PlaybookBeginnerEU-wide9 min readTechnology, AI and Infrastructure#9

AI bootcamp for EU affairs, day 1: your baseline and your first prompts

Two downloadable worksheets, four questions that become your baseline, and one deliberately silly exercise that shows you what a model quietly fills in when you do not tell it.

Sebastián Rodríguez Pérez

Founder

European Campaign Playbook · Spain

Sebastián Rodríguez Pérez is a European campaign strategist with direct experience running and advising campaigns for some of Europe's largest pro-European organisations. He founded the European Campaign Playbook to codify practitioner knowledge across the EU's 27 political environments and make it accessible to campaign professionals, advocacy teams, and civic organisations working at the European level. His work spans electoral strategy, EU public affairs, digital campaigning, and the integration of AI tools into political communication workflows.

European campaign strategyEU public affairsDigital campaigningAI in politicsCross-border advocacyGEOSEO

Disclosure

Sebastián Rodríguez Pérez is the founder of the European Campaign Playbook and Campaign Intelligence Library. He has worked with pro-European organisations across multiple EU member states. He maintains editorial non-partisanship in all published content; where articles relate to campaigns or organisations he has directly worked with, this is disclosed within the article. He holds no current positions in active electoral campaigns.

european campaign playbook →LinkedIn →

What it covers

Day one of the AI bootcamp for small public affairs teams: the baseline questionnaire, the friction log, and a first prompting exercise on how role and context change what a model gives you.

Who it is for

Anyone in a small public affairs, policy or communications team who wants to get properly useful with AI, from complete beginners to people already building workflows.

When to use it

At the start of the bootcamp, before you change anything about how you work, so you have something honest to compare against at the end.

Key takeaway
A prompt sets who is speaking, who is listening and what matters. Where you leave a gap, the model fills it with an assumption and does not tell you.

In brief

Day one is about your work rather than the tools, because the bootcamp is measured against a baseline you write today.
Four questions: your current understanding, the last time AI actually helped, what eats your week, and what you would do with the time back.
Run a friction log for a few days: note every repetitive or manual task without trying to solve it.
Exercise 1 shows that the same question with a different role produces a different answer.
It then shows what the model assumes about a person when you give it one word, and what changes when you give it real context.

Day one starts with your work

Welcome to day one of the digital bootcamp on artificial intelligence for small public affairs teams.

Today is about you and your work. Not out of nosiness. The whole point of this bootcamp is that the way you work changes, and that you can actually observe the change rather than take my word for it. So on the last day we come back to the same questions and compare.

Download these two before you start

The baseline questionnaire has the four questions and the friction log. The exercise 1 worksheet has every prompt below, laid out with space to write your comparisons. Both are one-page-ish PDFs you can print or fill in on screen.

Everything in this article is in those two sheets. Read here, write there.

The four baseline questions

Answer these honestly and keep the sheet somewhere you will find it in a few weeks. There are no right answers. The only wrong move is being vague.

One: how would you describe your current understanding of AI?

Be specific. You are somewhere on this range, or nowhere near it, and either is fine.

Where are you right now?

Level	Sounds like
Completely new	I still do not really understand what it does.
Curious	I have tried ChatGPT a few times. Everyone seems excited and I am still working out why.
Experimenting	I have tried a few tools and I am still working out how to apply them to my actual work.
Regular user	I use AI for quite a few tasks. Give me the advanced material.
Building things	I am already building automations, connecting tools and creating my own workflows.

Two: when was the last time AI actually helped you with your work?

Actually helped. What did it help you do? If that moment has not happened yet, write down what you wish it could help you with instead. That answer is just as useful.

Three: what type of task takes most of your time?

Writing or editing reports. Research. Creating content or campaigns. Planning projects. Preparing meetings. Monitoring EU policy. Admin. Bureaucracy. Something else entirely. Pick the big one, the thing that eats your week.

Four: if you got some of that time back, what would you spend it on?

This is my favourite question. Developing new ideas. Thinking strategically. Talking to more people. Reaching decision-makers. Learning something new. Actually leaving work on time. Pick one or two, three at the most.

This is the baseline. On the final day you answer the same four questions again, and the difference between the two sheets is the real result of the bootcamp.

The friction log: notice, do not solve

Second piece of homework, and it runs in the background for the next few days. Every time you catch yourself doing something repetitive, slow or unnecessarily manual, write it down.

Writing the same type of email for the tenth time.
Moving information from one document into another.
Spending an hour turning a long document into a briefing.
Preparing the same kind of report every week.
Searching through fifteen documents to find one piece of information.

Do not try to fix any of it yet. Just notice. Some of those entries become very interesting later in the bootcamp, and the ones you catch in the moment are far more honest than the ones you would invent from memory afterwards.

Pick one tool and stay with it

ChatGPT, Claude, Gemini, Le Chat, whichever you prefer. Pick one and keep it for the whole bootcamp, because half of what you are learning is how your particular tool behaves.

We start with the basics and build from there: prompts, research, accuracy, workflows, automations, responsible use, and eventually something as serious as an AI policy for your organisation. The goal is that you become the person colleagues come to when they need help with AI.

