PwC and AWS Develop Ethical AI Utilizing Automated Reasoning on Amazon Bedrock

This guest post is co-authored by Scott Likens, Ambuj Gupta, Adam Hood, Chantal Hudson, Priyanka Mukhopadhyay, Deniz Konak Ozturk, and Kevin Paul from PwC

Organizations are integrating generative AI technologies while ensuring accuracy, security, and compliance. In an increasingly competitive global landscape, speed is becoming more critical than scale, and innovation is paramount, as highlighted in PwC’s recent 2025 business insights on AI agents. To preserve a competitive edge, organizations must facilitate quick deployment and establish verifiable trust in AI outcomes. In regulated sectors, mathematically validating results can shift the speed of innovation from a possible liability into a competitive strength.

This post outlines how AWS and PwC are creating new reasoning checks that merge deep industry knowledge with Automated Reasoning checks in Amazon Bedrock Guardrails to foster innovation. Automated Reasoning is a field of AI dedicated to the algorithmic search for mathematical proofs. Automated Reasoning checks in Amazon Bedrock Guardrails, which convert knowledge into formal logic to verify if outputs from large language models (LLM) are plausible, became generally available on August 6, 2025.

This updated guardrail policy maintains accuracy within specified boundaries, distinguishing it from conventional probabilistic reasoning techniques. The system assesses AI-created content against guidelines extracted from policy documents, which include company procedures and operational standards. Automated Reasoning checks yield findings that indicate whether the AI-generated content complies with the extracted rules, identify ambiguities present in the content, and suggest how to eliminate assumptions.

“In a domain where breakthroughs occur at remarkable speed, reasoning is among the most significant technical advancements to enable our joint customers to thrive in generative AI,” remarks Matt Wood, Global CTIO at PwC, during AWS Re:Invent 2024.

Transformative industry use cases utilizing Amazon Bedrock Automated Reasoning checks

The strategic partnership between PwC’s extensive expertise and AWS’s innovative technology aims to change the way businesses engage in AI-driven innovation. The diagram below illustrates PwC’s implementation of Automated Reasoning, focusing initially on heavily regulated sectors such as pharmaceuticals, financial services, and energy.

In the subsequent sections, we will showcase three innovative use cases crafted by PwC teams.

Compliance with the EU AI Act for financial services risk management

The European Union (EU) AI Act mandates that organizations classify and verify all AI applications based on predefined risk categories and governance standards. PwC has devised an effective strategy to tackle this challenge via Automated Reasoning checks within Amazon Bedrock Guardrails, thereby turning EU AI Act compliance from a tedious task into a systematic and verifiable procedure. Given a description of an AI application’s use case, the solution transforms risk classification criteria into distinct guardrails, allowing organizations to continuously evaluate and monitor AI applications while supporting expert human assessment through automated compliance verification with auditable records. The major advantages of utilizing Automated Reasoning checks include:

• Automated classification of AI use cases based on risk levels
• Auditable proof trails for AI-generated classifications
• Improved efficiency in identifying necessary governance controls

The diagram below depicts the workflow for this use case.

Review of pharmaceutical content

PwC’s Regulated Content Orchestrator (RCO) is a globally scalable, multi-agent system—powered by a central rules engine tailored to company, region, product, and intended use—that automates compliance with medical, legal, regulatory, and brand standards. The RCO team was an early collaborator on Amazon Bedrock Automated Reasoning checks, integrating them as a secondary validation layer in the marketing content generation process. This fortified defense bolstered existing content controls, leading to quicker content creation and review processes while enhancing compliance standards. Key benefits of Automated Reasoning Checks in Amazon Bedrock Guardrails include:

• Utilizes automated, mathematically driven safeguards for validating RCO’s analyses
• Facilitates transparent QA with traceable, audit-ready reasoning
• Protects against potentially unsupported or fabricated outputs

The following diagram illustrates the workflow for this use case.

Management of Utility Outages for Real-Time Decision Support

Utility outage management employs Automated Reasoning checks within Amazon Bedrock Guardrails to improve response times and operational efficiency for utility firms. The solution can produce standardized protocols from regulatory standards, develops procedures in accordance with NERC and FERC requirements, and validates AI-generated outage classifications. Via a unified cloud-based architecture, this solution applies severity-based verification workflows in dispatch decisions—normal outages (3-hour target) assign tickets to available teams, medium severity (6-hour target) activates expedited dispatch, and critical incidents (12-hour target) trigger emergency protocols with proactive messaging.

The main advantages of using Automated Reasoning checks are:

• Improved and enhanced responses to customer needs
• Timely operational insights with confirmed regulatory alignment
• Faster decision-making supported by mathematical assurance

The following diagram depicts the workflow for this use case.

Looking Forward

As AI’s integration progresses, particularly with agentic AI, the AWS and PwC alliance is dedicated to:

• Expanding solutions with integrated Automated Reasoning checks across additional industries
• Developing industry-specific agentic AI solutions with built-in compliance verification
• Improving explainability features for enhanced transparency

Conclusion

The incorporation of Automated Reasoning checks within Amazon Bedrock Guardrails, coupled with PwC’s extensive industry experience, provides a powerful path for deploying AI-driven solutions. As an essential facet of responsible AI, Automated Reasoning checks deliver safeguards to enhance the reliability of AI applications. With the demand for mathematical assurance and verifiable trust in AI outputs, organizations can now foster innovation while upholding accuracy, security, and compliance. For further insights into how Automated Reasoning checks function, visit Minimize AI hallucinations and achieve up to 99% verification accuracy with Automated Reasoning checks: Now available and Enhance accuracy with Automated Reasoning checks in Amazon Bedrock Guardrails.

