The choices that will define your AI customer journey

Building on our previous insight, The customer journey is being remade by AI, we argued that AI is already reshaping the insurance customer journey. This is not happening through one single strategic transformation, but through the cumulative effect of individual model deployments across pricing, onboarding, policy management, claims and renewal.

We identified three possible end states for the customer journey: a Default Journey, where AI is applied to the customer in ways that optimise commercial outcomes but increase opacity, asymmetry and exclusion; a Designed Journey, where AI is deliberately deployed for the customer through board-level choices, explainability and outcome management; and a future Involved Journey, where customers have greater visibility, challenge rights and meaningful human review.

For insurers, the question is which version of that relationship they are choosing to create, and whether the choices, controls and accountabilities behind it are explicit enough to withstand regulatory, commercial and public scrutiny.

Below, we explore the strategic choices leaders need to address as AI becomes a more prominent part of the customer experience.

Three choices that determine which state your firm reaches

How does a firm end up in the Default Journey when nobody deliberately chose it? Usually, through three choices made repeatedly across the customer journey. Each affects customer outcomes, but each also has a commercial impact that can pull firms toward Default by default. Moving toward a Designed or Involved Journey therefore requires a deliberate counter-choice.

Choice one: Who AI serves

At pricing level, how granular should a model be allowed to become before precision turns into exclusion? Where does actuarially justified differentiation end and social sorting begin? At investment level, should AI capital and attention be directed towards optimising the profitable existing book, or towards extending access to customers the firm has not historically served?

The answers determine who gets to be a customer, on what terms, and who does not.

Commercially, precision can mean better selection and lower loss ratios. Investment in existing-book optimisation has a measurable return, while investment in market extension is often less certain. The commercial imperative therefore pulls firms toward more precise pricing of existing customers and away from extending access to new ones. At key journey moments, however, the same capability can produce very different outcomes. For example, alternative data and AI can be used to insure previously excluded drivers; left ungoverned, the same technology can also deepen exclusion.

By contrast, renewal pricing AI may identify customers who are unlikely to switch and price them higher than equivalent new customers. In that context, the underlying capability and the commercial incentive to use it can weaken trust in the fairness of the customer relationship.

“At what point does pricing precision stop being actuarially justified and start being social sorting, and can you point to that boundary in your model?”

Choice two: What AI hides

This choice concerns what a customer can see about how decisions are made. What data is being used to decide their price, claim or renewal? What can they challenge? Where does the firm’s information advantage become an information asymmetry the customer cannot navigate? The right to challenge under the Data (Use and Access) Act 2025 has limited practical value if the customer does not know what to challenge. Information advantage can create a competitive moat, while customer visibility costs money and creates audit exposure. The result is a commercial pull toward more sophisticated decisioning and less customer-facing visibility into how those decisions are made.

For example, Telematics pricing has been documented making consequential premium decisions on inaccurate data that customers could not see. Renewal pricing may be optimised on behavioural data the customer does not know exists. When AI rejects a claim, the customer may receive an automated message they cannot interrogate, while the model’s reasoning remains opaque. Within firms, the right to human review is still rarely operationalised at the scale at which AI operates.

“Does your AI improve outcomes for consumers or improve extraction from them, and do you have a framework that can tell the difference at model level?”

Choice three: Who answers for AI

This is the speed-versus-accountability question, the right-to-human-review question and the SM&CR personal liability question. In practice, they are all the same underlying choice: where does human judgement sit? When the model is wrong about an individual, who decides? When SM&CR is invoked, who is named? Are humans real reviewers, or nominal sign-offs?

Removing humans can produce faster decisions and lower cost-to-serve. Keeping humans in the loop can catch edge cases, but it carries operational cost. Once automation becomes embedded, commitments to human involvement can also weaken over time. The commercial gradient therefore pulls firms toward less human review, even as the Data (Use and Access) Act 2025 gives customers a right to it. Many firms have not yet built the infrastructure to deliver that review at the scale at which AI operates.

“Which AI decisions is each member of your ExCo personally accountable for under SM&CR, and can they evidence oversight of those systems today?”

The risks sit within and across insurance firms

The choices made by individual firms are not made in a vacuum. Some of the most important pressures on the AI customer journey sit above any one insurer. They come from shared public trust, common technology suppliers, shared data feeds and the role of aggregators in shaping how customers first enter the market.

Even a well-governed insurer may be exposed to risks it does not fully control. A customer may lose trust because of another firm’s AI failure. A flawed vendor model may affect several insurers at once. An aggregator may shape the customer’s options before the insurer ever sees them. These market-level dependencies change what responsible AI governance has to consider.

Trust

Trust in insurance is shared. If one firm has a visible AI failure, such as an unfair pricing outcome or an automated claims decision customers cannot challenge, the damage does not stay neatly with that firm. It can weaken confidence in insurers more generally. Each firm may have a commercial incentive to let AI optimise locally, but the whole market carries the reputational cost when the outcome is seen as unfair, opaque or poorly governed.

Vendor concentration

Many insurers rely on a relatively small group of technology vendors, data suppliers and platform providers. That can create efficiency, but it also creates concentration risk. If the same vendor model, data feed or triage engine contains a flaw, the problem may appear across several insurers at the same time. What looks like a firm-specific issue may actually be a shared market dependency that is invisible to any one firm’s risk register. This presents a market resilience concern.

Aggregator-mediated discovery

For many customers, the first AI system shaping their insurance journey may not belong to an insurer at all. It may belong to the aggregator that decides which products are visible, how they are ranked, which customers are eligible to see which offers and which placements are promoted. A significant part of becoming a customer therefore happens upstream of the insurer. The insurer may be held accountable for the customer relationship while having limited visibility over the AI systems that shaped the customer’s first choices.

Conclusion

This discussion is not intended to prescribe a single right answer. The appropriate choices will differ by firm, product, customer segment and operating model. The aim is to make those choices visible, so they can be addressed deliberately at the right level rather than emerging through accumulated operational drift.

Some firms will have arrived at their current position gradually, as individual AI deployments accumulated across pricing, onboarding, policy management, claims and renewal. Their priority is to recognise how those choices now shape the customer relationship. Others may have made conscious decisions that align more closely with the Default Journey. Their challenge is whether those choices remain defensible in a changing regulatory, commercial and societal context.

The key question, therefore, is not simply whether AI is being used. It is how AI is being governed in practice, who benefits from its deployment, what customers can see and challenge, and who is accountable when the model is wrong.

“Are you governing what your AI is actually doing, or what you think it is doing?”

For further information, please get in touch with your usual Crowe contact.