Why this training should be given by someone who writes the software and stands in a classroom – and how a course I have already delivered was built.
Training on artificial intelligence in schools is usually delivered either by computer scientists who have never had a class of their own, or by enthusiastic teachers who have only used the tools. I stand in between, and not by accident.
I write software and I build websites. I design and develop web applications end to end (interface, back end, integration with archives and APIs, deployment to the cloud) and I look after WordPress sites for associations, magazines and cultural projects, from content architecture to the domain, from maintenance to security. I use language models inside my own development work every day, which is the least theoretical way there is to know what they can do and where they go wrong.
And I teach Italian and history: I mark homework like the people listening to me, I sit on the same staff meetings. When someone in the room asks whether it makes sense to have a machine write an essay, I do not answer as a consultant, I answer as a colleague who has to grade that essay on Monday.
Above the two there is a third thing, a humanities training: a doctorate in Romance philology, a master’s in public history. A language model produces text, and assessing it, tracing its sources, working out where it is lying is quite literally the trade of a philologist. It is also, word for word, what John Dewey asked of schools more than a century ago.
Active, persistent, and careful consideration of any belief or supposed form of knowledge in the light of the grounds that support it and the further conclusions to which it tends constitutes reflective thought.
John Dewey, How We Think, 1910
Dewey was writing about books and newspapers, but the definition fits a machine-generated text better than anything else. The question for schools is not whether students will use artificial intelligence: it is whether anyone will teach them to ask what grounds hold up what they are reading. This is why a course on AI is not training in a piece of software, it is an education in judgement, and teachers are the first who need it.
How a training course on artificial intelligence really works inside a school, told through the last one I ran.
In 2023 the Ministry of Education used the national recovery plan’s digital transition funds to open a line of financing so that schools could train their own teachers: it is decree 66, and each school spends it as it sees fit, choosing the subjects and looking outside for someone who can teach them. The «Quintino di Vona – Tito Speri» comprehensive school in Milan decided to devote part of it to artificial intelligence and published a call to select an expert teacher. I applied, and the appointment came in May 2025.
The problem in front of me was not explaining what a language model is. It was a wide staff room: primary school teachers and middle school mathematicians, people who had never opened a chatbot and people already using one to prepare tests, subjects whose problems with these tools are the opposite of one another. A single lecture for everyone would have satisfied a minority and wasted the afternoon of all the rest.
What came out was a course in seven stages between July and October, built backwards: first measure the people in front of me, then split by subject area, then the individual meeting where each person brings their own problem. And at the end a document, the school policy on what is allowed, what has to be declared and how it is assessed, because a course that ends with its last lesson leaves enthusiasm behind and nothing else, while a school needs a shared rule.
IC «Quintino di Vona – Tito Speri», Milan · «Il docente di STEM» project · with Pietro Stori · DM 66/2023, national recovery plan M4C1I2.1
An expert-teacher appointment of 20 teaching hours, inside a forty-hour course, completed and paid by the school. Before that, a training cycle with primary school teachers on integrating AI into Google Workspace. The course materials are public: daniloaprigliano.notion.site/materialicorsoai. In the school where I teach I sit on the standing committee on artificial intelligence, the one that has to write the rules of use for teachers and students: it is the inside face of the same work.
The format comes from a practical problem: in a staff room the starting levels run from people who have never opened a chatbot to people who use one every day. A single lecture serves half the room.
I measure skills and expectations before starting, against a recognised framework. It serves to build the groups and to waste the time neither of the beginners nor of those already ahead.
Every teacher brings their own tests, their own exercises, their own lesson plans, and we work on those. Good practice is shown at the start and checked immediately afterwards, on the actual case of the person learning it.
Twenty minutes each, remotely as well, by appointment. It is the part participants remember: in that quarter of an hour the specific problem of that person gets solved, the one they would never have raised in front of everyone.
A closing session pools the best ideas that came out of the meetings, and from there the school policy gets written: what is allowed, what has to be declared, how it is assessed. What one colleague has solved for her subject almost always serves someone else too.
I use only free tools that everyone can reach: a course that works only with the school’s paid licence stops working when the licence expires.
They adapt to the staff and to the school’s budget. On request I prepare a quotation with the hourly structure and the technical requirements.
One meeting for the whole staff: what these tools actually are, what changes in assessment, which uses hold up and which do not. For schools starting from zero.
The Di Vona format: questionnaire, plenary sessions, workshops by subject area, individual meetings, a final policy. It produces materials usable in class from the following week.
Primary, lower secondary and the two halves of upper secondary have different problems. I have worked with primary teachers and with the science area: the course is built around the people it is for.
What I build outside the courses, and leave online for whoever wants to use it.
prototype · benjamin
A prototype digital magazine on artificial intelligence at school: an editorial, articles, materials. I built it to have a place to think in public about what actually changes in teaching, instead of answering the same question over and over in the staff room. In Italian.
classroom material · il-risorgimento
An interactive timeline from the Congress of Vienna to the capture of Rome, written for my own classes and left online for anyone who wants it: every stage has its date, its event and a short explanation. In Italian.
Other work that belongs to this page although it lives elsewhere: the dashboard I wrote to hold archives, mail and search together – the practical proof of what I can build –, the accredited ECM course, adult training built from the idea to the catalogue, the Fifteen Martyrs of Piazzale Loreto, archival work done with students (i15martiri.daniloaprigliano.it), and the School of European Citizenship (scuoladicittadinanzaeuropea.it).