01
Choosing the use cases
I start from processes and people, not tools. I map the activities, estimate value and risk, and choose one or two cases to start from, with success criteria written down before we begin.
Artificial intelligence for companiesFrom prototype to adoption, with a method.
Many companies have tried generative AI. Few have brought it into everyday work. As an AI Product Manager, that’s the step I work on: choosing the right problem, building a prototype people can try, measuring whether it works and making sure people use it.
I’ve worked for and with
- Leroy Merlin
- Bricocenter
- Intesa Sanpaolo
- Bocconi University
- BMW
01
Where I bring AI into a company
02
Prototyping in days
With Claude Design and Claude Code I build working prototypes for real users to try. Decisions are made on something you can touch, not on a slide.
03
Making AI trustworthy
Answers with sources, confidence thresholds, human review, automated tests on the outputs. The requirements of the EU AI Act, ISO 42001 and NIST AI RMF become concrete controls.
04
Driving adoption
Training on real work, champion colleagues, adoption KPIs collected without monitoring people. This is where many AI projects stall.
02
The method: six PMI-CPMAI phases
An AI project is a data project. That’s why I follow CPMAI, the Project Management Institute’s method: iterative, with a checkpoint at the end of each phase.
- IBusiness understandingWhich problem, for whom, and how we’ll measure success. If AI isn’t the right answer, we say so here.
- IIData understandingWhich data exist, where they are, who owns them and how reliable they are.
- IIIData preparationCleaning, permissions and privacy before any model.
- IVModel developmentA prototype built with AI agents, tried by users within days.
- VEvaluationAutomated tests on the outputs, edge cases, checking the criteria agreed at the start.
- VIOperationalisationRelease, training, adoption KPIs and monitoring over time.
- InsuranceCompliance CopilotAn assistant that answers only from sources, with mandatory citations, human review and readiness panels aligned with the EU AI Act, ISO 42001/42005 and NIST AI RMF. Working POC on synthetic data.
- Professional firmGrowing LabVisibility in AI engines: across 15 questions to ChatGPT, Claude and Gemini, the brand appeared in 22 of 45 answers. A monitoring dashboard and a conversational coach for the firm’s clients.
- Productstaged.aiVirtual home staging: a SaaS built single-handedly with Claude Code, at about €0.21 per analysis, validated with a community of beta testers.
FAQ
Frequently asked questions
What does an AI Product Manager do in an artificial intelligence project?
They choose where AI creates value, get solutions built and validated, manage risk and compliance, and follow adoption. They don’t start from a tool: they start from the problem and the data.
Where do you start when introducing artificial intelligence in a company?
With an inventory: which AI tools people already use, often with personal accounts, and which repetitive tasks weigh the most. Then one or two small use cases, with a measurable success criterion and a prototype tested within a few weeks.
Why do so many AI projects stop after the demo?
Because technology is the easy part. What’s often missing is success criteria agreed in advance, reliable data, users’ trust and someone who follows adoption after the release. That’s why I measure real usage with dedicated KPIs.
Does the AI Act require companies to train their employees?
Yes. Since 2 February 2025, Article 4 of the AI Act requires providers and deployers of AI systems to ensure a sufficient level of AI literacy among the staff who use them. It doesn’t prescribe a standard course: training must be adapted to the roles and the systems used. See how AI is learned at work.
What is the PMI-CPMAI certification?
It’s the Project Management Institute’s certification for managing artificial intelligence projects with an iterative six-phase method, from business understanding to operationalisation, with a focus on data and trustworthy AI.
C
Contact
Shall we talk?
Want to talk about the method or invite me to a talk, a meetup or an interview? Write to me.
vsemeraro@gmail.com