Why are talents and competencies important in the AI context?

The extent of competence building a company needs for AI deployment depends heavily on its size and goals. Large companies often build AI competence broadly and create their own roles or teams. Smaller companies usually fare better by qualifying specifically for a single AI application case rather than investing right away in comprehensive training programs. For, a technically flawless AI application may still be unproductive if employees cannot reliably assess its results – conversely, a dedicated AI team is typically unnecessary for a finished standard solution.

Building AI competence remains an ongoing process: New application cases bring new requirements, so competence development should be closely linked to specific AI projects and regularly reviewed. AI competence is also enshrined in the European AI Act: AI system providers and operators are expected to take measures to develop their employees' AI competence, oriented to prior knowledge, experience, and specific use.

The topic of talents and competencies becomes particularly meaningful when a company:

• wants to introduce its first AI application,

• wants to integrate AI into existing business processes,

• wants to develop or adapt its own AI solutions,

• is uncertain whether existing team knowledge suffices for the planned AI deployment,

• needs to meet legal requirements such as the AI competence obligation from the AI Act, or

• wants to cushion the shortage of skilled workers by targeted internal AI competence building.

Not all employees need the same training. The sooner you clarify what employees are supposed to do with the respective AI system, the more precisely and economically competence building can be planned.

How should I proceed?

Build the right AI competencies for your company in 5 steps:

• Step 1 – Start from the specific application case

Don't start with job profiles or training offers, but with an AI application case your company aims to implement – such as an AI assistant that answers customer inquiries faster. Note which tasks are necessary for this: knowing typical questions and good answers, providing relevant information, technically checking AI answers, managing the technical solution, clarifying legal requirements. Additionally, check how many inquiries are currently manually processed to assess the project's impact and whether applications already used bring suitable AI functions.

• Step 2 – Set the target and find existing knowledge

Formulate a concrete, measurable goal instead of a general intention – for example: “In six weeks, an employee from customer service independently manages the AI assistant and reliably verifies results.” Do not check by job title, but by task, who already possesses the necessary skills today – in customer service, IT, and data protection. Often it becomes apparent that a significant part of the competence is already present, but distributed among multiple people. Finally, note only the actual gaps; a simple list suffices for the start.

• Step 3 – Qualify only for the specific task

Choose a narrowly defined qualification that can be implemented in a few weeks, instead of setting up a training program for all employees. For example, those using an AI assistant for research and drafting should know what information they can input, what the system is suitable for, what errors may occur, and how to verify results. General AI training provides basic knowledge but is often not enough for specific applications – those using only a finished application do not need training in machine learning. Avoid company-wide large-scale training as an introduction; they rarely meet the actual needs of individual teams.

• Step 4 – Determine responsibilities and solutions for skill gaps

Appoint a constant contact person who accompanies the qualification, plans the timeframe, gathers feedback, and reports progress to management. They also decide how gaps will be closed: if a skill is regularly needed, they first check for internal development with reasonable effort. If specialized knowledge is only needed for integrating an AI solution, temporary external support is often more sensible. Recruiting becomes relevant only when knowledge is permanently needed and cannot be built internally.

• Step 5 – Evaluate with real results after 4 to 8 weeks

Agree on a fixed date for a review 4 to 8 weeks from the start. Go through five typical outcomes of the AI application with involved employees: Were errors recognized? Were unsuitable results adopted? Is it clear when a human needs to intervene? Does the application actually relieve the workload? Document the findings in writing and derive further qualification needs. If skill development does not reasonably relate to the benefit, the application case itself should also be reviewed.

You will find this and many more pieces of information on the AI compass topic "Talents and Skills" in our download portal: