Business diagnostic

A serious organisational diagnostic costs weeks of consulting and arrives as a document nobody reads twice. The information needed to produce it, on the other hand, the company already has: it only needs structuring and criteria applied to it.

• Neurona
• Airtable
• n8n
• OpenAI or Claude
• PDF generation

A flow that only works with perfect data is no use. These are the cases that were anticipated and what the system does in each one.

• The questionnaire is left half-answered

  A dimension without enough answers returns “no data” and is shown that way in the report. It is never filled in with an average, and never estimated.

• The model drafts something that does not match the score

  The model only drafts; the diagnosis is already settled by the rules before it is called. If the text contradicts the score, the diagnosis wins and human review cuts it.

• The model provider goes down or blows the budget

  The report is assembled from the pre-written templates and goes out anyway. Model-written prose is an improvement, not a dependency.

The questionnaire below works in full right here: you answer it, it computes and you read the plan without leaving the page. It is a smaller, self-contained version of what the flow describes: it calls no model to draft anything — the actions are written in advance — and it does not generate the PDF. What it does do is the part of the case that matters: score by dimension, order by what weighs most and say “no data” when answers are missing.