I find where AI creates value in your business.Then I put it into production.
I help companies take on the AI revolution: I plan adoption around how the business works, and deliver it with off-the-shelf tools or custom AI agents.
- Over 50 processes transformed
- Over 30 AI agents put into production
- Over 20 adoption and training plans delivered
- Background
- Strategy and IT consulting at KPMG and EY-Parthenon. Joined Yellow Tech to help Italian companies through their AI transition.
- Sectors
- Manufacturing, banking and finance, public sector, software
- Companies I’ve worked with
- From €5 million to over €1 billion in revenue
First I learn how you work.Then I decide where AI belongs.
-
Listening
An AI-adoption survey for the people involved: what they already use, where they lose time, and how ready they are.
-
Process map
Workshops and interviews with the people who do the work, activity by activity, as it runs today.
-
Redesign with AI in mind
I redesign your processes with AI in mind and optimise where it pays. Every activity has three possible paths: improve it without AI, with off-the-shelf tools, or with a custom agent. An agent is not the answer to everything. Fig. 1
-
Roadmap
Every possible initiative on a timeline and a budget: the investment each one needs and, where the data allows, the expected return. So the company chooses what to do and when.
-
Governance
Where needed, the organisational model and strategic direction for governing AI over time.
-
Delivery
The chosen initiatives go into production: off-the-shelf tools configured for the people who use them, custom automations and agents, with acceptance criteria tested on real cases.
Fig. 1 Process chart
- Process
- Handling a technical support request
- Activities
- 8
- Type
- Illustrative example, not a client case
-
Receive the request Today: It comes in by phone, email and portal; someone retypes it into the ticket by hand. Outcome: Without AI.
Today It comes in by phone, email and portal; someone retypes it into the ticket by hand.
Why Simplified: one entry channel, with only the fields that matter. Before adding AI, remove the needless variety.
-
Check warranty and contract Today: Someone looks up the serial number in the ERP and checks the contract and its dates. Outcome: Without AI.
Today Someone looks up the serial number in the ERP and checks the contract and its dates.
Why Classic automation: it’s a fixed rule (serial number, contract, expiry). It costs less than a model and is more predictable.
-
Set priority and assign a technician Today: The coordinator reads every request and decides who gets it. Outcome: Custom agent.
Today The coordinator reads every request and decides who gets it.
Why This needs an agent connected to your service system: it reads the request and proposes category, priority and technician. The coordinator confirms or corrects, because a mistake here means a wasted site visit.
-
Answer status enquiries Today: The customer calls to ask where things stand; someone checks and calls back. Outcome: Custom agent.
Today The customer calls to ask where things stand; someone checks and calls back.
Why Repetitive questions, answers already in the system, no irreversible decisions: a custom agent answers on its own and hands over to a person when a question falls outside its scope.
-
Diagnose the fault Today: The technician searches the manuals and asks more experienced colleagues. Outcome: Off-the-shelf tools.
Today The technician searches the manuals and asks more experienced colleagues.
Why The diagnosis stays with the technician. With an off-the-shelf tool configured on your manuals, the right page and similar solved cases come up in seconds.
-
Approve spare parts Today: Even low-value parts need two signatures. Outcome: Without AI.
Today Even low-value parts need two signatures.
Why Simplified: a threshold below which the technician decides alone. Fewer steps, no new software.
-
Retype the service report Today: The paper report is retyped into the ERP back at the office. Outcome: Without AI.
Today The paper report is retyped into the ERP back at the office.
Why Eliminated: the technician fills in the report once, on a tablet. The retyping disappears, and so do the typos.
-
Update the knowledge base Today: Solutions live in emails and in senior technicians’ heads. Outcome: Off-the-shelf tools.
Today Solutions live in emails and in senior technicians’ heads.
Why The senior technician writes the article with an off-the-shelf tool configured with your template, starting from the closed case instead of a blank page.
Eight activities. Four improve without AI. Two with tools that already exist. Only two need a custom agent.
How we work together
The services follow this order, but each one can be taken on its own.
-
First step
A 30-minute call
- Duration
- 30 minutes
Thirty minutes to see whether, and where, AI can help you.
Email me -
Strategy
AI Strategy & Roadmap
- Duration
- 4–6 weeks
I study your processes through a survey, interviews and workshops with the people who do the work, and redesign them with AI in mind, optimising where it pays.
What you get
- Your processes mapped and redesigned to take on AI, activity by activity
- A roadmap of the initiatives over time, with the investment each one needs
- Where the data allows, the ROI of each automation, based on the savings it brings
- Where needed, governance: organisational model and strategic direction
You come out knowing where to invest, and where not to.
-
Delivery
Off-the-shelf tools
Copilot, Claude, ChatGPT and other off-the-shelf tools, configured for your people and for the activities defined in the roadmap, with the training to use them.
Custom agents and automations
AI automations and agents built on your processes, where off-the-shelf tools are not enough.
For larger projects I work with specialist firms, which I introduce when needed. Smaller jobs, like a single agent, I handle myself.
About me
I’m a management engineer. I started in strategy and IT consulting at KPMG Nolan Norton and EY-Parthenon: transformation plans for banks, due diligence, post-merger integrations, cloud strategy for public bodies. At KPMG I founded the internal AI-Lab. Today, at Yellow Tech, I design AI agents and take them into production for Italian companies.
My strength is connecting business and technology: I talk to the operations director and to the AI engineer on the same day, and translate one into the other.
I have also built products of my own.
- Vertigo AI
- AI agents for HR teams that cut time-to-hire by 40%.
- LiberAI
- An AI sommelier for restaurants.
What I work with
- Off-the-shelf tools
- Microsoft Copilot, ChatGPT, Claude, Gemini
- Platforms
- Salesforce Agentforce, Azure OpenAI
- Architectures
- Multi-agent RAG, agents with human oversight
Who I’m not for
- I don’t sell licences.
- I don’t build demos that stay demos.
- If a problem is better solved without AI, I’ll tell you.
Frequently asked questions
Does our data leave the company?
It depends on the path. Custom agents are deployed on your own cloud or on-premise infrastructure; off-the-shelf tools are used in their business editions, configured to your data rules. Either way, who can access what is agreed up front, based on how sensitive the data is.
How do you handle the AI Act and GDPR?
As project requirements, not as an afterthought: a documented risk classification, transparency towards operators, human oversight of decisions, and a data processing agreement.
How soon will we see something?
AI Strategy & Roadmap takes 4–6 weeks. Delivery depends on the path: configuring off-the-shelf tools is faster than building a custom agent.
What do you need from us?
A technical contact, a process expert to validate the results, and access to the data and systems involved.
Do you only work with large companies?
No. I’ve worked with companies from €5 million to over €1 billion in revenue, across different sectors.
Let’s talk about a process that costs you too much today.
Thirty minutes to see whether, and where, AI can help you. If it’s not the right fit, I’ll tell you.
- vCard
- Save my contact