# AI Training Program

> A modular generative AI program for corporate teams and universities, from a maturity assessment to hands-on training and the co-design of the first use cases.

- Author: [Jacopo Marcolini](https://www.jacopomarcolini.com), Deployment Strategist & Product Design Lead
- Role: Program Designer & Lead Trainer
- Year: 2026
- Area: AI Enablement
- Tags: Generative AI, Training, Workshops
- Web page: https://www.jacopomarcolini.com/case-studies/ai-training-program

![The title AI Training Program on dark navy, next to two semi-transparent glass sparkles in violet and pink](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/hero-v2.jpg)

## Generative AI training for teams

I design and deliver generative AI training for corporate teams and universities. The program is modular: it starts by measuring where a team stands, continues with hands-on sessions on the team's own work, and ends with the design of the first use cases worth building.

So far I have spent more than 400 hours in the classroom with over 600 people, in groups of 5 to 80. The program has involved more than 25 companies and validated over 90 use cases, and it has also run at the University of Milan and the Free University of Bozen-Bolzano.

## Every team starts from a different place

A single session can bring together daily users of an AI assistant, people who tried it once, and people who are worried about what it means for their job. Some companies have already rolled out a licensed assistant; others have a usage policy and little else. The program had to:

- **Start from the team**: measure skills, tools and recurring tasks before the first session, and adapt the content to the results.
- **Be clear for non-technical people**, explaining how the models work in plain language.
- **Use real work**: practice on the documents, presentations and visuals people produce every week.
- **Cover the risks**: hallucinations, bias, copyright, jailbreaks and the AI Act, together with the company's own data policy.
- **Leave reusable material**: prompts, templates and assistants people can keep using, and a list of use cases for the next steps.
- **Work with 5 or 80 people**, keeping the hands-on part in both cases.

![Opening question on dark navy: how many of you have talked with an AI today? Below it, a single raised hand](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/moment-question-v4.webp)

![Slide listing what generative AI is not, ending with the answer: it's a tool](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/moment-tool-v2.webp)

![Diagram of an agent: a user delegates to a model that loops through context, tools and feedback](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/moment-agent-v2.webp)

![The AI Act risk tiers as three boxes: transparency obligations, high risk and unacceptable](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/moment-ai-act-v2.webp)

## From assessment to roadmap

The program has three phases, which can run together or separately. Alongside the tools, it covers change management and reskilling: I work with key users to support adoption and reduce resistance across the organization.

![The three phases on a line: maturity assessment, AI literacy and co-design, leading to an AI roadmap](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/path-v3.webp)

*The three phases of the program and the roadmap they lead to.*

### 1. Maturity assessment

A kick-off with the team lead to agree on goals, business priorities and the company's policy on AI and data. An individual survey then maps skills, the tools already in use and the tasks that take the most time. The results become a maturity report and a training plan matched to the team's level.

### 2. AI literacy

Hands-on sessions, in person, to bring everyone to the same level: a shared vocabulary, a working method and exercises on the team's own documents and visuals.

### 3. Co-design

The team maps its processes and ranks the ideas collected during training on a value and feasibility matrix. Each selected solution is described on a use case canvas (inputs, outputs, data, tools and KPIs), then prototyped together and tested on real cases.

### 4. AI roadmap

The work closes with a presentation to management: priorities, owners, the KPIs that will track results over time, and the next steps.

## What the training covers

### Foundations and prompting

How the models work, what they do well and where they fail. Responsible use, the AI Act and copyright. The structure of a prompt (context, examples, iteration) and techniques such as few-shot, role prompting, chain of thought and the RISEN framework, plus deep research and source checking.

### Documents

Drafting and reviewing structured documents, comparing files, extracting information, summarizing several sources, adapting tone and format, and saving what works as reusable templates.

### Visuals

Generating and editing images, infographics, presentations and data visualizations. Writing a visual brief, following brand guidelines and understanding the usage rights of generated content. Photography terms such as lens, depth of field, lighting and perspective become part of the prompt.

### Assistants and agents

Shared knowledge bases and custom instructions to standardize recurring tasks, connections to the tools the company already uses, and multi-step automations. What an agent is, and which decisions to delegate to it. The use case ideas for co-design are collected here.

![A tilted grid of training slides: prompting techniques, image tips, AI risks, the AI Act tiers, agents, a use case canvas and an impact and ease matrix](https://www.jacopomarcolini.com/images/case-studies/ai-training-program/slides-collage.webp)

*A selection of slides from the training materials.*

## How I run the sessions

Six habits keep the sessions practical, with groups of 5 as well as 80.

### Open with a question

Each session starts with a question to the room, for example how many people have talked with an AI that day. The answers tell me where the group stands and where to focus.

### It's a tool

Early on I clarify what generative AI is: a tool that produces content from patterns in its training data, with no understanding, no autonomy and no guarantee of being right. This sets realistic expectations for the rest of the program.

### Example first

Each technique is shown on a concrete prompt and its output before it gets a name, so the example is what people remember.

### Their own material

Exercises use the documents, decks and visuals mapped during the assessment, so the results can be used at work the next day.

### Risks with examples

Bias, deepfakes, copyright and jailbreaks are shown with real cases, together with the guardrails that limit them and the AI Act risk tiers.

### Small teams

The final part is hands-on in small groups, each working with the tools on a problem it chose.

## Working alone or with tutors

I design each program end to end: the assessment, the agenda, the exercises and the materials. I deliver it myself to small groups. With larger ones I lead a team of tutors: I run the plenary sessions, then the room splits into small teams, each followed by a tutor, which keeps the exercises manageable with up to 80 people.

The materials are updated as models and tools change, and the same structure works in companies and in university courses.

## What teams take away

Teams finish the program with a shared vocabulary and method for working with generative AI, a library of prompts, templates and assistants built during the sessions, and a backlog of use cases. Teams that complete co-design also have tested prototypes, KPIs to measure them and a roadmap with priorities and owners.

Several programs led to follow-up projects and long-term partnerships with the companies involved.

## Key numbers

- **400+** Training Hours
- **600+** People Trained