AI in Credit Best Practice

AI is moving from general experimentation into more practical workplace use, and credit is no exception.

In a recent GICP webinar, Adam Ahmed (GICP, Fitch Learning) explored prompt patterns, context management, agentic AI and the realities of applying these tools in professional settings. The discussion focused on how AI can support work in credit today, what risks need to be managed, and what skills may become more important as adoption develops.

What AI can and cannot do in credit

It is clear that AI should not be treated as a replacement for deterministic, logic-based credit models. Where a process depends on repeatable rules, explainable calculations and consistent outputs, those foundations still matter. Large language models are not designed to provide that kind of reliability on their own.

Where AI can add value is in the surrounding workflow. It can help gather information from large volumes of unstructured material, identify relevant content more quickly, support data extraction and preparation, and assist with the early stages of analysis. In that sense, AI is most effective when it strengthens the path into a credit process rather than replacing the underlying analytical logic itself.

Why prompt design is only part of the picture

Prompt engineering can be presented as the key to effective AI use, but the webinar makes a broader point: context matters just as much. A strong prompt helps frame the task, but output quality also depends on what information is included, how relevant that information is, and whether the model is being overloaded with unnecessary material.

For credit professionals, this has practical implications. If an AI tool is being used to support a focused task, it is often better to provide a smaller, more relevant set of information than a large volume of material with no clear structure. It is important to monitor context limits over longer interactions, particularly where work is being carried forward across multiple steps or tasks.

What agentic AI could change

There is a distinction between conversational AI and agentic AI. In simple terms, an agent does not just respond to a prompt; it can work through tasks, operate within constraints, and take steps towards a goal. That makes it potentially more useful in real workflows, but it also introduces new questions around oversight, security and control.

This matters because the value of agentic AI depends not just on the tool itself, but on the environment around it. Processes need to be documented, priorities need to be clear, and data needs to be organized well enough for the system to work effectively. Without that structure, adding AI agents may simply expose the weaknesses that already exist in a workflow.

Risk, accountability and responsible use

A recurring theme in the webinar is that AI output remains the user’s responsibility. If material generated with AI is used in analysis, reporting or decision support, it still needs to be reviewed, tested and owned by the professional using it. This is especially important where hallucinations, incomplete reasoning or over-confident phrasing could create risk.

Furthermore, as AI systems become more connected to tools and workflows, the consequences of error can become more significant. Security, permissions, data handling and governance therefore remain central to any serious application of AI in credit.

Top tips for credit professionals using AI

• Use AI to support information gathering, preparation and workflow efficiency rather than to replace rule-based credit logic.
• Keep prompts clear and purposeful, but pay equal attention to the quality and relevance of the context you provide.
• Treat AI as a tool for insight, challenge and learning rather than a source of definitive answers.
• Review and own any output before it is used in analysis, communication or decision-making.
• Develop stronger process, documentation and prioritization habits, as these become more important when working with agentic tools.

Watch the webinar.

Further learning

For those looking to build a broader understanding of how AI is being applied in credit, the Global Credit Certificate (GCC) includes a dedicated chapter on AI in credit, covering practical uses, limitations, risks and guardrails for responsible adoption. Find out more about the GCC here.