Grounding
Anchoring a model’s output in specific, approved sources of information, such as retrieved documents or database records, rather than in its general training. A grounded answer can be traced back to evidence the organisation trusts.
Why it matters
Grounding is the main defence against unsupported outputs in business settings. It lets users check claims, lets reviewers see why an answer was given, and keeps responses consistent with current policy and data.
Grounding is only as good as the sources and retrieval behind it. Outdated documents, missing permissions or weak retrieval produce confidently grounded answers that are still wrong.
In practice
For example, a UK pension administrator’s member-query assistant might answer only from the scheme rules and the member’s record, cite the rule it relied on, and hand over to an administrator when no approved source addresses the question.
Where Rodan fits
Rodan grounds generative systems in governed sources through AI and Decision Systems work. See also AI hallucination: what it is and how businesses should manage it.