The easiest way to make an AI system look ready is to show it the question it was prepared to answer. The useful test begins when the evidence is incomplete or the requested action is outside its scope.
Our July review discussions insisted on a truth focused assessment of Omnia. That is the standard Vertex Automotive Solutions should apply to its own development and to the technology it evaluates for customers.
A credible demonstration uses representative records and normal exceptions. It shows what happens when a source is missing, an identity cannot be resolved or the user asks for a decision they are not authorised to make. Failure handling belongs in the demonstration because it belongs in daily work.
Technical test results remain important, but their meaning must be clear. A successful component test does not establish reliable production behaviour across every connected service. An attractive maturity score is weak evidence if another reviewer cannot reproduce its basis.
Omnia's progress should therefore be described in terms of the journeys tested, limits retained and acceptance still required. That makes the next development decision more useful and the commercial promise more credible.
Ask the intended user to complete an ordinary task without a developer guiding the route. The awkward case is part of the product, and it belongs in the acceptance evidence.