Legal & General use AWS AI to transform Customer Experience, Speed, and Productivity

L&G’s UK RRD has 5.2 million workplace pension members who send in more than 1 million paper documents per year, requesting, for example, a change of address or beneficiary. The division must process and respond to each one within a set period. It wanted to make the process more efficient, scalable, and cost-effective. “Using AWS, we now have a much more robust, elastic, and event-driven architecture,” says Tariq Surty, IT director for the RRD at L&G.

The UK RRD, alongside its local AWS team, created an internal proof of concept to ratify use of Amazon Textract, a machine learning (ML) service that automatically extracts text, handwriting, layout elements, and data from scanned documents. The division then engaged AWS Professional Services—a global team of experts that helps companies use AWS—to complete its project quickly and reliably, from designing the proof of concept to scaling into full production.

L&G’s cloud-focused solution for document management is Docusort. It processes documents with ML using Amazon Textract and Amazon SageMaker, a fully managed service that brings together a broad set of tools for high-performance, low-cost ML. Docusort classifies documents, extracts the relevant datasets, and pushes that data to backend systems to update customer information.

Using AWS, L&G can now process 1,000 documents concurrently in 30 minutes, and with the serverless architecture, its solution can scale up and down to meet any workload. Before, processing was limited by the number of people handling the documents. With these time savings, the UK RRD meets SLA requirements more accurately and consistently. As a result, the division has been able to redeploy five employees, improving business productivity.

L&G plans to use its Docusort solution as a template for other use cases where ML might be useful. In particular, the company is exploring how to apply generative AI to processes such as managing complaints, supporting the underwriting process, and analyzing customer feedback.

“We’ve learned how to manage, train, and maintain ML models, and we’re in a great place to use generative AI for new use cases now that we have a proven modular, elastic, and serverless architecture, which is really exciting,” says Surty.