Vendor reportedLive since Nov 2024Added Oct 2026

Markerstudy Group

Markerstudy Group uses Azure OpenAI to summarise claims calls and saves about four minutes per call

UK insurer Markerstudy Group built a call summarisation app on Azure OpenAI for its claims handlers, with a person checking each summary. Microsoft reports that around 300 handlers have used it and save about four minutes per call.

Key figures

Copied exactly from the sources. Copy any figure with its citation.

about four minutes

Time saved per call by claims handlers1

Around 300

Claims handlers who have used the app1

What it took

CostNot disclosed
TeamBuilt by Markerstudy with Microsoft engineers, after a dedicated generative AI innovation team was set up1
TimelineNot disclosed

The problem

Claims calls at Markerstudy Group are complex. Writing a summary of each call by hand takes time, and the quality of summaries varies between handlers.

What they built

A call summarisation app for the claims department that produces a summary of each call for the handler. It runs on Azure OpenAI.

How they did it

Markerstudy and Microsoft engineers built the app together, after the two companies approved a joint AI investment plan. Markerstudy then set up a dedicated team focused on generative AI. A person checks every summary for accuracy.

Results

Microsoft reports that around 300 claims handlers have used the app and save about four minutes per call. Markerstudy's chief architect says it also improved the quality of summaries. The method behind the figure is not given.

Lessons for your project

  1. Pick a task every handler repeats. Writing up a call is routine, so even a few minutes saved per call add up.

  2. Check every summary. Markerstudy keeps a person in the loop to confirm accuracy.

  3. Set up a dedicated innovation team after approving the plan, as Markerstudy did, so the build has an owner.

Caveats

  • The only source is a Microsoft customer story. The method behind the four minutes is not stated.
  • Microsoft's story also says this adds up to 56,000 hours on 840,000 calls a year. That is arithmetic on the per-call estimate, so it is not listed as a result here.
  • The story gives no launch date. deployedDate uses the date of the story, which is an announcement date.
  • The model is not named. Azure OpenAI serves OpenAI models, so the vendor is recorded as OpenAI and the model as undisclosed.

Sources

  1. Markerstudy Group embraces AI to drive contact centre innovation and excellence

    Microsoft UK Stories · 2024-11-26 · Vendor case study · ukstories.microsoft.com

Summarized in our own words from the sources above. Spotted an error? See corrections.