By Laura Carballo, Head of Advanced Analytics for South-West Europe and Lukáš Vermach,Director, ICT Insurance Management Consultancy Team, WTW
There will be opportunities at every stage of the value chain - from product design and marketing to underwriting, pricing, claims reserving and customer service. The most forward-thinking insurers are therefore developing roadmaps for AI transformation, working across every business unit to identify new use cases and to move forward with implementation.
Pricing and underwriting in focus
AI transformation is proceeding at pace in two areas in particular at the core of the insurance sector: pricing and underwriting. There are multiple applications here.
Improving risk and underwriting efficiency
AI enables insurers to dramatically improve their risk and behavioural models, detecting patterns and anomalies not previously visible in order to price with far more accuracy. It automates manual assessments with predictive models that deliver smarter decisions with greater speed to enhance underwriting efficiency.
Monitoring trends and portfolio performance
AI is also transforming monitoring work, with insurers now able to detect emerging trends in areas such as competitor activity, claims and customer behaviours. It supports more active portfolio management and provides insurers with intelligence on where to focus their efforts.
Governance, explainability and compliance
At the governance level, AI helps to address compliance, and insurers also now recognise the importance of responsible deployment. They are emphasising explainability, conscious of the need for bias detection, and ensuring human intervention.
Critically, insurers have started to put theory into practice, often with impressive results. For example, WTW worked with a large UK motor insurer that had become increasingly concerned about a rise in its lapse rates. Having identified the segment of the business where this rise appeared to be concentrated, it was possible to apply large language models (LLMs) to analyse transcripts from the insurer’s call centre. This identified a recurring theme in conversations with customers; they complained that a specific area of the insurer’s new business customer journey wasn’t working as intended, making it difficult to renew their policies.
It’s a good example of how AI can help insurers identify issues and opportunities that are hidden in plain sight. Using language embeddings to enable segmentation analysis, the LLMs were able to pinpoint the problem at the root of the increase in lapse rates. The insurer was then able to fix the issue at speed.
In another deployment, WTW worked with an Italian direct insurer that felt it should be making better use of its motor claims data to enhance underwriting, pricing and portfolio management. Using advanced modelling techniques to analyse this data more effectively, it proved possible to uncover new insights into risk patterns, these enabled the insurer to refine its underwriting and pricing activity.
The results observed so far are highly positive. The deployment is leading to better pricing, improved renewals and an enhanced loss ratio, and is becoming a key part of their active portfolio management process. The insurer also benefitted from improved underwriting workflows that increased operational effectiveness and profitability.
AI for success
As insurers enjoy these successes, their use of AI will inevitably grow, with each new model deployed running continuously to identify further refinements and improvements. Relatively quickly, insurers will end up with an evolving ecosystem of AI deployments capable of driving value.
That, however, brings a new challenge – the need to manage and monitor these models to ensure they deliver the maximum possible value. It’s the same challenge that a farmer planting multiple crops faces – each one has to be nurtured and maintained for the harvest to be optimised.
What insurers need here is a means to manage ongoing advanced analytics, pricing, underwriting and portfolio management. New technologies and services now coming on stream make it possible to automate analysis of insurers’ emerging AI experience so that their growing crops of models continue to yield relevant business insight rapidly and efficiently. Such tools monitor expanding model portfolios, surface early signs of deterioration, and ensure models remain reliable, compliant and aligned with business objectives.
In practice, that means monitoring and managing all the models deployed so that any segments where model health is deteriorating can be quickly identified and rectified. In an environment where insurers will be increasingly dependent on those models for competitive advantage, managing the uncertainties and complexities around them will be an ever-more critical task. It will go beyond governance and oversight, enabling better risk control and more active portfolio management.
The bottom line? The potential of AI to transform insurance is clear – businesses not taking advantage risk being left behind. But as deployments accelerate, insurers will also need to monitor and manage growing numbers of models in order to maximise the value they create. That will be a key task in the transformation challenge.
|
