Study by Der Entrepreneurs Club & appliedAI: AI Potential in German Family Businesses

The German labor market presents a paradox: while parts of industry are cutting jobs, many companies simultaneously fail to fill their open positions. This is precisely where artificial intelligence offers substantial potential — not as an abstract transformation vision, but with one concrete question: which of the open positions can be augmented with AI?

To answer it, this study takes a pragmatic approach and looks at the most tangible entry point: the open positions that are being newly advertised anyway. The basis is an analysis of 5,280 currently advertised positions at German family businesses — evaluated for AI exposure and substitutability, and translated via average salaries into an affected salary volume.

21% of tasks are AI-substitutable on average — a calculated 1,118 full-time equivalents

€74.6M of the annual gross salary volume is attributable to AI-substitutable activities

The key findings at a glance

• On average, 21% of the tasks per position are AI-substitutable — a calculated 1,118 of 5,280 full-time equivalents. For 40% of positions, the substitutable task share is at least 30%.
• Measured by salary volume, the potential is greater than by headcount: €74.6M, or 25.4% of the annual gross salary volume, is attributable to AI-substitutable activities — higher-paid roles are systematically more affected.
• Highest relative lever in knowledge-intensive office functions: Finance & Accounting (51%), Legal & Administration (46%), Advertising & Marketing (39%). Highest absolute lever by volume: IT & Software and Corporate Management.
• Manual-physical occupations (Production & Trades 6%, Logistics 10%) are barely substitutable — here the AI lever lies elsewhere.
• AI hits qualified, expensive knowledge work: academic roles, at 36%, are roughly three times as substitutable as non-academic ones (11%).

The data basis was all positions advertised on www.karriere-familienunternehmen.de, a job portal for family businesses operated by Der Entrepreneurs Club. Each analysis drew on the job title and the job description (the "job posting"). The cut-off date for the analysis was June 18, 2026.

At the time of the evaluation, the portal listed job postings from 350 family businesses. These companies belong to 52 family business groups with revenues ranging from around €36M to €185.6B (median around €1B) and a German workforce of 250 to 200,000 (median around 2,050).

Before the evaluation, all positions at student or qualification level were excluded — specifically working-student positions, internships, theses, dual study places, and master's theses — since these do not reflect regular staffing needs. After this cleaning, 5,280 evaluated positions from 325 family businesses remain.

The methodological starting point is the evaluation approach developed by Andrej Karpathy for the US labor market. The guiding question per role is how strongly positions or activities are "AI-exposed" — that is, how strongly a position or activity is potentially subject to change under the increasing adoption of AI.

A more detailed analysis of the specific activities and tasks stated in the job postings was then carried out. From this, the percentage share that could potentially be taken over through the use of AI was derived.

Finally, based on the average salaries for the affected positions, the study calculated the affected gross salary volume and the associated personnel costs. This makes it possible to express the potential not only in positions but also in euros.

This yields the two central metrics per position:

• AI exposure (scale 0–10): a measure of how strongly AI changes the activity overall.
• Substitutability (AI ratio, 0–100%): the share of core tasks that could be taken over by purely cognitive, digital AI.

An automated NLP pipeline based on modern Large Language Models (LLMs) was used to evaluate the job postings. The evaluation ran in three parallel tracks: classification and substitutability (AI ratio), AI exposure (Karpathy Score), and salary estimation.

• Zero-shot procedure: each vacancy was evaluated as an isolated data package in order to rule out cross-effects or systematic distortions from previous evaluations.
• Guardrails against hallucinations: physical and interpersonal activities (order picking, assembly, machine operation, pedagogy) were explicitly excluded from the AI ratio via prompt restriction (ratio = 0%); robotics explicitly does not count as AI.
• The Karpathy Score follows a heuristic modeled on US Bureau of Labor Statistics criteria with a calibrated 10-point scale; full digital feasibility (remote capability) serves as the primary indicator of high exposure.
• The salary figures are based on an AI-supported interpolation from job title and location (gross annual salary in EUR, incl. currency conversion and special formats).

Aggregation was carried out via the position volume per group. The metric "calculated positions replaced" is the sum of the substitutable task shares (number of positions times average substitutability). The affected salary volume is derived analogously from average salary times substitutability per position.

The features generated per position by the pipeline include substitutability, AI exposure, academic degree (academic / non-academic), seniority, and estimated salary. Since the data basis contains no predefined occupational classification, positions were additionally grouped into function clusters via a rule-based procedure based on the job title (e.g. Finance & Accounting, IT & Software, Sales & Procurement, Production & Trades). This procedure assigns around 87% of positions unambiguously; the remainder are listed as "Other." The clusters are oriented toward common occupational groups following the German Classification of Occupations 2010 (KldB 2010) of the Federal Employment Agency.

Across all 5,280 positions, average substitutability is 21%. That corresponds to a calculated 1,118 full-time equivalents that could be replaced with AI available today. More telling than this average is the distribution: for 40% of positions, at least 30% of the task share is substitutable - so the potential is concentrated in a clearly definable subset of open positions.

Average AI exposure, at 5.1 out of 10, is higher than substitutability. AI therefore changes many activities noticeably without fully replacing them - full substitution is the narrower special case of the broader change.

At the function level, a clear pattern emerges: knowledge-intensive office functions show the highest substitutability and exposure. The following table shows the function clusters with the highest relative lever.

Table 1: Function clusters with the highest relative AI lever (excerpt). Subst. = substitutability, Expo. = AI exposure. Values rounded.

