How AI slashed the cost and turnaround-time on certificates of insurance for improved CX


A top global insurance broker specializing in risk mitigation products needed to streamline its process for handling certificate of insurance (COI) requests. With approximately 2.5K COI requests coming in daily, adding up to 0.5M emails a year – each potentially containing hundreds of pages of data – manual processers were overwhelmed. By 2020, the company was employing nearly 190 document processing users (DPUs) at an annual cost of about $4.7M. Even with that investment, errors abounded and delays tested customer patience.

In May 2023, the firm engaged EXL to apply its AI-enabled content extraction technology, EXL XTRAKTO.AI™, to improve its COI function. EXL had been working with the firm since 2018, using AI-led digital transformation services to improve other processes with convincing results. The goal of the engagement was to reduce COI handling times, lower the firm’s cost to serve, free valuable staff for other strategic business priorities, and provide a better customer experience (CX).


Having already engaged with the company since 2018, EXL had an intimate knowledge of how the firm worked. The solution deployed by EXL involved configuring existing and creating new models, as well as using open source APIs to implement the necessary integrations for automated COI operations.

The solution’s features included:

Salient Features of the solution:

  • The solution is hosted as a service, meaning that any application within the client environment can theoretically send an extraction request and retrieve the results using APIs
  • Highly reusable components that can be leveraged across intelligent document processing use cases within client environment, including:
    • A document pre-processing module that leverages a custom-built industry grade services such as optical character recognition, table detection, and other components to prepare incoming request into formats that can be used by the model
    • Document classification models featuring reusable machine learning components with the ability to identify relevant documents and also classify by document type
    • Extraction models for the intake of relevant information from documents
  • Serverless architecture for all computing requirements
  • Auto-scaling in or out endpoints depending on demand


By July 2023, the EXL XTRAKTO.AI™ solution had proven itself once again for the client, removing frustrations in the process for both DPUs and customers, and satisfying all project goals:

  • ~50% efficiency for DPUs issuing COIs associated with email requests
  • ~5-10% improvement in error rates and rework
  • ~21.5 minute average handling time per COI request, down from 41 minute average
  • Single screen validation replacing 132 screen manual request processing
  • ~25-step completion process vs. 40 steps manually
  • ~8.5 minutes spent on AHI research and processing, down from 17.5 minutes prior

The client has decided to leverage this solution as an enterprise-wide capability due to the reusability of the models and EXL’s comprehensive prebuilt and pretrained library of models. To learn more about the EXL’s XTRAKTO.AI™ solution or to request a demonstration, please visit