Exception cases pulled analysts offline across 2.2M cases a year. Now agents investigate and recommend; humans decide.
A benefits administration provider serving about 12 million employees, dependents, and retirees for large U.S. employers, completing roughly 1 million procedures and 2.2 million cases each year across more than 2,000 vendors.
Most cases followed an automated happy path, but exceptions required an experienced analyst to research the issue, decide on action, execute it...
“Most cases follow what we call ‘the happy path.’”
Global advisory, insurance brokerage, and risk management solutions firm.
Collaborated with WTW to build its AI agent platform on Microsoft Foundry
Enterprise software, cloud infrastructure, and consumer electronics platform.
WTW's Case exception handling is part of this use case:
Related implementations across industries and use cases
Forms, photos, scans—15 minutes per claim, backlogs frozen every night. Now the system runs 24/7; handlers work the cases worth their time.
QA sampled 50 conversations—most went unscored. Four agents now route and score every interaction, exposing a routing failure no one saw.
Recruiters lost 40% of their time to manual scheduling. An AI agent now handles coordination, letting candidates self-book via chat.
Adjusters spent up to two hours manually reviewing hundreds of claims notes. Now, secure AI summarizes complex case files in minutes.
Every AI workflow meant engineering approval gates from scratch—employees held back from delegating. Cowork shipped with them baked in.
Phone scheduling took 3 days per candidate. An SMS agent now handles it end-to-end in 37 minutes, 24/7.
Manual underwriting and billing created operational bottlenecks. AI now automates submissions and policy checks, freeing brokers for advisory work.
Multiple staff, one spreadsheet, inconsistent logs. Staff now describe incidents; AI classifies each, improving with every game.
Equipment faults left technicians hunting for technical data while machines sat idle. A chatbot now puts the answer in their hands.
Exception cases pulled analysts offline across 2.2M cases a year. Now agents investigate and recommend; humans decide.
A benefits administration provider serving about 12 million employees, dependents, and retirees for large U.S. employers, completing roughly 1 million procedures and 2.2 million cases each year across more than 2,000 vendors.
Most cases followed an automated happy path, but exceptions required an experienced analyst to research the issue, decide on action, execute it...
“Most cases follow what we call ‘the happy path.’”
Global advisory, insurance brokerage, and risk management solutions firm.
Collaborated with WTW to build its AI agent platform on Microsoft Foundry
Enterprise software, cloud infrastructure, and consumer electronics platform.
WTW's Case exception handling is part of this use case:
Related implementations across industries and use cases
Forms, photos, scans—15 minutes per claim, backlogs frozen every night. Now the system runs 24/7; handlers work the cases worth their time.
QA sampled 50 conversations—most went unscored. Four agents now route and score every interaction, exposing a routing failure no one saw.
Recruiters lost 40% of their time to manual scheduling. An AI agent now handles coordination, letting candidates self-book via chat.
Adjusters spent up to two hours manually reviewing hundreds of claims notes. Now, secure AI summarizes complex case files in minutes.
Every AI workflow meant engineering approval gates from scratch—employees held back from delegating. Cowork shipped with them baked in.
Phone scheduling took 3 days per candidate. An SMS agent now handles it end-to-end in 37 minutes, 24/7.
Manual underwriting and billing created operational bottlenecks. AI now automates submissions and policy checks, freeing brokers for advisory work.
Multiple staff, one spreadsheet, inconsistent logs. Staff now describe incidents; AI classifies each, improving with every game.
Equipment faults left technicians hunting for technical data while machines sat idle. A chatbot now puts the answer in their hands.