In reading this article you will understand:
- How AI summarisation tools are being used to improve underwriting efficiency and accuracy in the protection market.
- Why the type of AI selected matters in a regulated financial services environment and why generative AI alone is not sufficient for underwriting decisions.
- How AI capability in underwriting could affect provider panel selection and the service your clients receive.
- The importance of staff training and trust-building in successful AI deployment and what this means for the broader industry.
Aviva’s AI underwriting tool has cut its pipeline by 75 per cent, halved processing times and is now delivering accuracy above 99.4 per cent, all within five months of launch. This was some of the information shared by Robert Morrison, Aviva’s Chief Underwriter, last week as a follow up to my article How Aviva Are Putting AI into Practice, to transform service and their staff job security in December.
We were following up as last week Aviva extended the use of their AI underwriting service from Life to Critical Illness. This gave me the chance to explore if the service delivered in practice, as it had in testing. The results to date are impressive.
The summarisation tool reads lengthy GP reports and pulls out the medical facts that matter. After 18 months of development, it went live for mortality cases in November. Its job is to remove the heavy lifting, so underwriters spend their time on judgement, not on searching through clinical notes. Morrison shared three numbers that tell the story:
- The Aviva’s underwriting pipeline for Life cases has reduced by around 75 per cent since November.
- The accuracy of the system is close to 100 per cent.
- The time it takes an underwriter to assess a complex case has halved.
For advisers waiting on decisions for clients with serious medical histories, those numbers should translate directly into faster outcomes.
To put the efficiency gain in context, the tool compresses reports, that previously averaged 78 to 80 pages down to around eight. That is not a marginal improvement. It is a fundamental change in how underwriters spend their time. Instead of reading through large volumes of clinical notes looking for the relevant detail, they now start with a focused summary and can click through to the source material if they want to verify anything.
Morrison identified a key challenge Aviva has recognised, which is an important wider lesson for the industry. Users need to learn to trust the AI they are working with. This is sound advice, but it also highlights the need for extensive testing of AI systems in advance and the need to understand the type of AI being used.
Before deployment Aviva carried out extensive testing that enables them to drive this level of certainty. You cannot expect to just put data into a generative AI tool and have it create 100% reliable analysis. Generative AI is optimised to prioritise user satisfaction over accuracy, which creates challenges in a regulated environment, it is crucial to choose the right type of AI for the right task and Aviva have been rigorous about this.
When you do have a system that has been suitably tested Morrison provided a great analogy. “How does an airline pilot land an aircraft in zero visibility? They have to trust their instruments. You have to trust your instruments here, guys.” He noted that Aviva had taken underwriters from an average of 78 to 80 pages down to about eight, and that “human beings being human, the temptation to go in and just have a look” was natural. But he said underwriters needed to “have confidence in the tooling” and start from the summary. Over time since the November launch, that checking behaviour has been declining as confidence has built. Aviva expected this pattern, and they are seeing it play out.
Aviva has now extended the tool to Critical Illness (CI) cover. The CI version went live on 18 March, just days before we spoke. Moving from mortality to CI required only minor adjustments, a few additional prompts to reflect what is material for CI underwriting. Accuracy on CI cases is already above 99.4 per cent with minimal extra input required.
Morrison expects the CI tool to deliver even stronger outcomes than the mortality version, because the underwriting teams are already familiar with the technology and their confidence is building with each week of use. That compounding effect matters. The efficiency gains are not one-off. They are accelerating.
Income Protection (IP) is next. Morrison confirmed that discovery work on extending the tool to IP underwriting will begin this month. This is the logical progression, and it means Aviva is on track to have AI-assisted summarisation across all three core product lines in the near future.
Aviva has also added a new AI Academy to its training infrastructure. This sits alongside the Aviva University, the partnerships with Cambridge and the Alan Turing Institute, and the data science function of almost 1,300 people. The Academy meets staff wherever they are, foundational learning for those new to the technology, advanced modules for those already proficient. The message is plain: build these skills and we will invest in helping you get there.
Morrison also outlined how Aviva is now thinking about AI through two strategic lenses. The first is internal: how to optimise the capability of AI within the business, controlling costs and enabling teams to do more with the same resources. The second is external: how to use AI to better support distributors and consumers. These are different problems requiring different solutions, and the fact that Aviva is segmenting them deliberately rather than treating AI as a single initiative suggests mature thinking.
It is important to recognise not every case lends itself to AI processing yet. Some GP reports still arrive on paper. OCR technology is improving but not all medical information is digitised. Morrison confirmed that Aviva is engaging with relevant parties across the industry on how the industry can get better digital access to medical records.
Such work is crucial across the market. Last week I presented the keynote at the Protect Association’s inaugural Future of Insurance event on the implications of AI for the Life insurance sector. Vicki Jordan, Head of AI Strategy, at the FCA also presented and demonstrated that the regulator is actively engaging with the industry on how AI will impact the market.
Five months on from the initial service launch, the data from Aviva is impressive. IP is an obvious next step for Aviva and the discovery process to add this to the underwriting service is already underway. The results Aviva have delivered are real, meaningful and should be applauded.
It is no secret that Aviva, like other insurers, have struggled with underwriting turnaround times in recent months, but the changes being delivered by this work are strategic and long term and position the company well for the future.
AI capability should certainly now be a factor to consider in any panel review process, especially for those where only a small panel of providers are being selected.
Aviva are not the only company doing impressive things with AI, Vitality deserve similar credit, but where are the others? Given my 30 years plus experience as an advice tech specialist I am watching closely who else is delivering real change using AI?
Hopefully other insurers will have a story to tell soon?
Further thinking and activity
To develop your understanding further, consider the following:
- Watch the video of Ian McKenna’s Keynote at the Empowering Advice Through Technology conference Please note this is a 45 minute presentation so you may want to think about when you have time to watch it.
- Consider how AI-driven improvements in underwriting turnaround times could affect your client recommendations. If one insurer can return a complex underwriting decision in half the time, does that change how you position them relative to competitors with similar product quality?
- Reflect on the distinction between generative AI and purpose-built AI tools discussed in this article. Does your firm have a documented position on how AI should be used in the advice process? If Consumer Duty requires you to evidence good outcomes, understanding how your providers use AI is becoming increasingly relevant.
- Think about the issue of digitised medical records raised in the article. How does the quality of information flow between GP surgeries and insurers affect your clients’ experience? Are there steps you can take to help clients provide better quality disclosures at the application stage?
- When you next meet with provider representatives, ask specifically about their AI roadmap. What is live today, what is in testing, and what is planned? The answers will tell you a great deal about which firms are investing for the long term.
