Process first, technology second
The most advanced model applied to the wrong problem is still just a cost. We start from where the company is losing margin, not from what shipped this week.
Rayo Consulting is an AI consultancy founded in April 2025 in Città di Castello, Italy. We work with Italian SMEs that want to understand where artificial intelligence actually pays off, and where it doesn't.
It all started with a question. We kept hearing it from business owners and operations managers, always in the same words.
“Is AI right for us?”
There was almost always an answer. Nobody was giving it in a usable form.
Impressive demos that went nowhere.
Enterprise-scale projects, pitched to companies with fifteen people on the payroll.
We chose the middle ground. Go into the company, look at the real processes, and put into production only what holds up on a Monday morning.
We are a young studio, founded in 2025.
We don't have a long history to tell. We have systems running in production, clients who use them every day and a method shaped on site, not in a slide deck.
We don't do everything. We do four things and take them all the way to production.
Four phases, always the same ones. They exist to avoid the mistake we see most often: projects that start from the technology instead of the problem.
We talk to you and to the people who do the work every day. Where time gets lost, what gets done twice, what nobody wants to keep doing by hand.
We check feasibility: which processes can genuinely be automated, with which data, at what cost. If a project doesn't stand up, we say so upfront, not later.
A proposal with priorities, timings and agreed KPIs. Ninety days, not three years. You know what is coming, when, and how we will measure it.
We build. Frequent releases, checks against the numbers, corrections along the way. At the end, the code and the data are yours.
The most advanced model applied to the wrong problem is still just a cost. We start from where the company is losing margin, not from what shipped this week.
We show what runs, on what data and with what limits. Including when the honest answer is that AI is not the right tool for that particular step.
Every project has numbers agreed before we start: hours recovered, requests handled autonomously, response times. If it is not measured, it is not a result.
Open source stack and zero vendor lock-in. What we build, you can take with you, have maintained by whoever you like, or carry on yourself.
Based in Umbria, working with clients across Italy: web agencies, manufacturing, financial services and retail: different sectors, the same method. What we built, and the numbers it produced, is documented in the case studies.