Ask Oscar
The report answers back, and only about itself
When the analysis is done, the reasoning behind it is still there. You can ask it to defend a verdict, argue with a specific score, or say what would have to be true for the answer to change — scoped to your analysis, not to the internet.
Scoped, and why that is the feature
A general assistant asked about your idea will produce a plausible answer from nothing in particular. That is the opposite of useful here: the whole value of a scored analysis is that its claims are traceable, and a chat that invents new ones undoes it.
This one is bounded to the report and its sources. Ask why competitive advantage scored nine out of fifteen and it answers from what the analysis found, or says it does not know. The second answer is the one that makes the first trustworthy.
What it is actually good at
Three things, in order of how often they earn their keep. Interrogating a single score, because a number without its reasoning is hard to act on. Testing a change before you commit to it — narrowing the customer, changing who pays — and hearing which dimensions move. And asking the inverse question: what would have to be true for this verdict to flip.
That last one is the most useful and the least asked. A verdict tells you where you are; the conditions that would change it tell you what to go and find out.
What it will not do
It will not tell you the idea is good because you would like it to. The analysis is scored against evidence, and the chat reads from the same evidence — arguing with it does not move the score, it only surfaces the reasoning.
It also will not run new research. When a question needs evidence the analysis does not have, the honest answer is a new analysis or a Radar area, and it will say so rather than filling the gap.
Included with every analysis.