Line
- Key Takeaways
- FAQ: Explainable AI in Insurance
AI adoption in insurance is moving quickly. NAIC survey data shows that 88% of auto insurers, 70% of homeowners insurers, 58% of life insurers, and 92% of health insurers already use, plan to use, or plan to explore the use of AI in their operations.
But as more and more insurers leverage AI to support underwriting, pricing, claims, policy servicing, and customer experience, they need to answer a critical question: Can they explain how an AI-influenced decision impacted the policyholder?
That question is quickly becoming a compliance issue as regulators raise expectations around AI accountability. Insurers (and many other financial institutions) must not only govern AI models internally. They must also clearly explain AI-supported decisions that affect customers.
That is where the communications challenge begins. A decision may start upstream in a model, rules engine, claims system, pricing process, or underwriting workflow, but the moment it reaches a customer, it becomes something much more visible.
It becomes a letter, notice, email, portal message, policy document, claim explanation, or disclosure. That communication is where the customer looks for answers. It is also where regulators, legal teams, service teams, and internal stakeholders may look later if the decision is questioned.
This was the focus of our recent webinar, Closing the AI Explainability Gap in Insurance. After I introduced our speakers, I listened behind the scenes as Andy Young, founder of TreelinePress, led a thoughtful conversation with Rao Tadepalli, an insurance industry expert and veteran CIO, and Emily Washington, Head of Products at MHC.
Let’s look at the topics raised in the XAI webinar, including what explainable AI in insurance means, why it matters for compliance, and how insurers can turn AI-supported decisions into clear, defensible policyholder communications.
What Is Explainable AI in Insurance?
Explainable AI in insurance requires organizations to understand, document, justify, and communicate how a model, algorithm, or AI-enabled workflow impacted a specific underwriting, pricing, claim, or servicing outcome. In practical terms, it means insurers need to show why a decision was made, what data or factors influenced it, how it was reviewed, and how the outcome was explained to the policyholder.
This concept matters across regulated industries, especially when AI influences decisions that affect customers, citizens, patients, members, or account holders. But insurance has a particular sensitivity because the insurance industry is built on a promise to pay when policyholders need support, whether that moment involves a claim, a loss, a renewal, a pricing change, or a coverage decision.
In a technical context, AI transparency and explainable AI are often discussed together, but they are not the same thing. Transparency helps insurers understand what data went into a model, how the model was trained, and what controls were in place. Explainable AI models go further by helping insurers justify why a specific decision was made and how that decision can be understood, reviewed, and defended.
That distinction is especially important because AI-supported insurance decisions can affect coverage, premiums, claims payments, policy eligibility, cancellations, renewals, and customer trust. Whether teams call this explainable AI, explainability AI, or AI governance, the business challenge is the same: insurers need AI-supported decisions to be understandable, traceable, and defensible beyond the model itself. That clarity is essential for building trust with policyholders, regulators, and the teams responsible for explaining decisions later.
► Watch the clip: Why AI decisions must trace back to the source
Emily Washington explains why explainability has to connect upstream AI decisions to the customer communications policyholders actually receive.
Why Explainable AI Has Become a Compliance Requirement in 2026
AI accountability is becoming harder for insurers to ignore as regulators continue to focus on governance, documentation, fairness, and transparency. The NAIC Model Bulletin on the use of AI systems by insurers has established expectations around governance, risk management, testing, documentation, and accountability, while state-level activity continues to raise expectations around automated decision-making and consumer notice.
The NAIC’s AI Systems Evaluation Tool is now being used for a 12-state pilot giving regulators a more structured way to evaluate insurer AI programs during market conduct examinations. That makes explainability more than a future concern, especially for insurers that are already using or exploring AI across underwriting, claims, pricing, and servicing.
Recent enforcement activity also shows why communications matter. In Pennsylvania, the Attorney General announced a settlement with GEICO after an AI-enabled underwriting tool allegedly contributed to a policyholder’s coverage being cancelled without sufficient notice. The issue was not simply that AI was involved; the concern centered on how the decision-making process affected customers and whether the communications were clear enough.
