Generative AI in accounting: Opportunities and risks to assess today

By Will Bible, Audit & Assurance Partner and Digital Transformation and Innovation Leader, Deloitte & Touche LLP

Talking points

• Generative AI (GenAI) technology holds tremendous value for audit and assurance.
  
• Along with these opportunities come risks, which must be proactively managed.
• A longtime innovator in AI, Deloitte has launched a new internal GenAI chatbot, “DARTbot,” to support the daily tasks of our Audit & Assurance professionals.

Our experience designing and implementing AI systems

Generative AI has been portrayed in the media as everything from the potential savior of the planet to a dystopian menace. GenAI, however, is simply a technological tool, like many others, which offers tremendous promise, together with various risks that humans must manage and mitigate to safely tap its potential.

As a longtime technology innovator, Deloitte has worked over the years to advance AI in the audit process and other areas of accounting. Our Deloitte Omnia audit technology, for example, incorporates AI to provide a differentiated audit experience. Through these development efforts, we’ve gathered a host of valuable insights about GenAI’s capabilities, risks, and opportunities.

Capabilities

What makes GenAI truly ground-breaking is its ability to generate novel initial content such as text, images, and code from existing content. GenAI can accelerate the creation of first drafts, summarize complex content, automate first reviews, conduct analyses, and even highlight risks. Although still in its infancy, the technology is already being used for a wide range of applications in the marketplace—from product ideation and prototyping to marketing text and audio, to training videos and social media posts.

Risks

Capitalizing on GenAI’s remarkable capabilities requires human oversight to manage privacy, security, legal, and behavioral risks. GenAI’s large language models (LLMs) synthesize so much data so quickly that it can be challenging to track its sources, making it difficult for humans to understand the validity of information it produces.

Hallucinations are a real risk. These phenomena—defined as responses that aren’t justified by the training data—range from subtle biases to plausible-sounding falsehoods randomly embedded in responses.

Finally, as with any new technology, malicious actors can devise novel techniques abusing GenAI for forgery, phishing, and fraud. While scammers use many technologies to carry out these acts today, GenAI can enable them on a mass scale.

Deloitte actively works to manage AI risks of our applications of GenAI and the quality, accuracy, and reliability of its content. For more information, read our Pulse blog: “Steps to promote trustworthy AI."

Opportunities and challenges for audit and assurance

Risks aside, GenAI continues to make tremendous inroads. Indeed, many of Deloitte’s largest clients are experimenting with or already using GenAI tools in their products and services. Auditors need to be aware of their clients’ use of GenAI in financial reporting, accounting, and the related internal controls. The technology will likely be embedded in audit and assurance processes because of its ability to increase productivity and usefulness for content generation, analysis, and summarization, among other applications.