Prediction: Banks And Insurance Companies Will Vastly Improve The Extraction Of Value From Unstructured Data
- Mike Upchurch, VP of Strategy for Financial Services and Insurance at Domino Data Lab
- 09.12.2024 03:15 pm #UnstructuredData #BankingInnovation
In 2025, well-tuned LLMs will help someone like a commercial bank credit risk manager answer a question like "How many commercial real estate loans have foreign investor exposure from unstable regions?" in seconds. Previously, analysing hundreds of pages of contracts manually would take weeks.
More than 80% of banking and insurance data is unstructured. Traditionally, these industries extracted limited features from this data, storing them in structured formats for modelling. This approach was constrained by the limitations of pre-LLM technologies, leading to rule-based, rigid classifications.
LLMs have revolutionised this landscape, enabling context-aware, domain-specific, and highly accurate classifications and extractions. Tasks like named entity recognition and summarisation, once limited, are now powerful.
Agentic solutions will be delayed as regulations and compliance requirements become more demanding
The deployment of agentic AI solutions will be delayed as regulations and compliance requirements become more stringent in 2025. Regulators will mandate stricter frameworks focusing on transparency, accountability, and risk management, including enhanced documentation and explainability to address questions like "why and how was this decision made?" Institutions will need to demonstrate control over agentic AI decisions, such as credit approvals or operational choices, and safeguard against rogue actions. These requirements will likely push widespread adoption to 2026, when progressive institutions can scale agentic solutions and realise associated operational efficiencies.
Model risk management challenges will increase as a broader range of people will build more models than governance teams can manage
In 2025, model risk management issues will escalate as LLMs and AI coding assistants enable more employees to develop models, removing previous limits tied to specialised teams. This surge in model creation, along with increased regulatory requirements, will overwhelm traditional governance workflows, causing bottlenecks in approvals, delays in productionising models, and increasing operational and compliance risks. To address the resulting backlog and unlock value at the pace the business requires, banks will need to transform roles, processes, and technologies to scale governance and optimise model pipelines for innovation.
Banks are facing profitability pressures and will increasingly adopt AI for process optimisation
Banks face mounting profitability pressures driven by rising borrowing costs, compliance demands, digital transformation expenses, narrowing net interest margins, and competition from fintechs. To counter these challenges, banks will increasingly leverage AI in 2025 to automate repetitive tasks across compliance, customer service, HR, finance, and risk management. These solutions will streamline decision-making, optimise resources, and accelerate workflows, delivering significant cost savings while fostering innovation. However, the rapid adoption of AI will bring increased regulatory scrutiny and operational complexity. Banks with mature model development and governance platforms, cloud-first infrastructures, and well-organised data will gain a decisive advantage, pulling further ahead of competitors still struggling to modernise. As regulatory frameworks evolve and MLOps maturity becomes a competitive necessity, early AI adopters will lead the industry's transformation, capturing efficiencies and driving sustainable growth by year end.
Agentic solutions will be delayed as regulations and compliance requirements become more demanding
The deployment of agentic AI solutions will be delayed as regulations and compliance requirements become more stringent in 2025. Regulators will mandate stricter frameworks focusing on transparency, accountability, and risk management, including enhanced documentation and explainability to address questions like "why and how was this decision made?" Institutions will need to demonstrate control over agentic AI decisions, such as credit approvals or operational choices, and safeguard against rogue actions. These requirements will likely push widespread adoption to 2026, when progressive institutions can scale agentic solutions and realise associated operational efficiencies.
Model risk management challenges will increase as a broader range of people will build more models than governance teams can manage
In 2025, model risk management issues will escalate as LLMs and AI coding assistants enable more employees to develop models, removing previous limits tied to specialised teams. This surge in model creation, along with increased regulatory requirements, will overwhelm traditional governance workflows, causing bottlenecks in approvals, delays in productionising models, and increasing operational and compliance risks. To address the resulting backlog and unlock value at the pace the business requires, banks will need to transform roles, processes, and technologies to scale governance and optimise model pipelines for innovation.
Banks are facing profitability pressures and will increasingly adopt AI for process optimisation
Banks face mounting profitability pressures driven by rising borrowing costs, compliance demands, digital transformation expenses, narrowing net interest margins, and competition from fintechs. To counter these challenges, banks will increasingly leverage AI in 2025 to automate repetitive tasks across compliance, customer service, HR, finance, and risk management. These solutions will streamline decision-making, optimise resources, and accelerate workflows, delivering significant cost savings while fostering innovation. However, the rapid adoption of AI will bring increased regulatory scrutiny and operational complexity. Banks with mature model development and governance platforms, cloud-first infrastructures, and well-organised data will gain a decisive advantage, pulling further ahead of competitors still struggling to modernise. As regulatory frameworks evolve and MLOps maturity becomes a competitive necessity, early AI adopters will lead the industry's transformation, capturing efficiencies and driving sustainable growth by year end.
