Artificial Intelligence at Aviva – Emerj Artificial Intelligence Research

Aviva is a British multinational insurance company headquartered in London, England. Primarily recognized as the UK’s leading diversified insurer, Aviva provides various products and services across insurance, wealth management, and retirement solutions. 

With 19.2 million customers spanning the UK, Ireland, and Canada, Aviva has positioned itself as a major player in the financial services industry. In its 2023 annual report, Aviva also claimed a gross written premium revenue of $14.04 billion, reflecting a 13.9% increase from the previous year. 

Per its 2023 press release, Aviva further committed $193 million to its venture capital fund, investing mainly in AI solutions and other emerging technologies that will improve the company’s workflow. According to an Amazon Web Services (AWS) blog post, Aviva uses machine learning and AI across more than 70 use cases.  

In this article, we will examine two of Aviva’s most significant AI use cases: 

• Improving motor insurance repair estimates: Adopting visual AI analytical damage assessments to accelerate resolutions and improve accuracy across large repair networks.
• Expediting claim management processing: Leveraging machine learning to automate claims and unify disjointed systems, ultimately streamlining settlements and informing employee decision-making. 

Improving Motor Insurance Repair Estimates 

Traditional motor insurance claim processes often suffer from manual assessments, inconsistent diagnoses, and time-consuming repairs. According to a press release by J.D. Power, the average auto insurance repair cycle time was 23.1 days in 2023, an increase of 6.2 days from 2022 and more than double from 2021. 

In a 2022 report from Accenture, 31% of claimants across the industry surveyed were not fully satisfied with their automobile insurance experiences. Among these claimants, 60% cited settlement speed issues.

A newsroom briefing confirms that Aviva experienced similar issues that prolonged auto claim resolution. The briefing further found that the auto industry posed mounting challenges to Aviva’s 2040 net-zero sustainability objectives. 

To tackle these challenges, Aviva partnered with Tractable, a UK-based tech unicorn specializing in visual AI for accident and disaster recovery. Tractable’s AI models leverage computer vision and deep learning to analyze photos of vehicle damage, generating accurate repair cost estimates within minutes. 

We found evidence from a Tractable briefing that this capability enabled: 

• Remote Assessments: Claims assessments can now take place remotely by using photos submitted by customers, eliminating the need for physical inspections by assessors
• Automated Damage Assessments: Deep learning and AI models conduct hyper-accurate calculations of repair costs 
• Repair vs. Replace Guidance: AI models identify parts that can be safely repaired rather than replaced, aligning with Aviva’s sustainability goals.

Screenshot from an Amazon Web Services (AWS) blog post on Aviva’s AI auto insurance workflows. (Source: AWS)  

According to an AWS blog post, the workflow entails the following steps:

1. The customer provides Aviva with information about the incident, photographs, and details about the damage
2. AI and ML models, armed with a set of business rules, process the request
3. The projected repair cost is evaluated against the vehicle’s current market value using data from external sources
4. Data on comparable vehicles available for sale in the local area is incorporated into the analysis
5. The model makes a recommendation to repair or write off the car, and the recommendation and supporting data are issued to the claim handler

In a video interview with Tractable, Aviva’s Motor Technical Manager Adam Murray specified that while AI will take charge on more straightforward auto claims, “technical experts will focus on scrutinizing complex claims to ensure safe compliance.”

Here, the capability helps free up staff to expedite complex claims while also facilitating personnel reassignment to other high-demand claim types.

Tractable’s capability is also poised to help Aviva meet its net zero goal by avoiding unnecessary journeys to in-person claim assessment and reducing the volume of new parts needed. According to a McKinsey case study, Aviva was able to triple the use of recycled parts. 

