What challenges did PwC face before AI implementation?

Manual review of regulatory texts was time-consuming and error-prone. Inventory safety narratives required significant manual effort, particularly in manufacturing's repetitive processes. Lack of automation in classifying inclusion and exclusion criteria led to inefficiencies in key business processes. Policy reviews took substantial time due to manual work and growing regulatory complexity.

Source

MicrosoftScaled productionEvidence: Medium60/100

PwC automates knowledge work processes for manufacturing and regulatory compliance

Use case typeMedical document automationUpdated Jun 13, 2026

PwC, a professional services firm in the UK, used Azure OpenAI Service to modernize knowledge work across multiple industries, with a particular focus on manufacturing and regulatory domains. The solution automates the classification of regulatory texts and generates automated safety narratives, providing significant touch-time savings in inventory management cases. PwC's multidisciplinary agile pods deliver generative AI solutions via a factory model, ensuring responsible AI implementation and governance. The approach encompasses ideation, data readiness, workforce enablement, model training, and deployment directly on Azure OpenAI, with continuous operational risk mitigation and governance. Key results achieved include up to 90% touch-time savings in manufacturing inventory cases, 50%+ reduction in policy review times, and an accuracy improvement in regulatory text classification and inclusion-exclusion criteria. The implementation is now live in production, with efficiency and accuracy benefits already materializing for both PwC and their clients.

Organization: PwC
Industry: Manufacturing
Location: United Kingdom

Reported outcomes

−50%

timeTime & speed

80-95%accuracy−90%time78%accuracy

Strategic outcomes

New product / capabilityAutomated regulatory text classificationNew product / capabilityGenerated automated safety narrativesScale & capacityDelivered AI use cases at scaleRisk & complianceEstablished responsible AI governance

Catalog median for time & speed deployments: −50% across 312 reported metrics. Compare benchmarks →

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Accuracy: 80-95%

azuremarketplace.microsoft.comUnknownInferred claimMedium evidence strength

Achieved 80-95% accuracy in regulatory rules identification.

Normalized claim

Time: 50% decrease

azuremarketplace.microsoft.comUnknownInferred claimMedium evidence strength

Reduced policy review times by 50% or more.

Normalized claim

Time: 90% decrease

azuremarketplace.microsoft.comUnknownInferred claimMedium evidence strength

Enabled approximately 90% touch-time savings for 250,000 manufacturing inventory cases annually.

Normalized claim

Accuracy: 78%

azuremarketplace.microsoft.comUnknownInferred claimMedium evidence strength

Enhanced accuracy and automation in inclusion/exclusion criteria tasks where no prior automation existed (up to 78% accuracy).

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: PwC
Provider: Microsoft
Maturity: Scaled Production
Linked source: azuremarketplace.microsoft.com

Applied a multidisciplinary, agile pod model to deliver and govern AI use cases at scale

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Automated regulatory text classification
2Automated safety narrative generation for inventory cases
3AI-based inclusion/exclusion criteria extraction

Manual review of regulatory texts was time-consuming and error-prone.
Inventory safety narratives required significant manual effort, particularly in manufacturing's repetitive processes.
Lack of automation in classifying inclusion and exclusion criteria led to inefficiencies in key business processes.
Policy reviews took substantial time due to manual work and growing regulatory complexity.

Deployed Azure OpenAI Service to automate regulatory text classification.
Implemented generative AI to create automated safety narratives for inventory, reducing manual touch-time substantially.
Applied a multidisciplinary, agile pod model to deliver and govern AI use cases at scale.
Established a robust, tiered governance environment for responsible AI adoption and compliance across processes.

Technologies

Azure OpenAI Azure

Achieved 80-95% accuracy in regulatory rules identification.
Reduced policy review times by 50% or more.
Enabled approximately 90% touch-time savings for 250,000 manufacturing inventory cases annually.
Enhanced accuracy and automation in inclusion/exclusion criteria tasks where no prior automation existed (up to 78% accuracy).

Architecture

Multidisciplinary agile pods deliver use-case-specific generative AI solutions through a factory model leveraging Azure OpenAI Service. The approach spans ideation, data readiness, workforce upskilling, model training, integration, deployment, and ongoing governance and risk mitigation, all run and secured within the Azure cloud environment.

Sources & evidence2

Evidence: Medium60/100Evidence strength