For insurers, the takeaway is practical. IBM research found that 48% of leaders interviewed believe decisions made by generative AI are not sufficiently explainable. The reality is, a well-governed AI model may produce a decision, but the customer experiences that decision through a notice, letter, email, portal message, disclosure, or claim explanation. If that communication is unclear, incomplete, or difficult to defend, the explainability gap is still there.
The webinar made one thing clear: explainable AI in insurance is not only a model governance challenge. It is also a communications challenge.
Every Communication Is an Accountability Moment
One of Emily’s strongest points in the webinar was that customer communications are where accountability becomes visible. An AI-supported decision may begin upstream in a model, rules engine, claims platform, pricing process, or underwriting workflow, but once that decision reaches the customer, it becomes a communication that must be understood, supported, and potentially defended.
INSIGHT FROM MHC
“Every letter, email, and portal message you communicate to your customer…that’s the accountability moment. When a customer challenges a claim, or looks at the policy document and believes there’s an inaccuracy, you need to be able to trace that back to its source.”
~ Emily Washington, Head of Products, MHC
That is why the communication record matters. If a customer questions a claim denial, premium change, coverage decision, cancellation, or non-renewal, the insurer needs more than an internal model explanation. It needs to show what the customer received, why they received it, which data and logic informed it, and whether the communication was accurate, approved, and delivered.
This is where explainable AI becomes operational rather than theoretical. It is not only about documenting what the model did, but also about making sure the resulting communication can stand up to scrutiny.
Reason Codes Are Not Customer Explanations
AI-supported systems often produce structured outputs such as scores, classifications, decision logic, or reason codes. These outputs are useful for internal governance and auditability, but they are not automatically meaningful to the policyholder.
A reason code may help an insurer understand why a decision occurred, but the customer still needs a plain-language explanation that connects the decision to their specific situation, policy, claim, or coverage.
► Watch the clip: How CCM turns AI reason codes into explanations
Emily Washington discusses how insurers can turn upstream decision outputs into approved, readable, traceable communications.
For example, if a claim is denied, the customer does not want to decode internal terminology. They want to understand why the claim was denied, what part of the policy applies, what information was considered, and what options they may have next.
The same principle applies when a premium changes, a policy is cancelled, or coverage eligibility is affected. The customer needs a clear explanation of the basis for the action, and the insurer needs that explanation to be accurate, approved, consistent, and traceable.
This is where communication design, approved language, template logic, tone, and governance become essential. Emily made the point that the communications layer has two responsibilities: the communication must reflect the correct decision and data, and it must use the right tone for the customer moment.
That balance is not easy in insurance because communications often need to meet several goals at once. They must be clear enough for the customer, precise enough for compliance, and consistent enough for agents, service teams, auditors, and regulators.
Four Places AI-Driven Decisions Create Communication Risk
AI-supported decisions can create risk anywhere they influence a communication that affects a policyholder. The highest-risk areas are often the moments where the customer receives unfavorable, confusing, or consequential information.
Communication Type | AI Decision It May Reflect | Explainability Risk |
Policy cancellation or non-renewal notices | Underwriting model output or AI-enabled review process | The insurer may need to clearly explain the basis for the action. Vague or insufficient notices can invite regulatory scrutiny and customer confusion. |
Claims denial or partial payment letters | Claims processing, triage, fraud detection, or settlement support | The communication must explain the reasoning behind the decision. Opaque denials can increase disputes, complaints, and litigation risk. |
Premium change notifications | Pricing algorithm, rating factor analysis, or risk scoring model | The explanation must be traceable to approved rating factors and communicated in language the customer can understand. |
Coverage eligibility communications | Risk scoring, underwriting analysis, or third-party data evaluation | The insurer may need to show that the decision was fair, documented, and compliant with applicable consumer protection and anti-discrimination requirements. |
Communication Type | AI Decision It May Reflect | Explainability Risk |
Policy cancellation or non-renewal notices | Underwriting model output or AI-enabled review process | The insurer may need to clearly explain the basis for the action. Vague or insufficient notices can invite regulatory scrutiny and customer confusion. |
Claims denial or partial payment letters | Claims processing, triage, fraud detection, or settlement support | The communication must explain the reasoning behind the decision. Opaque denials can increase disputes, complaints, and litigation risk. |
Premium change notifications | Pricing algorithm, rating factor analysis, or risk scoring model | The explanation must be traceable to approved rating factors and communicated in language the customer can understand. |
Coverage eligibility communications | Risk scoring, underwriting analysis, or third-party data evaluation | The insurer may need to show that the decision was fair, documented, and compliant with applicable consumer protection and anti-discrimination requirements. |
The common thread is that the communication is often where the decision becomes real to the customer. A technically sound AI governance process is only half the picture if the resulting notice or letter does not clearly explain the decision in a way the policyholder can understand.
