Leveraging Meta Llama to solve business needs

KPMG is helping clients use Llama—Meta’s open-source large language model (LLM)—to solve a variety of business problems. Offering extensive training on diverse data, fine-tuning capabilities, adherence to ethical guidelines, continuous learning, and multilingual support, the model can support a wide range of use cases across industries. Explore the following real-life examples.

Internal Audit Assistant

Client: Global technology and social media company

Virtually every internal audit (IA) team is seeking ways to streamline time-consuming processes, such as preparing initial audit reports and performing ongoing audits of internal control effectiveness. These workflows are ripe for augmentation with generative AI (GenAI). A global technology and social media company’s IA team turned to KPMG to explore how Llama can support four specific IA use cases: Observation Generation, Report Creation, File Upload Summary and Analysis, and Scope Creation.

KPMG helped deploy a GenAI system that is being trained with historical reports and workpapers provided by the IA department. The goal: provide a smart “assistant” that generates high-quality output aligned with the established content structure and writing style.

Other companies can collaborate with KPMG and use Llama to explore opportunities to streamline audit report writing, increase content consistency and alignment to prescribed style guides, and empower individual auditors to work more quickly and efficiently.

Loan Summarization

At an American global multinational bank, the wholesale credit team had already adopted natural language processing (NLP) to accelerate extraction of hundreds of data elements per loan document. With the advent of LLMs, they recognized an opportunity to take automation and intelligence to the next level.

The bank worked with KPMG to deploy a tailored GenAI solution. The vision: use GenAI to empower human reviewers to more quickly and efficiently review and interrogate complex loan applications. We used our defined framework for training and fine-tuning to address complexities and optimize training within a secure environment. Llama demonstrated strengths in enabling generation of sophisticated loan document summaries, which support and streamline business decision-making.

Using Meta’s open-source LLM, other financial services institutions can increase the speed—without compromising the integrity—of their highly complex business processes.