The opinions expressed here are those of the authors. They do not necessarily reflect the views or positions of UK Finance or its members.

Agentic AI systems act as virtual coworkers, capable of executing specific roles, collaborating with human agents, and continuously learning to improve decision-making accuracy. This will represent a game-changer for those institutions that still maintain labour-intensive processes and a myriad of handovers during the credit assessment and lending execution. One prominent example is the automation of underwriting processes for secured and unsecured lending using AI-driven solutions. The objective is to reduce manual workloads, minimise errors, and accelerate decision-making. This involves the use of intelligent document processing to classify, extract, and organise data from underwriting documents, as well as automated credit memo drafting that summarises key decision-making inputs and identifies anomalies through AI-driven analytics.

To effectively implement AI, particularly Agentic AI, in lending processes, financial institutions should consider the following use cases:

1. Document gathering: Agentic AI can autonomously source and compile data from various internal records, credit bureaus, and third-party providers. These systems streamline the gathering of financial statements, tax returns, bank statements, and credit ratings, ensuring data consistency and reducing manual effort. AI agents can monitor for missing documents and request updates, thus accelerating the document collection process.
2. Data extraction: Advanced AI solutions can automate data extraction, transforming unstructured data into usable formats. Agentic AI systems can identify key financial data, verify document integrity, and flag discrepancies in income statements, loan applications, and other documents. These agents can execute data extraction across multiple formats and data sources simultaneously, enhancing speed and accuracy.
3. Analytics and calculation of risks: Agentic AI can analyse borrower financial data, credit history, and economic indicators to perform comprehensive risk assessments. Predictive models calculate risk scores, identify potential defaults, and recommend lending terms. AI agents can also cross-reference real-time market data to adjust risk calculations dynamically, ensuring more accurate lending decisions.
4. Review and decision: Once AI-generated insights are produced, Agentic AI systems facilitate the review and decision-making process. AI agents can compile decision-making inputs into comprehensive memos, allowing human experts to validate findings. By presenting a detailed audit trail of all data inputs, risk assessments, and decision criteria, AI agents enhance transparency and accountability in the review process.
5. Model for continuous improvement: AI systems should not be static. Agentic AI can continuously monitor and refine decision-making models. Feedback loops allow agents to update risk assessments, detect emerging fraud patterns, and optimise algorithms based on new data, ensuring the system evolves with market conditions.

The integration of Agentic AI in lending processes can not only reduce review cycle times by up to 60 per cent, as per a recent McKinsey report, but can also enhance accuracy and operational scalability by leveraging data from multiple sources. This approach empowers financial institutions to deliver faster processing times, improved customer experiences, and more consistent decision-making, ultimately gaining a competitive edge in the market. Early adopters of Agentic AI are 1.8 times more likely to improve gross margins and productivity, and these benefits will be especially felt in complex processes like underwriting, so lenders should act quickly to leverage its use.  Explore how the NOW platform can digitise and connect workflows and bring actionable AI to every corner of your lending operations.