AI Skills · Germany · Updated June 2026

How Can German Companies Train AI Engineers at 100% Funded Cost? The 2026 QCG Guide

German companies with fewer than 50 employees can train an existing employee as an AI engineer at up to 100% funded cost through the Qualifizierungschancengesetz (QCG) – a Bundesagentur für Arbeit programme that covers course fees and up to 75% of the employee's wages during training.

The fastest way for a German company to get an AI engineer runs through its own payroll: train someone you already trust, and let the state cover most of the bill. Most owners assume the only options are an expensive external hire or a consultancy retainer.

The Bundesagentur für Arbeit funds a third route through the Qualifizierungschancengesetz (QCG): up to 100% of the course cost and a large share of the employee's wages while they train. For a firm under 50 staff, that can mean turning an existing employee into an AI engineer for close to nothing.

What is the QCG, and which German companies actually qualify?

The QCG exists because Germany cannot hire its way out of its skills gap. Our 2026 Germany Tech Job Market report puts a number on it: roughly 109,000 IT specialist roles sit unfilled, and 79% of firms expect the shortage to worsen as the workforce starts shrinking in 2026 (Bitkom, 2025). The state would rather fund the skill than pay for the unemployment that a frozen economy creates.

For a small company the math is blunt. You keep an employee on payroll, the state reimburses most of their wage while they train, and you own the resulting skill forever. Compare that to recruitment, where a senior AI hire commands a six-figure package and the open role sits empty for the 7.7 months it takes, on average, to fill an IT position in Germany.

Who qualifies on the employee side:

• Staff in roles exposed to automation or technological change.
• Employees whose last formal training ended four or more years ago.
• Workers of any age, though employees 45+ at mid-sized firms unlock higher funding rates.

How much does the state actually pay?

The wage subsidy is the part owners overlook. The state does more than pay for the seat in the classroom; it pays you back for keeping the person employed while they sit in it. A €6,000 course at a firm under 50 staff can cost the owner close to nothing, and the salary you keep paying during those weeks gets cushioned by three-quarters. Run your own numbers in the estimator further down.

Why act now? The talent gap widens while competitors move

Two forces are pulling the German market in opposite directions, and the combination is what makes upskilling urgent. The broad market has cooled – unemployment reached 6.3% in May 2026 and open roles are down about 10% year on year. Underneath that, the shortage of specialist and AI talent keeps deepening (Germany Tech Job Market 2026). The companies that can build AI into their operations are the ones taking share while the market is soft.

German employers expect AI to raise technology demand. In Bitkom's survey of 855 firms, 42% expect AI to increase demand for IT specialists and the same share expect entirely new IT job profiles to appear. The firms creating that demand are the ones that move first.

Regulation adds a second deadline. The EU AI Act has carried AI-literacy obligations since early 2025, with high-risk obligations phasing in toward 2027. Building AI competency inside your team is now a compliance requirement as well as a growth play.

What can one trained AI engineer build for a small company?

The skills employers ask for are concrete. Our report's analysis of German AI postings names Python, large language models, retrieval-augmented generation (RAG) and MLOps as the core, with LangChain, LangGraph and vector databases around it. Four projects show up again and again as the fastest payback for small German firms:

• Inbox triage and reply drafting. An engineer connects an LLM to the shared mailbox. Incoming mail gets classified, urgent items flagged, and a draft reply written for a human to approve. A two-person desk handles the volume of five.
• Quote and proposal generation. The system reads a request, pulls pricing from past deals, and produces a draft quote in minutes instead of the afternoon it used to eat. Faster quotes win more deals.
• Internal knowledge search. Years of PDFs, contracts and wiki pages become searchable in plain German. New hires stop interrupting senior staff for answers buried in a 2021 folder.
• Data entry elimination. Invoices, delivery notes and forms get read and entered automatically, removing the manual keying that produces both errors and overtime.

These are real projects our learners shipped during the course – each one a working application you can open and click through. Select any card to view the project.

