Eurocomponents

Maurizio "Canta" Cantaboni is a senior consultant, not an employee. He works with Gianluca Zanoni (GM and owner) as external advisor. Known each other ~10 years. He is a close personal friend of mine.

He is the decision shaper. Zanoni signs. Trust level is extreme:

He has been pushing AI and software automation to Zanoni for months. Already took 2 or 3 vendor calls with no one. "Non ho avuto un gran feeling con nessuno." We are near the front of the line by default.

He also manages a second external client, separate, flagged for a later play.

• Eurocomponents SpA, Bedizzole (BS). Prefabricated bathroom pods since 1978. 60,000+ pods delivered across 20+ countries.
• €18M revenue, 39 employees. Grew from €10M to €18M in recent years, rebuilt by Zanoni after losing the majority years ago.
• Just lost a €6.5M project. Buyer pulled out after agreement signed. Production has been empty for 4 to 5 months.
• Currently on cassa integrazione, 2 days per week reduced hours.
• Zanoni quote: "le persone che li han lì non possiamo fare affidamento su quelle persone." He knows he cannot rely on current staff.

• 150 quotes per year (not per month). Low volume, very high value per quote. Each quote is a full project (hotel, hospital, residence, dorm) = 100 to 300 bathroom pods per project = €M-scale per offer.
• Turnaround: 2 weeks standard (clients ask for 3 days, reality is 2 weeks). Time to prepare the "descrittivo" alone: 10 min simple to 4+ hours complex.
• Inputs are chaos. Descriptive specs, CAD, email text, Excel, PDFs, JPEGs. Languages: Italian, German, French, English, Swedish, Norwegian. Often the bathroom info has to be dug out of general building capitolati (plumbing, electrical, structural plans), not bathroom-specific files.
• They do not work from a catalog. Every bathroom is custom designed per project. This is the key architectural difference from Aronlight.

Phase 1 only. No end-to-end. No catalog matching.

Scope: messy multi-format / multi-language input bundles go in, a clean standardized "descrittivo" per bathroom typology comes out. The preventivista takes the output and adds pricing manually (for now).

Architecture: LLM + rules + human-in-loop. Same pattern shape as Aronlight v6 (inputs to structured output with human review), different engine below (extraction and standardization, not SKU matching). We reference Aronlight as proof we have shipped this pattern, not as the template.

Why this: Canta explicitly de-prioritized email parsing and end-to-end auto-quoting. He wants the first block of the problem solved well. If we try to pitch the full stack Monday, we lose him.

• Canta's comfort range: €3K to €15K implementation + usage-based ("pago per uso" / per proposal generated).
• Zanoni's anchor: a back-office person costs ~€30K/yr. He is willing to spend €15K impl + €10K/yr maintenance if it replaces human effort. Not willing to pay €30K/yr just to keep the human.
• Competitor: €200K to get started was rejected outright.
• Our Aronlight v6 shape maps directly: low implementation + per-transaction. Different unit (descrittivo generated vs RFQ processed), same model.

• Do NOT pitch end-to-end.
• Do NOT reuse the Aronlight A2 slide deck as-is. The catalog-matching framing does not apply here.
• Lead with business model: low impl (€3–5K) + usage fee per descrittivo generated. Same shape as Aronlight v6, different unit.
• Reference Aronlight as pattern proof (email-to-structured-output + human-in-loop), not as direct template.
• Acknowledge "no catalog" explicitly as Canta's insight. Shows we heard him.