Proportionate effort
A minor, reversible matter does not require the same review as a material decision with lasting consequences.
Purpose and limits
Information is plentiful. Knowing what it supports is harder.
Reports, emails, conversations and professional judgement often sit apart. A decision may be made before anyone has distinguished what is known, assumed or still missing.
The problem
Facts can become mixed with assumptions. Sources can conflict. A risk or dependency may only become obvious after an action is taken. Later, it can be difficult to reconstruct why a decision was reasonable at the time.
Rabbit is being developed to test whether a structured, AI-assisted process can make the basis of a decision easier to examine without removing human responsibility.
Principles
Visible uncertainty
Where evidence is insufficient, “not ready to decide” can be a useful and honest finding.
Traceable reasoning
The aim is to connect information, analysis, recommendation, human decision and eventual outcome.
Authority stays human
Rabbit does not grant professional approval, accept risk, authorise expenditure or issue instructions on anyone’s behalf.
What Rabbit does not claim
Rabbit is not a substitute for a professional, regulator, client or accountable manager. It does not guarantee a correct outcome or turn incomplete evidence into certainty. Its broader architectural direction is not evidence that every capability is available in today’s Web Pilot.
Validation is ongoing. We want reviewers to challenge where the method or implementation falls short.