Browse the full project showcase →

How does AI engineering fix documentation – a worked example?

Here is how the numbers tend to work for a machine-building SME, as an illustrative example. Each delivered machine ships with a technical handover document: specifications, wiring notes, maintenance schedule, safety sign-off. Writing one by hand takes a senior technician about four hours. A firm shipping forty machines a year burns roughly 160 hours – a full month of skilled labor – retyping information that already sits in the project files.

An engineer trained on the course can build this:

1. The system ingests the project's CAD notes, the parts list and the technician's voice memo from the final inspection.
2. An LLM drafts the full handover document against the firm's approved template.
3. The technician reviews and corrects the draft instead of writing from scratch.
4. The corrected version trains the next draft to be sharper.

Four hours per document dropped to thirty-five minutes of review. The technician went back to building machines. Labor cost per document fell into the $8–$12 range you see across document automation projects – multiplied by every document, every month, forever.

The deeper point: no vendor sells an off-the-shelf tool that understands these machines, this template and this firm's sign-off rules. The trained employee builds exactly that, because the employee already knows exactly that.

QCG vs. hiring: what does each route really cost the business?

When we run cohorts of working professionals, the pattern repeats: the upskilled internal employee outperforms the external hire on the projects that matter, because they already understand the messy reality of your invoices, your customers and your handover process. The outside hire has to be taught all of that. Your employee lived it.

The bigger QCG payoff: revenue and competitive position

Cost savings have a floor. You can only cut so much. Revenue and market position have no ceiling, and that is where the funded engineer earns the company its return twice over.

In that example, the freed month of technician time converts into roughly three more machines a year. At a typical order value, that is six figures of new revenue from a capability the state paid to install. The saved labor cost is the smaller half; the added throughput, and the orders won on faster quotes, is the larger one.

Frame it for a skeptical owner this way: the Bundesagentur für Arbeit funds a capability that pays the business twice over – capacity reclaimed and revenue added.

Two funded courses: which one fits your team?

Match the course to the person. A product manager who wants AI to draft specs and triage tickets fits Building with AI. A developer who will own the company's internal tools fits AI Engineering. Both run under the same QCG funding rules.

How to apply this tomorrow

• Pick the workflow. Walk your floor and find the task a skilled employee complains about most. That complaint is your first AI project.
• Pick the person. Choose someone who understands that workflow and wants to grow. Domain knowledge beats prior coding experience.
• Confirm funding. Contact an AZAV-certified AI engineering provider and ask them to check QCG eligibility for your company size and the chosen employee. They submit most of the documentation.
• Set the first project before training ends. The best cohorts have the employee build their real internal tool during the course, so it ships the week they return.

Our team at Turing College runs a project-based AI Engineering course built for exactly this – working professionals shipping real LLM applications, with QCG funding that can cover the full cost for eligible companies.

Frequently Asked Questions

Is the QCG only for unemployed people?

No. The QCG funds currently employed staff who want to upskill without leaving their job. The unemployed route is the Bildungsgutschein, a separate voucher that covers 100% of tuition. Both run through the Bundesagentur für Arbeit; the QCG is the one for people still on your payroll.

How long does a funded AI engineering course take?

QCG-eligible programs must exceed 120 hours. In practice, AI engineering courses range from roughly twelve weeks of intensive, project-based study to around 30 weeks for longer formats. The program must be AZAV-certified to qualify for funding.

Does my employee need a programming background to qualify?

It depends on the course. AI Engineering suits staff with programming experience. For non-technical employees – product managers, operations, marketing – the Building with AI course needs no coding background and teaches them to apply AI tools and automate their own work. For both, domain knowledge matters: an employee who understands your workflows builds more useful tools than an outsider who knows only the code.

What is the catch with 100% funding?

The main conditions are size and timing: the 100% course-cost rate applies to companies under 50 employees, the program must exceed 120 hours and be AZAV-certified, and the employee's last comparable training must be at least four years in the past. The wage subsidy and exact rate scale with company size.