AI has moved from experimentation to expectation in insurance. Recent research shares that 91% of insurers have or are planning to invest in AI, with many betting that automation will solve long-standing operational challenges. But adopting AI at scale brings a new layer of responsibility.
AI can scale processes at speed, yet it can just as quickly erode trust if it introduces inconsistency or decisions customers struggle to understand. The question for insurers is no longer whether to adopt AI, but how to modernize without weakening the human judgement and clarity that sit at the heart of the customer relationship.
Where AI Is Delivering Value
The progress inside insurance isn’t coming from futuristic use cases but from something far more pragmatic: efficiency. Across pricing, underwriting and claims, AI is finally doing what the sector hoped it would by helping teams make better decisions and reduce avoidable errors.
McKinsey reported how Aviva deployed over 80 AI models across its motor claims operation, cutting liability assessment times for complex cases by 23 days, improving routing accuracy by 30%, and reducing customer complaints by 65%. Earning more than £60 million in savings in 2024. Fraud detection is seeing similar improvements, as Deloitte estimates that multimodal AI could cut fraud-related claims costs by 20-40%.
In distributions the gains are also evident. Automated eligibility checks, smarter risk screening, and more accurate quotes reduce rework and help avoid disputes later on by ensuring cleaner data is captured at the start of the process. The teams seeing the biggest uplift tend to:
- Remove unnecessary steps in the pre-quote process, like duplicate questions or manual handoffs.
- Standardize partner data flows so carriers aren’t dealing with inconsistent inputs.
- Focus on communication over automation, so customers and brokers understand why the system is asking questions or flagging risks.
The most successful use of AI by insurers is by strengthening the basics rather than complicating them.
How Insurers Modernize Without Losing the Human Touch
The test insurers are currently facing is pairing automation with the clarity and reassurance customers still expect, especially when money, risk and stress are involved. Customers crave transparency. They don’t need to understand the inner workings of a model, but they want to know why their premium changed and what data informed that decision and what their options are if something doesn’t look right. Clear explanations will do much more to maintain customer confidence than technical disclosures ever will.
The 2026 State of Digital CX in Insurance report found that 70% of insurance customers say they still want human interaction during their journey, even if much of the process becomes digital. That preference is even more critical during high-emotion claims and for vulnerable customers.
There are certain moments where automation alone cannot substitute human judgment. For instance, claims involving major loss, sensitive customer segments or coverage disputes need someone who can interpret context, apply judgement and show empathy. Protecting those human touchpoints matters, as it’s where trust is built or lost. Insurers who communicate clearly and consistently see retention gains of around 20%.
However, this human-AI partnership is only possible when insurance employees understand the tools shaping the decision. This is where many insurers are feeling the strain. AI is moving faster than people can keep up with and staff can’t reassure customers or challenge unexpected outcomes if they aren’t confident in how the system works. That’s why building an AI-capable workforce becomes part of protecting the human touch.
Insurers need to provide hands-on training for team members working with AI. They need training on real dashboards, alerts, and simulated failures to understand how the model works in day-to-day scenarios. This exposure also shows them what AI can’t see, such as gaps in data and bias risk.
Insurers shouldn’t strive for complete AI fluency but rather AI interrogation. What’s most important is to teach staff to question model outputs. Therefore, having a human-in-the-loop as a safeguard means well-trained teams catch anomalies before they escalate to costly errors or regulatory scrutiny.
Embedded Insurance: The Next Experience Battleground
Embedded insurance is moving from a distribution offering to a core part of the customer experience. As journeys move online, protection is expected to appear naturally at the point where a customer is already making a decision, not as a separate task.
The two moments that shape this the most are checkout and first use. At checkout, customers are already comparing options and are more open to adding protection when it fits the flow. After the purchase, when someone begins using a product or service for the first time, reassurance carries far more weight, as knowing they’re covered from day one builds confidence and reduces second-guessing that can lead to cancellations.
AI strengthens this by tailoring context and adjusting coverage based on the type of property, asset, or risk profile without forcing the customers through extra steps. But the real test in embedded insurance happens when something goes wrong. If a claim feels slow and cumbersome, the benefit of a seamless purchase is gone. The companies with mobile-first claims and photo uploads and fast adjudication as standard practice will lead as customers expect the experience to match the promise made at the point of sale.
Behind the scenes, the difference comes down to partnerships. Embedded insurance only works when insurers, platforms, and partners operate with shared objectives. Effective partnerships spend time aligning the data already captured in the platform’s journey with the information an insurer needs to price or assess risk. When this is done well, eligibility checks shrink, quoting becomes cleaner and drop-off falls because customers aren’t asked to re-enter information they’ve already provided.
For example, property platforms that integrate embedded renters’ insurance through partners can use existing lease data—unit type, coverage requirements, occupancy dates—to pre-fill most of the information needed for a quote. This removes redundant questions and makes compliance easier for both the property manager and the resident. The customer experiences a simpler journey, and the insurer benefits from cleaner, more complete data from the outset.
The strongest partnerships will:
- Share data responsibility to ensure accuracy and reduce errors
- Co-design the journey, not bolt on insurance at the end
- Align KPIs so resolutions speed and customer satisfaction matter as much as conversion
When insurers and partners act as a single team, embedded insurance stops feeling like an upsell and functions as part of the service customers are already choosing.
The industry has moved beyond asking whether AI works. It does. The current issue lies in whether insurers can deploy it without weakening the judgement and empathy that customers rely on. The future is not AI-first or human-first. It’s a sector where the two are inseparable, and where insurers who balance speed with accountability will define the next standard of customer experience.
