PwC Generative AI Knowledge Modernization

kiadó: Pricewaterhousecoopers LLP

Creating advantage through knowledge work by enabling GenAI for industry and business use cases

Securely and responsibly transform knowledge work to a competitive advantage.

PwC and Microsoft are using Microsoft's OpenAI services to create the future of knowledge work by enabling Generative AI (GenAI) for specific industry and business use cases built on a foundation of trust.

Focused on responsible AI since 2017, PwC’s Responsible AI Toolkit was ranked as a 2020 World Changing Idea by Fast Company. PwC has also been ranked as an AI Consulting Leader by Forrester for the past four years.

PwC provides a lifecycle of offerings on Azure from ideation to execution:

  • mapping client challenges
  • assessments & collaborative proof-of-value workshops
  • getting your data ready for GenAI
  • preparing workforce & building skills for GenAI
  • conducting discovery & model training
  • executing delivery integration & deployment
  • establishing operations that mitigate risks
  • applying controls and governance to build trust

We are reimagining the way we work with Azure OpenAI Service, internally and with our clients. GenAI is, today, amplifying knowledge work with unprecedented efficiency, helping you scale further, work faster, reduce costs and enable new business models.

What is GenAI and what does good governance look like?

GenAI is a subset of deep learning that involves training a model to generate new data. PwC has built trusted solutions on Large Language Models (LLM) like GPT on Azure OpenAI that are in live production.

Our governance approach includes AI experts - including data scientists and engineers - and technology practitioners - alongside prompt engineers, business analysts and leaders for each use-case pod. In a controlled, tiered environment, they build tools on Azure OpenAI Services like CoPilot, to enable capabilities for your broader workforce, making sure your data remains trusted and secure.

Real-world use cases

PwC is actively transforming knowledge work to help drive value realization for enterprise businesses and their IT groups using deep industry use cases powered by Azure OpenAI Services. Here are some of the results we are currently seeing with our clients:

  • Regulatory Text Classification in financial services Regulatory rules are correctly identified ~80-95% of the time and policy review time can be reduced by more than ~50%. Use cases include ESG, tax and audit reporting.
  • Safety Automated Narrative Generation in manufacturing ~90% touch-time savings for 250,000 cases of inventory a year.
  • Inclusion Exclusion criteria generation across industry Our approach performs at an accuracy level of ~78% where no automation existed previously.
  • Negation Detection in life sciences Generating templated clinical trial inclusion exclusion criteria to help drive patient enrollment and trial success, improving the precision of one model pipeline from ~73% to 94%
  • Insurance pre-authorization in healthcare effectively create pre-authorization letters with optimal approval rates for patient insurance claims, targeting more than ~70% time savings
  • Text Summarization and synthesis for Marketing Content Reduction in human-hours that goes into enhancing search engine metadata on a monthly basis, as the AI generated outputs can serve as a starting point.

Delivering GenAI at scale

PwC’s Generative AI Knowledge Modernization is a pioneering factory model. Multidisciplinary agile pods develop specific business and industry-led uses cases built on a foundation of responsible AI to help drive efficiency while enabling good governance. This model is being applied within PwC as a first use case and replicated with our clients. We are building out solutions on GPT that are live in production, built for our regulated industry.

With PwC’s factory model, Azure OpenAI has the potential to help drive 50% to 90% productivity improvements in repetitive tasks such as knowledge capture, process automation, code creation and content generation.