Contract management · Production

PwC's AI-Driven Annotation (AIDA) on AWS reduces manual contract review time by up to 90%

The problem

Contract analysis consumes significant time for legal, compliance, and procurement teams because important insights are buried in lengthy, unstructured agreements. As contract volumes grow, keyword and pattern-based extraction methods and contract management systems fall short of providing consistent insights at scale.

First attempt

Teams relying on keyword and pattern-based extraction or contract management systems found these approaches unable to provide consistent insights at scale.

Workflow diagram · grounded in source
1
Document upload and storage
trigger
“After authentication, AIDA stores uploaded documents, Optical Character Recognition (OCR) outputs, and associated metadata in Amazon S3”
2
Async OCR processing
ai_action
“OCR and extraction workflows run asynchronously on Amazon ECS using AWS Fargate, with tasks coordinated through Amazon Simple Queue Service (Amazon SQS)”
3
Rule-guided LLM extraction
ai_action
“Extraction rules guide how relevant content is identified and sent to foundation models (FMs) hosted on Amazon Bedrock, where LLMs can interpret the contract text and extract structured values”
4
RAG-grounded Q&A
ai_action
“AIDA uses RAG to help verify that responses are grounded in the underlying contract text. Documents stored in Amazon S3 are embedded using Amazon Bedrock Embeddings Models, with vectors indexed in Amazon OpenSearch Serverless for semanti…”
5
Human-in-the-loop review
human_review
“A configurable human-in-the-loop review queue can validate and approve extracted outputs before they are forwarded downstream”
6
Downstream system integration
integration
“The structured extracted insights generated by AIDA can be quickly pushed to downstream systems such as Contract Lifecycle Management (CLM) tools, ERP systems, CRMs, or data warehouses”
Reported outcome

AIDA helped reduce manual contract review time by up to 90%, enabling teams to retrieve key information more quickly and shorten review cycles.
One major film and TV studio reduced rights research time by 90%.

Reported metrics
Manual contract review time reductionup to 90%
rights research time reduction (film and TV studio)90%
Reported stack
Amazon ECSAmazon S3Amazon RDSAmazon SQSAmazon OpenSearch ServerlessAWS FargateAmazon CognitoAWS WAFAWS LambdaAmazon EventBridgeAmazon Bedrock Knowledge BasesAmazon Bedrock GuardrailsAmazon CloudWatchAWS X-RayAWS KMSAWS IAMAWS CodeBuildAWS CodePipelineAWS CloudTrailAmazon Quick SightNGINXOCRRAGLLMsMicrosoft Entra IDOkta
Source
https://aws.amazon.com/blogs/machine-learning/extracting-contract-insights-with-pwcs-ai-driven-annotation-on-aws/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

AIDA helped reduce manual contract review time by up to 90%, enabling teams to retrieve key information more quickly and shorten review cycles.

What tools did this team use?

Amazon ECS, Amazon S3, Amazon RDS, Amazon SQS, Amazon OpenSearch Serverless, AWS Fargate, Amazon Cognito, AWS WAF, AWS Lambda, Amazon EventBridge.

What results were reported?

Manual contract review time reduction: up to 90%; rights research time reduction (film and TV studio): 90% (source-reported, not independently verified).

What failed first in this deployment?

Teams relying on keyword and pattern-based extraction or contract management systems found these approaches unable to provide consistent insights at scale.

How is this contract management AI workflow structured?

Document upload and storage → Async OCR processing → Rule-guided LLM extraction → RAG-grounded Q&A → Human-in-the-loop review → Downstream system integration.

Related contract management cases