Question 1

What does a C-level AI consultant actually do?

Accepted Answer

I advise your C-level on AI — and prove it by building the first POC myself. In practice that is two jobs in one person: the strategic sparring at management level, and the engineering that turns one decision into a working system. Most AI consulting stops at the recommendation. As an AI consultant in Hamburg working with German companies, I run both halves: working sessions with the C-level to decide which AI use cases are worth pursuing, and then the hands-on build of the first proof of concept on your data. The short version of the role: “AI sparring partner for C-level executives” — with an engineer’s hands. The three packages are described on the services overview.

Question 2

How does a time-boxed AI POC engagement work?

Accepted Answer

In three time-boxed packages — Discovery → POC → Enablement — that together form the De-risked AI ladder. Each package has a fixed scope, a fixed timeframe, and defined deliverables agreed before the work starts, and every engagement ends with a handover. It is explicitly not a retainer. Discovery ranks the AI use cases worth building in your company and scopes the strongest one, with its success metric defined. POC is the first system, built by me on your data and measured against that metric. Enablement trains your developers and your C-level to run, extend, and govern it without me. At the end your team is in the driver’s seat; if a new use case appears later, the ladder simply starts again at Discovery — there is no standing fee in between.

Question 3

What does AI consulting cost?

Accepted Answer

Engagements start from €5,000 per month. Every engagement is one of three fixed-scope packages: the price is agreed up front and tied to measured deliverables — not a retainer, never open-ended. You know before we start what you will hold in your hands at the end — a written assessment after Discovery, a working measured system after the POC, a team that runs it after Enablement. The exact figure depends on the package and the scope we agree in the first call. What I deliberately don’t sell is an open-ended monthly fee — a model that pays the consultant most when you stay dependent longest. How the three packages work →

Question 4

How do you measure whether an AI POC worked?

Accepted Answer

Against one success metric defined before the build starts — measured on the POC, on your data, in numbers you can check. I never promise blanket percentage savings; I promise to measure. Vendors quoting industry-average savings are quoting other people’s companies. Every POC I build begins with the metric written down — answer accuracy on your documents, time per case, throughput, whatever the use case actually needs — and ends with the measurement against it. If the system clears the bar, you scale with evidence. If it doesn’t, a time-boxed budget bought you that answer before any real commitment. That is what “De-risked AI” means, and the goal in one line: “Vom AI-Hype zum laufenden System” — from AI hype to a running system.

Question 5

Do we need our own AI team before starting?

Accepted Answer

No. The model is designed for companies without an AI team: I build the first system myself, and the Enablement package then puts your existing people in charge of it. AI Enablement is working sessions with the C-level plus hands-on training for your developers — the people who already run your systems learn to run, extend, and govern the new one. The engagement ends with a documented handover: source code, architecture notes, and a team that doesn’t need me. If you later decide you want permanent AI leadership in-house, Enablement will have prepared your team to work with that hire.

Question 6

Build vs. buy: should a Mittelstand company build its own AI tools?

Accepted Answer

Buy what is generic; build where the value sits in your own data and processes. Off-the-shelf copilots cover commodity tasks — the cases that move your numbers are usually the ones no vendor ships, because they depend on knowledge only your company has. In AI consulting for German Mittelstand companies the honest answer is almost always mixed: standard tools for writing, transcription, and general assistance — a custom system where your differentiation lives, such as retrieval over decades of your own service records, quotations, or diagnostic knowledge. The expensive mistake is answering this in the abstract. AI Discovery exists to make the build-vs-buy call per use case, ranked by value and feasibility, before any money goes into building.

Question 7

How do you handle GDPR and the EU AI Act in an AI pilot?

Accepted Answer

Soberly and from day one: the POC runs on infrastructure and data agreements your IT signs off on, personal data is minimized or excluded, and every decision is documented. I’m an engineer, not a lawyer — legal sign-off stays with your data-protection officer or counsel. Concretely that means EU data residency where required, processing agreements with any model provider, no training on your data, and a written record of what data the system touches and why. On timing: the EU AI Act entered into force on 1 August 2024 and the bulk of its obligations applies from 2 August 2026 (implementation timeline, artificialintelligenceact.eu). Most POC-stage assistant and retrieval systems are not high-risk under the Act — but that classification is checked per use case and documented, not assumed.

Question 8

Why hire one advisor-builder instead of a strategy consultancy or a dev agency?

Accepted Answer

Because the gap between strategy and delivery is where AI projects die. A strategy consultancy hands you a deck and leaves the building to others; a dev agency builds what it is told but never sits with the C-level. I do both ends myself, so the recommendation and the system come from the same person. There is no handoff in the middle — nobody to translate between the boardroom and the codebase, and no one for the strategist to blame when the build disappoints. What you give up is an army of juniors; what you get is accountability in one person and a proof layer you can check: systems that shipped to production, not slideware.

Question 9

Can you work with our existing IT team?

Accepted Answer

Yes — that is the design. Your IT keeps running its systems; I build the POC alongside them with access agreed up front, and the engagement ends with a full handover: source code, architecture notes, and documentation. No dependency on my infrastructure remains. Heads of IT get the answers they actually need before anything touches their stack: what the architecture is, where data flows, what the security posture is — in writing. I don’t arrive with a fixed stack: the technology is chosen to fit your business and your existing IT landscape — deliberately boring and maintainable, in conventions your team can read and run. Each case study carries a “For the Head of IT” box for exactly this audience, because I expect every claim to be googled.

Question 10

Who is Hendrik Kleinwächter?

Accepted Answer

Hendrik Kleinwächter (also written Hendrik Kleinwaechter) is an AI consultant and engineer in Hamburg, Germany. He advises C-level executives on AI and builds the first proof of concept himself — roughly 20 years of software engineering, AI-native since 2023, 50+ AI applications shipped to production. He was CTO at AHEAD Automotive, where he built a RAG diagnostic platform used by 50,000+ mechanics and processing 2M+ queries per month. He founded three startups; one (tripl) was acquired by trivago in 2017 (Handelsblatt). He runs “The Bread Code”, a YouTube channel with 300,000+ subscribers on the engineering of sourdough bread, and publishes open source at github.com/hendricius, including The Sourdough Framework, an open-source book. He works in German and English. More: about · full CV.

AI consulting FAQ: straight answers, no hype