And we start all of that with something slightly silly.

Exercise 1, part one: the same question, three roles

Open your tool, start a new conversation, and run the plainest possible version of the question. This is your control.

explain how the EU makes laws.

Read the answer and keep it. Now give the model a role and change nothing else. Run role A, then start again and run role B. Both prompts are in the prompt pack below and in the worksheet.

Then put the two answers side by side. Same model, same question, different role.

What to compare

Look at	Ask yourself
Vocabulary	Which words changed, and who would use them?
Level of detail	What got expanded, and what got skipped?
Acronyms	How many appeared, and were they explained?
Priorities	What did the model decide was the important part?
Tone	How is it speaking to you, and does it assume you are an insider?

Write down three differences. That is the first discovery: the role you give a model shapes the answer you get out of it.

Part two: what the model fills in when you say nothing

Now let us make it human. Start a fresh conversation and write:

explain how the EU makes laws to my grandmother.

Read that answer carefully, because something has happened in it that you did not ask for. The model has decided who your grandmother is. You never told it. So ask it directly, with the assumption audit prompt in the pack below.

Spend real time with what comes back. All you gave it was one word.

Maybe your grandmother is a professor. Maybe she left school at fourteen. Maybe she worked in EU affairs for thirty years. Maybe she runs a company. Maybe she lives in Indonesia, or Kenya, or a rural town near Cordoba. Maybe she has followed European politics every morning for forty years and understands the legislative process better than you do.

The word grandmother carries almost no information about the actual person. The system filled the gap anyway, and it did not mention that it was doing so.

Part three: swap the assumptions for information

Now run the same request with real context about a real person, using the third grandmother prompt in the pack. Then compare the two grandmother answers.

What changed?
What disappeared?
What became more specific?
What did the model stop assuming once you gave it something to work with?

Context about a human being makes the output dramatically more useful. And the gaps you leave get filled by something, whether or not you wanted them to be.

What day one is actually teaching you

Two things, and they carry most of the way through the rest of this bootcamp.

A prompt is a set of instructions about who is speaking, who is listening and what matters. Change any of those and you change the output.
Where you leave a gap, the system fills it with an assumption, silently. Noticing those gaps is most of what separates careless AI use from competent AI use.

Everything later, the research, the accuracy checks, the workflows, the automations, the policy, is built on those two ideas.

Your homework

Fill in the four baseline questions and put the sheet somewhere you will find it again.
Run the friction log for the next few days.
Do exercise 1 end to end, including the comparisons. The comparisons are the exercise.
Pick your tool and stay with it.

Then come back for the next one. It might be next week, it might be sooner. Things are busy around here, I am running this on my own, and I am trying to get a few of these out every week. If you would rather work through this material with other people in a room, that is what the live workshops are for.

Thanks for bearing with me, and welcome to the AI bootcamp for EU affairs. We are just getting started.

AI workflow suggestions

1.Pick one tool and keep it for the whole bootcamp.
2.Run the baseline prompt first and keep the answer as your control.
3.Change one variable at a time so you can see what each change did.
4.Ask the model to audit its own assumptions before you trust its output about people.

Prompt pack

1. Baseline, no instructions

explain how the EU makes laws.

2. Role A, the communications trainer

you are an excellent communications trainer.
you're known for explaining complicated political processes in language that ordinary people can understand and remember.
explain how the EU makes laws.
keep it clear, conversational and under 120 words.

3. Role B, the EU civil servant

you are an EU civil servant with 30 years of experience working with EU institutions.
you know the legislative process in great detail, you value technical precision, and you love institutional terminology and acronyms.
explain how the EU makes laws.
keep it under 120 words.

4. One word of context

explain how the EU makes laws to my grandmother.

5. The assumption audit

analyse the answer you just gave me.
what did you assume about my grandmother even though i never told you?
look for assumptions about her age, education, literacy, profession, political knowledge, knowledge of the EU, digital literacy, where she lives, nationality, cultural background, socioeconomic background, interests, cognitive ability and the type of language she needs.
for every assumption, show me what in your answer suggests you made it.
separate reasonable inferences from unsupported assumptions.
be specific.

6. Real context about a real person

explain how the EU makes laws to my grandmother.
she is 76, lives in a rural town near Cordoba, Spain, has a university education, reads the newspaper every morning and follows Spanish politics closely.
she has very little knowledge of how EU institutions work.
keep it warm and respectful.
use familiar comparisons only when they genuinely make the process clearer.
simplify based on her knowledge of the EU.
keep it under 120 words.

Checklist

Download the baseline questionnaire and the exercise 1 worksheet.
Answer the four baseline questions and store the sheet where you will find it.
Start the friction log and run it for several days.
Pick one AI tool for the whole bootcamp.
Run the baseline prompt and keep the answer.
Run role A and role B, then write down three differences.
Run the grandmother prompt and then the assumption audit.
Re-run with real context and compare the two answers.

Sources

1.european campaign playbook, AI bootcamp for EU affairs, day 1 (2026)