Mapping your AI systems
What your company actually uses, who uses it and for what. The first step is always to discover how much AI is already inside without anyone having decided so.
← Back to home
Chapter 02
AI
I'm not an AI cheerleader: I use it every day to work, I build it when it's needed, and since the European regulation arrived I help companies stay within it. A tool, not magic.
President of ANPAI — since July 2026
I chair the National Association of Artificial Intelligence Professionals: the common voice of those who design, build, apply and govern AI in Italy. Professional representation, skills and training, project assessment, technical working groups. It's the place from which I try to shape the rules of the craft, rather than merely follow them.anpai.it →
AI ACT
NIS2
NIS2
AI Act & NIS2 certification — 2026
Certified competence on the EU Artificial Intelligence Regulation and the NIS2 cybersecurity directive. On this basis I opened a consulting desk for companies, with a conversational assistant trained on the legislation, guidelines and practice.
Compliance
Classification under the AI Act
Unacceptable, high, limited or minimal risk: each category carries different obligations. We define where your systems fall and what that entails.
Transparency and documentation
User disclosure, marking of generated content, system registry, technical documentation, human oversight.
Security and continuity
Risk analysis, incident management, business continuity, supply-chain security, notification obligations.
AI literacy — the forgotten obligation
From 2026, anyone deploying AI systems must ensure an adequate level of competence among those who use them. I design and deliver the training.
Gap assessment and plan
A document that says where you are, where you need to get to, in how much time and at what cost. No alarmism, no selling fear.
Building
Agentic loops
Agents that don't just answer: they plan, execute, check the result and start again until the task is done. Autonomous loops that oversee a process from start to finish, with human checkpoints where they truly matter.
Assistants on your documents
Chatbots trained on your knowledge base: semantic search, answers with cited sources, guardrails on out-of-scope topics. Designed, built and put online by me.
Processes that run themselves
Flows that collect data, write, publish and alert without human intervention. Make, GoHighLevel, Supabase and model APIs, integrated with what you already use.
Python, machine learning, pipelines
Extracting, cleaning and structuring data; bespoke models when a general-purpose one isn't enough.
Assisted production
Synthetic voice, content generation, automated editing, prompt engineering: twenty years of audiovisual craft joined with generative tools.
Paths for the team
From the first prompt to automating a real process. Tailored by role and sector, with exercises on your own documents.