Meta Global AI Developer Hackathon. $1M in prizes.Apply Now
LLAMA 3.1 8B INSTRUCT
PwC
90%+
accurate in extracting fields70%+
of documents are processed straight through60-70%
reduction in processing costs~800
hours per month saved*All results are self-reported and not identifiably repeatable. Generally, expected individual results will differ.
CASE STUDY
Bringing breakthrough intelligence to document processing
At a glance
Industry
Professional servicesUse case
Automating accounting functions with intelligent document processingGoal
Improve speed and accuracy to reduce manual rework and eliminate bottlenecksLlama version
Llama 3.1 8B InstructDeployment
On-premises and Amazon Web ServicesTHEIR STORY
Empowering organizations to succeed
PwC provides professional services across audit and assurance, advisory and tax to help organizations solve problems faster, achieve more and realize greater value. Driven by its mission to build trust in society and solve important problems, PwC capabilities include cloud and digital, deals, sustainability, cybersecurity and regulatory, governance/boards, risk, transformation and more.
THEIR GOAL
Reinvent data extraction with generative AI
Accounting depends on massive quantities of documents, like invoices and purchase orders, that have long been digitized using optical character recognition systems. These systems require manual correction that slows workflows and increases overhead. PwC began developing an AI document extraction solution, Digital Workmate. But early versions that used proprietary models came with high costs, limited fine-tuning and a lack of support for local hosting to keep data secure. PwC needed a cost-effective, lighter-weight model they could retrain and deploy on any hardware.
THEIR SOLUTION
Llama transformed Digital Workmate into an intelligent partner
PwC seamlessly integrated Llama 3.1 8B Instruct into its Digital Workmate platform. The open-source model supported deep fine-tuning on unique financial terms for better accuracy. Prompt engineering shifted most reasoning and logic to pre-programmed routines, allowing PwC to get highly accurate results from the smallest model in the 3.1 generation — and keep computing costs low.Because the model is completely portable and doesn’t charge per-token inference fees, PwC could deploy Digital Workmate to on-premises hardware and private cloud instances for hundreds of clients securely and without incremental software costs.
PwC’s Digital Workmate with Llama provides highly accurate, multilingual document processing to clients across industries.
THEIR APPROACH
Tuning performance for unique domains without retraining
PwC knew a one-size-fits-all approach to document processing wouldn’t serve their unique clients. But fine-tuning for each client would require too much time and cost. Instead, PwC combined an ontology manager, a retrieval augmented generation (RAG) workflow, and a dynamic prompt engine to generate tailored instructions for Llama based on predefined templates and contextual data. Llama then extracts, formats and validates the information, delivering structured outputs for downstream applications and clients.
Digital Workmate combines rule-based ontologies with a RAG workflow to generate dynamic prompts for each document scan.
THEIR SUCCESS
Enhanced document processing automation — and business productivity
Pairing Digital Workmate with Llama 3.1 8B has reinvented intelligent document processing. The result is faster processing times, higher accuracy and less human review. For PwC’s clients, everyone from purchasing to the CFO gets better information faster, which translates into productivity and bottom-line benefits.
- 90%+ accurate in extracting fields
- 70%+ of documents are processed straight through
- 60-70% reduction in processing costs
- ~800 hours per month saved
“Through our evaluation, we discovered that Llama offers the right balance between inference performance, accuracy and on-premise resource footprint. Its extended context window and strong task comprehension make it well-suited for accurately extracting and interpreting both structured and unstructured data in high-volume document processing environments.”
Ritesh Jain — Partner, Agentic Automation Practice, PwC India
“Beyond measurable gains in accuracy, cost and time efficiencies, Llama and Digital Workmate reinforce PwC’s position as a market leader in applied generative AI and document intelligence. Llama has accelerated experimentation and rapid prototyping across client engagements while empowering PwC teams with cutting-edge generative AI tools, fostering a culture of innovation and continuous learning.”
Ritesh Jain — Partner, Agentic Automation Practice, PwC India
“Our collaboration with Meta on Llama underscores our belief in the power of open-source innovation. It’s a step toward making AI more accessible, secure and truly transformative — especially for businesses navigating complex regulatory and operational environments.”
Manpreet Singh Ahuja — Chief Client and Alliance Officer, PwC India
Previous slide
Next slide
Start building
Explore more
ConsumerBusiness
Shopify | Llama case studiesShopify uses Llama to generate product pages, localize content, and automate support, helping developers scale workflows and save time.Read more
ConsumerTech
Scribd, INC | Llama case studiesDelivering faster, cheaper and more accurate results with a Llama-powered AI content discovery assistant.Read more
Stay up-to-date
Our latest updates delivered to your inbox
Subscribe to our newsletter to keep up with the latest AI updates, releases and more.