I lead the global Salesforce and AI platform of a multinational industrial group in the energy sector. A team of nine, and the roadmap of a platform the business runs on.
Where I sit
I am the buyer: vendors present to me, I sign, and I explain in committee where the money went.
- Scope
- ~€5bn
- of industrial sales pipeline behind the platform I run
- Responsibility
- €1M
- annual project budget, mine to defend
- Efficiency
- 69,147
- hours returned to the organisation by CRM modernisation
- Adoption
- -30%
- internal support tickets on roughly 2,000 users, within two months of launch
What I do with LLMs
I work on a CRM without opening it. From Claude, over MCP, I read the org metadata and the official documentation, query the data, and apply changes with the Salesforce CLI. The customer deck is a Lovable web app, so I can see which pages the reader stops on.
How I know it works
Two agents, two checks, neither of them a score after the fact.
The internal knowledge agent answers with its source attached: whoever reads an answer can open the document and check it.
A research run: every dental practice in Emilia-Romagna, in about an hour, for $0.34 against a $1 ceiling I set. The log is the proof:
- 3,122 records from 660 searches across 330 municipalities, zero errors;
- zero searches hit the pagination limit, so nothing was silently truncated;
- 688 non-dental listings discarded, and counted;
- every record and every email carries its source.
Limits stated up front: these are public listings, not a register of licensed practitioners. Emails are published, not verified, and none were invented.
An evaluation, to me, is a design where every row can be checked by whoever has to act on it.
Three methods I use
- Value Intelligence: AI initiatives funded on measured value, hours given back and fewer requests for help, not on features shipped. I designed it in my current role, and it is now how Salesforce and Agentforce initiatives are decided.
- Safe-to-Learn: technical feasibility proven in a separate demo org before anything touches company systems or budget.
- Portable Operational Memory: governed data, rules and permissions that outlive whichever model the company plugs in.
All three in full on how I work.
Background
- Now Global Salesforce Platform Lead & AI Product Owner Multinational industrial group, energy division
- Leading a multidisciplinary team of 9.
- Governance of CRM and AI investment, with a €1M annual project budget.
- 2024 - 2025 Lead AI Transformation & Business Architect Multinational industrial group, energy division
- Enterprise AI roadmap, and the evaluation and adoption of Agentforce, Einstein and Data 360.
- AI agent on Microsoft Teams and SharePoint: internal support tickets down by more than 30%.
- 2022 - 2024 Business Architecture Specialist Accenture
- Tier-1 clients in luxury, automotive and retail.
- Core process efficiency improved by 35% across multi-country deployments.
- 2020 - 2021 Salesforce Project Manager Seven IT
- End-to-end Salesforce implementations across different sectors.
Certifications
Twelve in total. The ones that carry the most weight:
- Data Cloud Consultant
- AI Associate
- Strategy Designer
- UX Designer
- Advanced Administrator
- Business Analyst
- Service Cloud Consultant
- Sales Cloud Consultant
- Platform App Builder
Education
- 2015 - 2017 Master's Degree in General Management University of Modena and Reggio Emilia · 105/110
- Thesis: Digital product development for the gig economy.
- 2014 Bachelor's Degree in International Social & Political Sciences Alma Mater Studiorum, University of Bologna · 97/110
- Thesis: Toyota's Lean Production.
Fourteen AI agents and tools built in the evenings, at my own expense.
What I am looking for
Applied AI architect work at a frontier lab or an AI-first company, on the customer-facing side of enterprise adoption: where AI creates value, how it connects to the systems already there, what makes it outlive the pilot. Stated up front: my hands-on work is prototype-led, not production software engineering.
Direct contact
Email is the fastest route.