GenAI is quite unlike other forms of artificial intelligence. Many business leaders expect it to behave like conventional software, delivering a consistent set of outputs for the same inputs. However, rather than following a rigidly defined deterministic path, GenAI uses probability to generate outputs – which can vary.

To understand the difference, consider a scenario in which three academics are given an identical writing assignment. All three academics complete their papers with the same level of professionalism, but just like probabilistic GenAI, each paper is unique. This variance and non-repeatability are an important area of consideration for heavily regulated financial services firms, as GenAI may not be the appropriate tool when certainty is required to achieve compliance.

Other key risks associated with GenAI include:

• Data privacy and security: Generative AI models require large amounts of data to train and generate content. This data must be protected, anonymised, or consented – to comply with data privacy laws and regulations, such as the Privacy Act 1988 or the Notifiable Data Breaches scheme.
• Ethical and social implications: GenAI algorithms need controls or guardrails to ensure outputs are unbiased and accurate. For example, a GenAI tool may produce inaccurate or unfair credit scores, loan offers, or investment advice based on biased or incomplete data.
• Legal and regulatory compliance: For example, GenAI models may produce documents that fail to meet the standards of clarity, completeness, or disclosure required by the Australian Securities and Investments Commission (ASIC) or the Australian Prudential Regulation Authority (APRA).

It is important to note that GenAI models are trained on general knowledge across a wide range of topics. Therefore, GenAI models typically require additional training to understand financial services terminology and data. GenAI also relies on prompt engineering – a methodology for asking GenAI appropriately structured questions, designed to deliver the most consistent outputs.

However, GenAI excels at aggregating and summarising text, assessing large volumes of data, and providing cognitive search across large corporate knowledge repositories. When used in the right context, GenAI can deliver a step change in knowledge-worker productivity. For example, one EY client was able to double their high-risk customer capture rate thanks to GenAI.