Grounded AI
Grounded AI means FundRobin can give an AI model relevant authorised organisation knowledge, grant context and source evidence to work from instead of relying only on the model’s general knowledge. Grounding improves context and inspectability, but it does not make AI error impossible.
Where relevant, Robin can retrieve reviewed organisational context from the Organisation Brain rather than asking users to repaste the same mission, beneficiaries, outcomes or track record into every conversation.
Robin can also work with opportunity facts, eligibility and match context, application requirements and other workflow context available to the organisation.
Relevant source passages and URLs can be retrieved for grant research and proposal work. Where supported, Sources lets users inspect material evidence behind the work.
The Knowledge Graph helps teams explore relationships in organisational knowledge. Selected graph context can be taken into Robin where supported, while reviewed Organisation Brain knowledge remains the factual authority.
Grounding cannot create facts that do not exist. Missing or conflicting organisation information can be surfaced for review or as Input Required rather than being silently resolved by the model.
Grounding reduces uncertainty but does not guarantee that every statement has a one-to-one citation or that an answer cannot be wrong. Important sources, facts and final application content still need human review.