Business analysis glossary
AI Requirements
What it is
AI requirements describe what an AI powered feature must do and how it must behave, including its purpose, the data it may use, accuracy targets, how it explains outputs, how errors and uncertainty are handled, fairness checks, privacy, human review and how its performance is monitored after release.
Why it matters
AI features behave probabilistically, so traditional requirements like the system shall calculate are not enough. Without clear AI requirements, teams build models that are hard to test and risky to use. Analysts translate business goals into measurable expectations and guardrails, working with data scientists and compliance.
Example
For a crop loan scoring model, the analyst writes that it must use only consented data, achieve an agreed accuracy on a held out test set, explain the top three factors behind each score, be checked monthly for unfair differences between districts and send borderline scores to a loan officer for review.