Template Sprawl Makes Explainability Harder
Many insurers manage hundreds or thousands of templates across lines of business, products, brands, states, channels, and departments. Different teams may own different communications, and some templates may live in legacy systems, manual processes, or disconnected repositories.
That fragmentation was already difficult before AI because it made consistency, version control, and approval management harder to maintain. With AI-supported decisions, the same fragmentation becomes riskier because each communication may need to explain a dynamic, data-driven decision in a precise and defensible way.
If different channels explain the same type of decision in different ways, customers may receive inconsistent information. A mailed letter may say one thing, a portal message may say another, and a call center representative may not have access to the same explanation the customer received.
This creates both a trust problem and a defensibility problem. As Emily noted during the webinar, all the customer has is the communication, and if the customer escalates the issue, disputes the decision, or takes legal action, that communication becomes the record of what was explained.
A modern customer communications management platform helps reduce this risk by centralizing templates, reusable content, business rules, approval workflows, version control, and delivery tracking. And that matters because explainable AI in insurance depends on consistency, and insurers need a way to govern once and communicate everywhere.
How Insurers Can Operationalize Explainable AI Across Communications
Operationalizing explainable AI requires connecting the AI decision layer to the communications layer. Every notice, letter, statement, disclosure, or digital message should be traceable to the decision that triggered it and should communicate the explanation in approved, plain language.
Build the Audit Trail
Insurers need an audit trail that travels with the communication. That means knowing what decision triggered a communication, what data was used, what reason codes were applied, what template was selected, what content blocks were included, who approved the language, when the communication was generated, and how it was delivered.
This is especially important when a decision is challenged later. Rao emphasized during the webinar that insurers cannot simply say an AI system made the decision; they still need to explain how the decision was made and demonstrate that proper governance was in place.
“You’ve got to be able to explain how these things are being done.”
— Rao Tadepalli, insurance industry expert and veteran CIO
For communications teams, that means the audit trail cannot stop at the model. It has to extend into the communication record so the organization can show not only how a decision was made, but also how it was communicated to the customer.
01.
Connect Decisions to Approved Language
Insurers need template logic that can incorporate explainability language into the right communication. That means templates must be able to dynamically include AI-related explanations, reason codes, policy language, jurisdictional rules, product details, customer data, and required disclosures.
Static templates are often not flexible enough for this level of variability. If a claim denial, cancellation notice, premium change, or eligibility decision depends on multiple data points, the communication needs to assemble the right explanation for that specific customer and situation.
That explanation also needs to be approved, readable, and consistent across channels. This is where an explainable AI tool for insurers cannot be limited to model governance alone, because it also needs to connect to the systems that produce customer-facing communications.
02.
Prove Delivery Across Channels
Insurers also need omnichannel delivery with proof of receipt. In regulated communications, it is not always enough to generate the right message, because insurers may also need to show when it was sent, through which channel, whether it was delivered, and what version of the communication the customer received.
That delivery record becomes more important as insurers communicate across print, email, SMS, portals, mobile apps, and other digital channels. A strong communications infrastructure helps insurers manage these requirements consistently while maintaining a complete record of the communication lifecycle.
03.
How MHC NorthStar CCM Supports Defensible Communications
MHC NorthStar CCM is a unified platform that automates the design, data integration, workflows, and omnichannel delivery of customer communications, including the compliance controls needed to produce defensible, explainable communications at scale.
For insurers, that means AI-supported decisions can be connected to approved templates, governed language, customer-specific data, review workflows, and delivery records. Instead of stopping at model-level explainability, insurers can extend explainability into the notices, letters, claim explanations, disclosures, portal messages, and emails customers actually receive.
The goal is not to let AI generate and send regulated communications without oversight. It is to use governed workflows, approved content, and human-in-the-loop controls to help insurers move faster while maintaining accuracy, consistency, and compliance.