Discover how Automated Reasoning checks in Amazon Bedrock can augment the reliability of your generative AI applications. For further details on using this capability or discussing tailored solutions for your specific requirements, reach out to your AWS account team or an AWS Solutions Architect. Contact the PwC team to explore how you can harness the combined strength of AWS and PwC to foster innovation in your sector.

---

About the Authors

Nafi Diallo is a Senior Automated Reasoning Architect at Amazon Web Services, focusing on innovations in AI safety and Automated Reasoning systems for generative AI applications. Her specialization lies in formal verification methods, implementation of AI guardrails, and assisting global clients in building trustworthy and compliant AI solutions at scale. Nafi holds a PhD in Computer Science with a focus on automated program repair and formal verification, alongside an MS in Financial Mathematics from WPI.

Adewale Akinfaderin serves as a Sr. Data Scientist in Generative AI for Amazon Bedrock, contributing to pioneering innovations in foundational models and generative AI applications at AWS. His expertise encompasses reproducible, end-to-end AI/ML methodologies, practical implementations, and supporting global clients in developing scalable solutions for interdisciplinary challenges. He holds two graduate degrees in physics and a doctorate in engineering.

Bharathi Srinivasan is a Generative AI Data Scientist at AWS Worldwide Specialist Organization, working on solutions for Responsible AI. Her focus includes algorithmic fairness, the integrity of large language models, and explainability. Bharathi guides internal teams and AWS clients on responsible AI practices and has showcased her work at various educational conferences.

Dan Spillane is a Principal at AWS, leading global strategic initiatives within the Consulting Center of Excellence (CCOE). Collaborating with clients and partners, he addresses critical business challenges through innovative technologies. Dan’s expertise encompasses generative AI and responsible AI, including automated reasoning, as he applies these tools to deliver substantial business value at scale. As a lifelong learner, he actively studies global cultures and business practices, enhancing his capacity to mentor others and propel cross-cultural initiatives.

Aartika Sardana Chandras is a Senior Product Marketing Manager for AWS Generative AI solutions, emphasizing Amazon Bedrock. With over 15 years of product marketing experience, she is dedicated to helping clients navigate the complexities of the AI lifecycle. Aartika is passionate about empowering customers to harness potent AI technologies in an ethical and impactful manner.

Rama Lankalapalli serves as a Senior Partner Solutions Architect (PSA) at AWS, leading a global team supporting PwC, a major global systems integrator. Collaborating closely with PwC’s global practice, he champions enterprise cloud adoption by leveraging the vast range of AWS services, including migrations, modernization, security, AI/ML, and analytics. Rama architects scalable solutions that facilitate organizations in expediting their digital transformation while delivering tangible business results. His leadership merges deep technical expertise with strategic vision to drive customer success through innovative and industry-specific cloud solutions.

Scott Likens is the Chief AI Engineer overseeing Global and US teams at PwC, leading the AI Engineering and Emerging Technology R&D teams and shaping the firm’s strategy surrounding AI, Blockchain, VR, Quantum Computing, and other disruptive technologies. With over 30 years of expertise in emerging technologies, he has been instrumental in helping clients revolutionize customer experience, digital strategy, and operations across diverse industries.

Ambuj Gupta is a Director in PwC’s AI and Digital Contacts & Service practice based in Chicago. With over 15 years of experience, Ambuj possesses deep expertise in Artificial Intelligence, Agentic and Generative AI, Digital Contact Solutions, and Cloud Innovation across a wide array of platforms and industries. He is acclaimed for driving strategic transformation through Cloud Native AI Automation and emerging technologies—including GenAI-powered agents, Intelligent Agent Assists, and Customer Data Platforms—to enhance channel performance and employee efficiency.

Adam Hood is a Partner and the AWS Data and AI Leader at PwC US. As a strategic technology leader focused on results, Adam specializes in driving enterprise-wide transformation and unlocking business value through the strategic use of digital systems, data, and GenAI/AI/ML, including building agentic workflows. With a proven track record in industry and consulting, he has assisted organizations through complex digital, finance, and ERP modernizations, from initial strategy and business case development to seamless execution and global implementation.

Chantal Hudson is a Manager in PwC UK’s AI and Modelling team, having spent just over five years with the firm, beginning her career with the South African branch. Chantal primarily collaborates with large banks on credit risk modeling, showing a strong interest in utilizing AI to advance modeling practices.

Priyanka Mukhopadhyay is a Manager in PwC’s Cloud and Digital Engineering practice. An AWS Certified Solution Architect – Associate, she has over 13 years of experience in Data Engineering. Throughout the last decade, she has developed expertise in AWS services and has more than 12 years of experience in executing and delivering robust projects following Agile Methodologies.

Deniz Konak Ozturk is a Senior Manager within PwC’s AI & Modelling team. With approximately 15 years of experience in AI/Gen AI and traditional model development, implementation, and validation across UK and EU/non-EU jurisdictions, she’s well-versed in compliance assessment with EU regulations and IFRS9 audits. For the past six years, her focus has been mainly on AI/Gen AI, particularly in developing AI Validation frameworks, executing this framework for various clients, managing a platform for automated ML, and leading research and product ownership in an R&D initiative targeting the financially underserved segment.

Kevin Paul is a Director in the AI Engineering group at PwC, focusing on Applied AI and possessing extensive experience across the AI lifecycle, building and maintaining solutions across various industries.