The customer journey is being remade by AI

The way a UK customer experiences insurance, including how they are offered cover, what they pay, how a claim is assessed, and whether their loyalty is rewarded, is being remade by AI. The character of the relationship between insurer and customer is becoming something different.

This is not a future scenario. Aviva has deployed over 150 machine learning models in claims alone, saving £60 million in 2024. Allianz UK detected £174 million of fraud in 2025. Marshmallow has insured one million previously excluded drivers, at a USD2 billion valuation and a profitable £500 million revenue run rate. The market is scaling. But most firms have not deliberately decided what their AI customer journey should be. The decisions are being made anyway, through individual model deployments, vendor procurement choices, productivity targets and third-party data feeds. The aggregate is reshaping the relationship in directions the board has not signed off.

Four moments of the AI customer journey

AI is now central to four moments where the customer experience is most consequential. These are organising categories chosen for where the AI-to-customer interaction is most material.

1. Becoming a customer

The customer searches, compares, receives a quote, decides whether to disclose what is asked. AI is now central to pricing, eligibility filtering, quote generation, behavioural targeting, and aggregator-mediated discovery before any insurer's system is engaged.

2. Living with cover

The customer holds the policy. Data flows continuously through telematics, app behaviour and third-party feeds. AI uses this data to price renewals, target cross-sell, score risk and identify vulnerability. Most of this is invisible to the customer.

3. The crisis

Something goes wrong. The customer claims. AI assesses, scores, triages, sometimes settles. This is the moment that defines whether the cover was real, and where the customer is least able to negotiate.

4. Staying or leaving

AI prices the renewal based on the customer's likelihood of accepting, not solely their risk. Loyalty is rewarded or penalised by the model's read of who is unlikely to switch.

This assumes a customer is engaged enough to notice data collection, want explanations and challenge decisions. Many customers, particularly the elderly, financially vulnerable, digitally excluded and time-poor, are not.

Three possible end states

Across the four moments, the AI customer journey is producing three potential end states.

Default State: AI to the customer

What documented evidence shows is emerging across UK insurers today.

AI deployed in pieces. Each component improves a metric. The aggregate is a customer relationship characterised by opacity, asymmetry and exclusion at the edges. The 26% of financially vulnerable UK adults uninsured. The £250 to £280 ethnicity premium gap. The automated claim rejection the customer cannot challenge. This state is not a failed implementation. It is the commercially optimal state. The asymmetry is profitable. The aggregate creates the regulatory and reputational exposure.

Designed State: AI for the customer

A composite of best-practice patterns visible across UK firms.

AI deployed deliberately. Choices made at board level, not in operational silos. Insurer A’s human-decides architecture in claims. Insurer B’s alternative-data inclusion in pricing. Explainability practices emerging in pilots. No firm exhibits Designed across all four moments. The Designed state is a composite of best-practice fragments that do not yet co-exist in any one firm. Achievable. Not yet operating end to end.

Participant State: AI with the customer

The direction the combined regulatory framework appears to be pushing toward.

The customer holds a real-time picture of what data is being used and can challenge any element. The right to human review under Data Use and Access Act 2025 (DUAA 2025) is operationally real, not just stated. Consumer Duty is a design constraint, not a retrospective check. Senior Managers and Certification Regime (SM&CR) accountability is mapped and evidenced. Not yet implemented at scale by any UK insurer. The framework infers this direction from published documents; regulators have not published a 'customer participant' framework themselves.

Please get in touch with your usual Crowe contact for further information.