Because the technology was deployed very recently, no Aviva-specific performance metrics have been publicly disclosed. However, Tractable has claimed various business results for other companies that have adopted the technology: 

• Cutting time for manual claim processing from 15-30 minutes to under a minute
• Identifying 97% of all losses with 95% accuracy
• Driving an approximately 3.5x ROI

In 2019, Aviva also rolled out a similar technology, Vehicle Remedy Tool at First Notice Loss. The tool is fueled by real-time predictive analytics to provide intermediary actors, like adjusters, with an instant and accurate repair prediction before an appraisal. It further enables adjusters to continue a settlement if the vehicle is deemed a total loss on the first call. 

According to an article from the Insurance Journal, early identification of total loss vehicles opens opportunities to mitigate delays across salvage, rental, and settlement workflows. Furthermore, it helps prevent non-repairable vehicles from clogging up repair shops and slowing down fixing times.

In an award nomination article from Insurance Canada, Aviva holds that the analytical tool delivered: 

• Reduced customer cycle time on total losses by 59% 
• Reduced rental days average by 52%
• Greater salvage returns due to cycle time reduction

Expediting Claim Management Processing

With a growing customer base, Aviva encountered noticeable deficiencies in timeliness and effectiveness during claim management. Workloads were stretched to the limit, and in some divisions, virtually half of Aviva’s employees incurred negative customer feedback, according to a case study from Aviva’s partner Beyond Philosophy. 

Dataiku’s use case documentation finds nearly 40% of customer feedback lacked detailed explanations, making it even more difficult to diagnose technical problems.

The outcome was greater costs due to miscommunications, delays, and inaccuracies. These costs permeate the industry, as the Capgemini Research Institute’s World Life Insurance Report 2025 quantifies 25% of insurance customers are frustrated by long wait times.

In light of such issues, Aviva sought strategic partnerships with various companies, leveraging AI and machine learning to automate claim management.  

Yet as organization-wide AI adoption expanded, Aviva’s disjointed IT platform faced various scalability, interoperability, governance, and slow onboarding challenges. 

The company partnered with Dataiku and Appian to promote seamless AI workload management. Dataiku served as the linchpin of the solution by consolidating platform infrastructure, streamlining governance frameworks, and scaling compute resources. The platform enabled Aviva to establish a robust CI/CD and MLOps foundation, significantly reducing deployment times for AI platforms. 

Use case documentation from Dataiku claims that owing to these changes, Aviva halved runtimes for large-scale data processing tasks and machine learning scoring jobs. Meanwhile, Appian contends it was able to unify 22 of Aviva’s legacy systems to provide a unified platform for automation and a 360-degree view of customers.

With the infrastructure assimilated, Aviva joined forces with McKinsey, integrating over 80 machine learning and AI models tailored to the claims team’s needs. 

A case study by AiCore corroborates that these models are capable of:

• Analyzing Incidents: Using natural language processing to analyze incident descriptions and cross-referencing against policy specifics
• Determining Liability: Leveraging machine learning to flag scenarios that indicate at-fault patterns and routing the claim to the applicable team 
• Auto-Generating Draft Reports: Integrating generative AI to generate draft reports, recommend action, and keep claimants informed via email or text
• Measuring Performance Outcomes: Employing advanced analytics, engineers built an impact measurement framework with 50+ KPIs 

Screenshot from an Empeek briefing on claim processing automation and an example schematic workflow. (Source: Empeek)  

The McKinsey case study maintains that Aviva uses a “double helix” approach, whereby processes seamlessly switch between digital and human interaction. It further cites that claims involving personal injury default to human interaction, reflecting a balance between employing AI to expedite basic claims and humans to process more emotionally sensitive claims. 

A blog article from Appian claims that automating Aviva’s claims process in France resulted in:

• Same-day claims settlements increasing from 1% to 25%
• Claims settled within three days, an increase of 530%

The broader benefits, according to the McKinsey case study, arrived through three routes:

• Customer satisfaction soared, evidenced by a sevenfold increase in Net Promoter Score and a 65% decrease in customer complaints
• Claim processing was expedited, with the average time to asses liability in complex cases reduced by 23 days
• Routing accuracy was improved by 30%