The implications of generative AI for financial services are huge, say analysts at Andreessen Horowitz in this timely and educational report. Financial institutions are already using AI to synthesize massive amounts of data to make better decisions in credit allocation, risk management, compliance and reporting, but generative AI will be a game-changer for the industry. However, the authors warn, the human element will remain critically important in ensuring AI works for the benefit of firms and customers alike. This astute recap is required reading for financial professionals and executives.
- Historical financial data will be rich fodder for generative AI.
- Generative AI offers financial institutions opportunities in five major areas: clients, operations, compliance, risk and forecasting.
- Notwithstanding AI’s benefits, financial institutions will face real challenges in ensuring the technology works properly.
Historical financial data will be rich fodder for generative AI.
The ability to harness large language models (LLMs) to learn and process reams of unstructured data, along with enormous computing power, has given a major boost to financial services providers. While most financial firms are already using existing data to predict outcomes and classify risk, generative AI has the potential to create fresh content for clients and providers.
“Unlike other platform shifts – internet, mobile, cloud – where the financial services industry lagged in adoption, here we expect to see the best new companies and incumbents embrace generative AI, now.”
When it comes to generative AI, both legacy institutions and fintech newcomers have competitive advantages and disadvantages. The former can rely on a massive amount of existing financial data, but they will face issues in regard to accuracy and privacy concerns. The latter may need to resort to publicly available data to teach their LLMs, but they should be able to generate proprietary data rapidly.
Generative AI offers financial institutions opportunities in five major areas: clients, operations, compliance, risk and forecasting.
Firms will use their enormous repositories of data to address five major areas:
- “Personalized consumer experiences” – The existing technology fails to take into account the contexts and heuristics in which humans make financial choices. For example, prioritizing financial obligations during a hardship requires personal inputs to help make the best decisions for a customer. A similar challenge arises in tax planning and wealth management, in which human subtlety allows for more tailored client advice than AI can currently provide. Generative AI could help optimize financial decision making.
- “Cost-efficient operations” – In an ideal world, a bank should be able to respond immediately when an existing customer needs a mortgage or a loan. But consumer information, typically dispersed across multiple databases, takes time and human intervention to access. Also, many financial decisions are “emotional purchases” that resist full automation, and regulatory issues often require a staffer’s involvement. Generative AI can make individualized assessments of a customer’s history, allowing more specialized LLM-trained models to address concerns and needs as well as shoring up data collection and flagging compliance issues.
- “Better compliance” – The application of AI to compliance could effectively put an end to money laundering and the financing of illicit activity worldwide. AI can enhance data screening, detect patterns of malfeasance and ensure adherence to regulatory policies, alleviating what has become a burdensome and expensive process for financial institutions.
- “Improved risk management” – Generative AI could more rapidly identify and mitigate liquidity and operational risks through predictive analytics, data integration and comprehensive market overviews.
- “More dynamic forecasting and reporting” – Generative AI could streamline corporate finance departments through the better use of existing tools: For example, pattern detection of data in spreadsheets could prompt new ideas for areas of analysis. Forecasting models, report generation and responses to tax code questions would benefit from greater automation and precision.
Notwithstanding AI’s benefits, financial institutions will face real challenges in ensuring the technology works properly.
To realize generative AI’s true potential, both traditional financial institutions and newcomers will need to train their LLM models with proprietary data and use cases, as well as publicly available information. The edge may go to newer financial services providers, as legacy institutions tend to be slower when it comes to adopting new platforms.
“While we don’t yet know who will emerge victorious, we do know there is already one clear winner: the consumers of future financial services.”
Given the importance of financial transactions to individuals and companies, AI output must be highly accurate and specific in its recommendations. Human verification will be critical, at least initially.
About the Authors
Anish Acharya et al. are professionals at Andreessen Horowitz.