What is a generated marketing department?

Most "AI marketing" tools are one model with a prompt taped to it. You type what you want, it types something back, and everything that makes a real department work — who owns what, who checks what, what the company may and may not claim — lives nowhere. A generated marketing department is the opposite bet: the structure is the product, and the model is just the electricity running through it.

It starts with an interview, not a form

Growics begins by reading your website — what can be seen is never asked. Then a consultant runs a founding interview, by voice or in writing, in Arabic or English. A typical interview is eight staffing questions and three truth questions, and it stops when every staffing condition is resolved — never on a question counter.

The staffing questions decide which seats your company actually needs. If you say "people sign up and go quiet," a Lifecycle seat earns its place, and the proposal quotes you as the reason. The truth questions write down three things most tools never ask: what your company may claim, what it must never say, and who decides what is true.

What gets generated

The output is not a report — it is a working department:

• The team. Four to ten specialists from a shelf of 28, each with an ownership boundary, a tool license, and a hard-stop list. Titles are functions, never ranks.
• Two gates, always. A facts gate checks every claim against your truth files; a quality gate checks everything else. They install in every team, no exceptions, and no seat can skip them — including the ones that write the checks.
• A measured memory. The department's memory is benchmarked from day zero on your own questions, and facts learned mid-task wait in quarantine until your reviewer approves them.
• One honest ledger. Everything shipped — by agents and by your people — lands attributed in one record. What is not recorded did not happen.

Why "generated" and not "configured"

Configuration assumes you already know what to set up. Generation moves that burden to the system: the interview resolves the decisions, and the engine composes the department from parts that already know how to work together. The difference shows up months later — a configured tool drifts into a junk drawer of settings; a generated department can be regenerated, upgraded as a fleet, and reasoned about.

The honest limits matter too. A generated department does not invent capabilities: if your stack is missing a tool a seat needs, the seat is created with that gap named, never worked around. And it does not replace judgment — your reviewer owns the truth files, and the gates enforce what you decided, not what a model guesses.

You can walk through a live department — the office, the interview, the gates — in the demo workspace.