Our recent report on The drive towards digital ONFR management revealed almost half (45%) of financial firms are using at least one of the mainstream genAI assistants to enhance their data-driven risk management strategies. The same number of firms are planning to implement operational risk applications focused on Large Language Models (LLMs) within the next 18 months.
Additionally, firms now see data as a key building block for successful risk management, with 95% stating that they have centralised, or are in the process of centralising, their operational and non-financial risk (ONFR) relevant data and pipelines to support their plans – typically at firm/group or enterprise risk level.
Key findings from the survey include:
GenAI assistants as an everyday tool
The intention behind using genAI assistants is to increase personal productivity and support day-to-day business activities. Just under half (45%) of those surveyed said that ONFR functions are already using at least one of the mainstream genAI assistants. This number increases to 64% for those firms working on rolling out tools in the next 12-18 months.
The most popular genAI assistant tool is currently Open AI ChatGPT at 35% of those surveyed, followed by Microsoft Copilot (22%), Amazon Q (13%) and Google Gemini (7%).
Figure: Use of genAI assistants for ONFR purposes 
There are some differences in their areas of application. Co-pilot links to the MS Office suite and is often used to support document summarisation, automate note taking or preparing data analysis. Examples seen include improving control descriptions, de-duplication of controls, and Know Your Customer (KYC) profiling.
ChatGPT has a wider number of applications, including document summarisation and key information extraction; content creation such as drafting of documents and code writing; and the creation of chat bots to provide faster access to internal risk information, such as policies or technical documentation.
Other GenAI assistance, such as Google Gemini, Amazon Q or Mistral, are used less and can be linked to quality assurance of internal documents and code assistance.
LLM adoption on a small scale
Of the firms surveyed, 45% are using LLMs in application development, and encouragingly, 77% of firms have short or long-term plans to implement the use of LLMs.
Figure: Firms currently building, training or deploying their own applications using Large Language Models
Only 14% of surveyed firms currently use LLMs as part of an application in production. However, 45% are using LLMs in application development, and encouragingly, 77% of firms have short or long-term plans to implement the use of LLMs.
Half of those surveyed have a formal or informal process for involving stakeholders in machine learning initiatives in place, suggesting adoption on a relatively small scale.
Members can read the full report to see more insights.