The largest absolute effect arises not in the most substitutable functions but in the highest-volume ones. IT & Software and Corporate Management & Organization lead because of their high position volume, even though their average substitutability is only mid-range.

Table 2: Function clusters with the highest absolute lever (calculated positions replaced). Values rounded.

At the lower end are the manual-physical functions: Production & Trades (6% substitutability) and Logistics & Transport (10%). This is not an all-clear but an indication that the AI lever in these areas is unlocked not through the job posting but through process and plant technology.

Substitutability is not distributed evenly across qualification levels. Academic roles, at an average of 36%, are roughly three times as substitutable as non-academic ones (11%). By seniority, substitutability rises with responsibility: executive roles are at 31%, professionals at 23%, whereas entry-level positions are only at 8%.

Translating the task potential into personnel costs reveals the actual economic lever. The total annual gross salary volume of the positions examined is around €293M. Of this, €74.6M — that is, 25.4% — is attributable to AI-substitutable activities.

This share is above the headcount-weighted substitutability of 21%. The reason: higher-paid roles are systematically more affected (correlation of 0.47 between salary and substitutability). Measuring the potential only in positions underestimates the financial lever.

The results should be understood as a frame of orientation, not as an exact forecast. The following constraints should be kept in mind when interpreting them:

• Exposure, not an employment forecast: a high value measures how strongly AI changes an activity — not the probability that the position will be eliminated. A function can be highly exposed and still grow if the increased productivity leads to higher demand.
• Calculated, not real replaceability: "substitutable" denotes the technical automation potential of the described tasks, not a statement about actual job cuts. Organizational, legal, and cultural factors are not taken into account.
• Role-based only — a conservative lower bound: effects are captured at the level of the individual role. The potentially greater lever of an AI-supported process and organizational design is not represented. The values are therefore rather a conservative lower bound.
• Job-posting data basis: what was evaluated is the job title and the advertised task description, not the role as actually lived. Postings are often generic and represent activities incompletely.
• LLM-supported evaluation: the metrics were generated in a zero-shot procedure by Large Language Models. Despite guardrails and deterministic output, a model-inherent fuzziness remains; a systematic human validation of the individual scores was not the focus.
• Salary figures are estimates: the gross salary volumes are based on AI-interpolated average salaries, not on actual compensation data. The euro values should be read as an order of magnitude.
• Rule-based function classification: the assignment to function clusters is keyword-based on the job title; around 13% of positions remain "Other," and individual assignments can be imprecise.

1. Task analysis before the posting: before every new hire, check which parts of the role are already AI-capable today — substitutable or augmentable.
2. Adjust the role scope instead of a 1:1 replacement: recut the requirements profile based on the analysis instead of repeating the posting from three years ago.
3. Build AI competence into the profile: anchor tooling and AI application competence as a requirement, especially in the high-leverage functions Finance & Accounting, Legal & Administration, Marketing, and IT.

The paradox described at the outset can be resolved precisely here: where positions are hard to fill, AI can take over or augment parts of the tasks — and recut the role so that it remains realistically fillable or gets by with fewer staff. New positions also bypass the biggest bottleneck of any AI adoption: the resistance of established teams to change. This way, AI does not become the next major project but a natural component of every staffing decision.

• Methodological starting point: AI exposure approach after Andrej Karpathy (US Job Market Visualizer), adapted for the German context.
• Calibration of exposure: criteria following the US Bureau of Labor Statistics (BLS).
• Occupational systematics of the function clusters: German Classification of Occupations 2010 (KldB 2010), Federal Employment Agency.

About appliedAI

appliedAI is Europe's largest initiative for the application of trustworthy AI technology. The initiative was established in 2017 by Dr. Andreas Liebl as a division of UnternehmerTUM Munich and transferred in 2022 into a joint venture with the Innovation Park Artificial Intelligence (IPAI) Heilbronn. At its Munich and Heilbronn locations, more than 130 employees pursue the goal of making European industry a shaper in the AI era, in order to preserve Europe's competitiveness and actively help shape the future. appliedAI supports international corporations, including BMW and Siemens, as well as mid-sized companies holistically in their AI transformation. This is achieved through collaborative exchange and the joint building of knowledge, through comprehensive accelerator programs, and through specific solutions and services. Further information is available at https://www.appliedai.de/en/

About Der Entrepreneurs Club

Der Entrepreneurs Club is a network of, and service provider for, entrepreneurial families and leading family businesses. Founded in 2005 by Stefan Klemm as an entrepreneurs' club for resolving business successions, Der Entrepreneurs Club today operates, among other things, the recruiting and contact fair series "Karrieretag Familienunternehmen" (https://www.karrieretag-familienunternehmen.de) under the patronage of the Federal Minister for Economic Affairs and Energy, as well as the employer branding and job portal "Karriere im Familienunternehmen" (https://www.karriere-familienunternehmen.de). Der Entrepreneurs Club develops entrepreneurially forward-looking solutions in strategically relevant fields, such as current approaches to integrating artificial intelligence into work processes while respecting the sustainable and human-centered value canon of family businesses. Further information is available at https://www.entrepreneursclub.eu

Press contact appliedAI

Stephanie Schäuble presse@appliedai.de appliedAI Initiative GmbH August-Everding-Straße 25, 81671 Munich

Press contact Der Entrepreneurs Club

Katharina Bunz presse@entrepreneursclub.eu Der Entrepreneurs Club e.K. Ismaninger Str. 115, 81675 Munich