Smart Citizenship. Presentation text

Open warm, unhurried. Let the title sit for a second before you talk.

“Welcome back. This next part is the practical one — it’s about using AI in your actual week, not the big-picture debate.

Here’s the metaphor I want you to hold: a lever. A lever doesn’t do the work for you — it multiplies the effort you put in. AI is the same. Used well, with the right tool and the right guardrails, it gives you back capacity for the mission. Used carelessly, it quietly creates risk you’ll have to own later.

So I’m not here to sell you AI. I’m here to make sure that by the end you walk out with three concrete things you can use Monday: a tool you’re actually cleared to run, a checklist you own, and one task you can show next time.

And the one line I want you to remember the whole way through — it’s right there at the bottom — AI advises. You decide, and you sign.”

The empathy beat. Slow down. They should feel seen.

“Let’s be honest about where you’re sitting. The ask keeps growing — more reporting, CSRD, more partners, more expectations on corporate citizenship every year. And the team? The team doesn’t grow at the same rate. Most of you are doing more with the same two or three people.

That’s the real reason AI is on your desk. Not because it’s exciting — because it’s the lever that gives some of that capacity back. And it points two ways.

Inward — your own work. Reporting, partner research, the internal pitch. Anything you do faster gives you more time for the part that actually needs a human.

Outward — and this is the part I don’t want us to lose — toward the nonprofits you back. Your company’s AI capability, pointed at them: lend a data team for a two-day model review. Fund a partner’s EU-hosted tooling. Scope one pro-bono AI project a quarter. That’s smart citizenship, not just a faster inbox.

Both directions run on the same principle: the right tool, used safely, with a named person accountable. Not ‘the company.’ Not ‘the AI.’ A person. Usually you.”

The map they asked for in the survey. Be concrete about the colours.

“Before any tool, here’s the map. Take your value chain and colour every step green, amber, or red.

Green — go. Partner research. Drafting a CSRD or impact paragraph. Low stakes, you’re checking the output anyway. Use AI freely here.

Amber — caution. Scoring applications. Stakeholder comms. AI can help you prepare, but a human reads every line before it moves.

Red — you decide. The grantee decision. Who gets the funding. That is never delegated to a model — not because the model can’t produce an answer, but because that answer is exactly the thing your role exists to own.

The skill isn’t ‘using AI.’ The skill is knowing which colour you’re standing in. The point of this whole session is that you own the responsible-AI call — and the colour tells you how tight to hold the wheel. AI advises. You decide, and you sign.”

Set it up in one breath, then play the ~93-second clip. Don’t talk over it — let it run. Pick the thread back up at the end.

“Let me show you what that discipline actually looks like, start to finish. Watch the checklist on the right — that’s the part we bring, not the AI.

▶ Play the demo. ~93 seconds. Stay quiet.

[After it ends:] See what happened there? The AI drafted in seconds. But it invented a number — ‘4,200’ — that was never in the notes. The checklist caught it. We sent the correction back as plain feedback, it revised, and only then did a named person sign and send. The AI never decided anything. It advised. We decided. We signed. That’s the whole method on one screen.

Now name the handover — this is the bridge to the take-home.

And one thing about that demo: it deliberately shows one task, start to finish, because the goal there is to build the reflex — draft, catch, sign. The fuller toolkit lives in the field guide you’ll take home. That has what a 90-second clip can’t: the six tools and which fits your boundary, ten ready-to-paste prompts, the budget pitch, and the outward playbook. Think of it as inspiration to adapt to your own desk — a starting point, not a rulebook. The demo teaches the habit; the guide is the reference you keep.”

Practical, honest. They want to know what they’re actually allowed to use.

“So which tool? The honest answer is: pick the one that fits your data boundary, and start narrow.

If you already run Microsoft 365 — Copilot is your Monday move. Your data stays inside the boundary you already have. The shortest path to a tool you’re allowed to use, today.

Langdock or Mistral — EU-hosted, multi-model, good for a pilot. They’ll name-drop Merck and BASF — verify that for yourself, don’t take the vendor’s word.

DeepL — narrow and safe. Translation and write-assist, built in the EU. Does one thing, does it well.

Briink, SAP Joule — specialists for impact, ESG, ERP work. More capable, but slower to clear — expect a data-protection assessment first.

Here’s the trap, though. EU-hosted is the start, not the finish. GDPR-capable is not the same as compliant — it still needs a DPA, a DPIA, and your Betriebsrat’s sign-off. And whatever the tool: green for drafting and research, red the moment real personal data or a funding decision is in play. The tool doesn’t change the colour. You do.”

Shift energy — this is the one that helps them in their next budget meeting. Make it useful.

“Now — most of you don’t control this decision alone. You need a yes from someone. So here are three arguments that actually move a budget.

One — the law is already on your side. The EU AI Act’s staff-literacy duty, Article 4, has applied since February 2025. Most organisations aren’t compliant yet, and enforcement begins August 2026. Training your people on AI isn’t a nice-to-have — it’s a duty you’re already behind on.

Two — there’s a leak open right now. Be honest: your staff are already pasting company data into private ChatGPT accounts. A sanctioned tool doesn’t open a risk — it closes one that’s open today.

Three — don’t ask for a rollout. Ask for a pilot. Five licences, one team, one measured metric. Small enough to say yes to.

“An eight-week pilot — five licences for the CC team, measured on hours saved in CSRD reporting. About €25 per user per month. Owner: me.”

Specific, cheap, owned. That’s a yes.”

The takeaway the senior people came for. Let it feel like a gift they can use.

“The single most useful thing you can leave here with is a position — one short paragraph your organisation actually stands behind. Something like:

“We use AI to extend our impact, never to replace judgment. Staff use sanctioned, EU-hosted tools inside our data boundary. AI drafts and scans; people decide and sign.”

Take that, adapt it, and get three signatures on it: CC lead, data protection officer, Betriebsrat. Now you’ve replaced ‘everyone quietly doing their own thing’ with something owned.

And then point it outward — this is where it stops being about your inbox and starts being about impact. Map where a nonprofit you back actually needs AI. Scope it with a specialist partner. Fund responsible AI, not just any AI. Lend your people through skilled volunteering. And pass on what worked, so the next nonprofit doesn’t start from zero. Partners like MI4People, N3XTCODER, CorrelAid, Civic Coding already do this — you don’t have to build it alone.”

Close crisp and warm. Make it a real assignment, then send them in.

“So here’s my one ask. Don’t try to transform everything. Pick one task, one sanctioned tool, one result you can show next time — with your name on it, not the AI’s. Maybe:

draft a CSRD paragraph in Copilot and run the checklist before you sign — summarise five partner reports and verify every figure — or scope one pro-bono AI ask with a nonprofit you fund. The field guide on your desk has the prompts to start from.

That’s it. AI advises. You decide, and you sign. Let’s take that into Breakout 2 and make it concrete for your own organisation.”

→ Hand off to Breakout 2.