One AI tool for everything is like one employee for everything. They'd be wrong. A lot.
In episode two, Emma and George get practical. Chapter two of EMBRAICE is about AI in your daily leadership life — tools, models, prompting, and the art of getting what you want out of a machine without sounding like you're ordering from a vending machine.
Plus: George reads a passage from the book that makes Emma uncomfortably self-aware. And Sven's marriage becomes briefly relevant.
What this episode is about:
Not all LLMs are the same. And using the wrong model for the wrong task isn't just inefficient — it's a waste of your most valuable resource: your attention.
The lineup:
- ChatGPT — the generalist. Fast, versatile, good for first drafts and brainstorming. Your Swiss Army knife.
- Claude — more nuanced. Better for longer documents, complex analysis, careful reasoning. Sven's personal co-pilot for presentations, speeches, and LinkedIn posts. Hopefully he manages without Claude when he talks to his wife.
- Gemini — Google's model. If you live in Docs, Sheets, and Gmail, this becomes a natural co-pilot.
- Perplexity — the researcher. Searches the actual current web. Cites sources. Gives you answers you can verify. The one tool Sven recommends most specifically for market analysis and competitive research.
The rule of thumb: right tool for the right job. Not one app to rule them all.
Prompting like a pro: the CRAFT framework.
The quality of your output is almost entirely determined by the quality of your input. CRAFT is the framework:
- C — Context. Tell the AI what situation you're in.
- R — Role. Who should the AI be? Expert consultant? Critical friend? Legal advisor?
- A — Adjustments. Constraints, tone, length, what to avoid.
- F — Format. List? Narrative? Table? Presentation outline?
- T — Type. What kind of thinking should the AI apply — analysis, brainstorming, summarizing, challenging?
Plus a set of prompting shortcuts worth memorizing: /ELI5 — Explain it like I'm five. For the technical briefing you need to survive before walking into a room full of engineers. /STEEL MAN — Build the strongest possible argument against your own position. Win the argument in rehearsal before you walk in the door. /SOCRATIC MODE — The AI only asks questions. Deeply annoying. Deeply useful. /ROLE FUSION — Think like a CFO who has also studied behavioral psychology. The model actually shifts.
Think in agents, not tools.
Don't just use AI for individual tasks. Build specialized agents with defined roles and personalities. Sven's AI dream team includes Maria the Research Analyst, a legal advisor, a financial controller — and Emma the strategy coach. None of whom take holidays. None of whom send emails at 11pm saying they need to talk.
The principle is identical to good team management: don't hire one person to do ten jobs.
The rule he gives: "If a task is recurring, rule-based, and documentable — AI can take it over. What remains is what you get paid for."
Eight guardrails. And not one of them says "stop using AI."
Critical thinking over autopilot. Data security — everything you type into a cloud-based AI leaves your company. Transparency with your team. Legal gray zones. Bias and fairness. And above all: AI stays a tool. You stay the boss.
The takeaway:
Diversify your tool set. Learn to prompt properly. Start thinking in agents.
The tools are the instrument. You still need to be the musician.
The book: EMBRAICE — The AI Guide for Leaders by Sven Dörrenbächer. Currently available in German only. Order it at your local bookstore.
📩 [email protected]
A special thanks to Michael Ohanian aka Don Swing for the smooth house beats that set the mood — and to Veit Kment and Paul Dörrenbächer for the OpenClaw architecture that gave this whole thing a solid backbone to stand on. They know what they built. And now so do you.
