modules/comunicazione-professionale/skills/comunicazione-professionale/references/product-thesis.md

# Product thesis

The workflow replaces the preparatory communication work often assigned to a
junior professional: monitor developments, collect sources, identify what
matters, prepare a first technical explanation, adapt it to channels, and lay
out a clear visual story. It does not replace the commercialista's technical
position, audience knowledge, relationship judgment, or approval.

The product is episodic, not a scheduled content generator. A material norm,
deadline, official interpretation, case-relevant development, or explicit
professional initiative creates the occasion. Most days correctly produce no
run and no communication. Its valuable output is one
of two professionally reviewable conclusions:

1. a source-backed communication package with a useful, audience-specific
   reason to exist; or
2. `no_publish`, with a concise explanation of why the available development
   is immaterial, insufficiently supported, too uncertain, repetitive, or not
   actionable for the selected audience.

One reviewed technical basis may support several channel adaptations. Each
channel changes expression and length, never the underlying claim set. A
carousel is not an automatic adaptation of a post. It exists only when it adds
useful detail, structure, comparison, sequence, or a bounded decision aid. If
it merely divides or paraphrases the post, omit it.

A studio's format is part of professional identity. Preserve its reviewed
letterhead, numbering, recipient and subject conventions, section hierarchy,
email signature, website byline and source treatment, social conventions, and
visual system. Do not substitute a generic Vera template when an approved
studio profile exists.

Prior posts and emails are evidence of style, not proof of technical accuracy
or audience success. Engagement metrics, when explicitly supplied, are context
for the professional; they do not automatically select a topic or determine
truth.

The professional selects every exact prior communication used for style
learning. Vera never scans all Studio documents or account history. Local code
first strips explicit-format email addresses, phone numbers, tax identifiers,
account identifiers and case numbers. One dedicated model pass receives only
those stripped documents and produces complete pseudonymized derivatives. A
second fresh model pass sees only the candidates and checks for residual
contextual identification. The identity mappings stay local. Generation sees
the accepted derivatives; claim, editorial, and visual review do not receive
prior communications at all. Transient stripped inputs are deleted after the
independent review, while the Studio retains the originals and local mapping.
This reduces the first model's exposure but does not
make that pass anonymous: names and contextual identifiers can remain until the
semantic pseudonymization is performed.