Dig Deeper: Follow-up Resources
PRESS RELEASE
Read why Digital Insure, a European insurance broker, selected MHC to modernize its communications technology—prioritizing enterprise scale, ease of use, and readiness for EAA and DORA requirements.
WEBINAR
To learn more, watch the full webinar, Closing the AI Explainability Gap in Insurance. Andy Young, Rao Tadepalli, and Emily Washington share practical perspective on how insurers can bring greater visibility, consistency, and control to the last mile of AI explainability.
The Bottom Line
AI is changing how insurers make decisions, but policyholders still experience those decisions through communications. That is why explainable AI in insurance cannot stop at model governance.
Explainability has to extend into the notices, letters, disclosures, claim explanations, policy documents, portal messages, and emails that customers receive. As AI accountability expectations continue to rise, insurers need communications that can clearly show what happened, why it happened, what information was used, whether the decision was reviewed, and whether the message was accurate and consistent.
The customer does not see the model. The customer sees the message, and that message needs to be clear, compliant, traceable, and defensible.
See how MHC NorthStar CCM helps insurance carriers produce compliant, defensible policyholder communications — Request a Personalized Demo >
Key Takeaways
- AI accountability is a regulatory priority for insurers. NAIC guidance, state-level activity, and AI evaluation efforts raise expectations around governance, documentation, fairness, and transparency.
- Explainability needs to extend beyond the model. Insurers need to justify, document, and communicate the reasoning behind AI-supported decisions to regulators, auditors, and policyholders.
- The communications layer is where risk becomes visible. Claim denials, cancellation notices, premium changes, coverage decisions, and policy documents all need to clearly explain the decision.
- Recent enforcement activity shows what is at stake. The May 2026 Pennsylvania AG settlement with GEICO highlights how communication-layer risk can surface when customers do not receive sufficient notice or explanation.
- Operationalizing explainable AI requires connected communications. Insurers need approved plain-language explanations, audit trails, and delivery records tied to the decisions behind each notice.
- MHC NorthStar CCM supports defensible communications at scale. Insurers can connect decision data to governed templates, approved content, workflows, and omnichannel delivery.
FAQs about Explainable AI in Insurance
What is explainable AI in insurance?
Explainable AI in insurance is the practice of ensuring that AI-supported decisions in underwriting, claims, pricing, fraud detection, or servicing can be understood, documented, justified, and communicated. It helps insurers show why a decision was made, what factors influenced it, and how it affected a specific policyholder.
Is explainable AI required by insurance regulators?
AI explainability is increasingly becoming part of insurance regulatory expectations. The NAIC Model Bulletin and related state activity emphasize governance, transparency, documentation, testing, and accountability for insurer use of AI, which means insurers need to be prepared to explain how AI-supported decisions are made and communicated.
What is the difference between AI transparency and AI explainability in insurance?
AI transparency is about visibility into the model, data, and governance process. AI explainability goes further by showing why a specific decision was made and how that decision can be explained to a regulator, auditor, or policyholder.
Where do most insurers have explainability gaps?
Many insurers focus on model-level governance, including bias testing, inventories, and documentation. The bigger gap often appears at the communications layer, where AI-supported decisions are translated into claim letters, policy notices, cancellation notices, disclosures, and other policyholder communications.
What types of insurance communications require AI explainability?
High-risk communications include claim denials, partial payment letters, policy cancellation notices, non-renewal notices, premium change notifications, and coverage eligibility communications. If AI contributed to the decision behind the communication, insurers should be able to trace and explain the decision in clear, compliant language.
How does MHC NorthStar CCM support explainable AI in insurance?
MHC NorthStar CCM helps insurers connect decision data to customer-facing communications through governed templates, approved content, workflow automation, audit trails, and omnichannel delivery. This helps carriers produce communications that are clear, consistent, traceable, and defensible at scale.
Sharon Malloch
Sharon leads content marketing at MHC, overseeing strategies that fuel sales and demand generation. With more than a decade of experience in customer communications, Sharon brings deep insight into customer pain points, industry trends, and the critical role of solutions in managing regulatory communications. Before joining MHC, she honed her marketing expertise at Doxim, Messagepoint, and OpenText.
