What is grounding?

Grounding is the practice of connecting an AI system's outputs to verified source data rather than allowing it to generate responses based solely on patterns from its training data or information provided in a user prompt. In accounting, grounding helps ensure that answers, analyses, and recommendations are based on actual financial records.

Grounding is an important safeguard because it reduces the risk of hallucinations and helps users trust the information an AI system provides.

Example: A business owner asks Ask Digits, "How much did we spend on software last quarter?" Rather than generating an answer from general accounting knowledge or assumptions in the prompt, Ask Digits retrieves the information from the underlying ledger and returns a response based on the company's actual financial records.

Related terms: Hallucination, System of Record, Ask Digits, Embedded Verification, Agentic General Ledger™ (AGL®)