KI-Skills-Gap-Matrix: die Lücke zwischen Können und Brauchen sichtbar machen

An AI skills gap matrix is an instrument that, for defined roles, sets the existing AI competence (current) against the needed competence (target) and makes the difference visible. It answers three questions, which role needs which AI skill, how well it stands today, and where the largest gap sits. The result steers targeted measures.

From gap to measure. In just one week your team closes the largest gap in practice instead of only naming it. Book a discovery call, or enter your email below.

Skills gap analysis is a classic of workforce development. For decades it has compared existing with needed skills to plan training. The matrix format, roles in rows and skills in columns, made this analysis clear and comparable. AI gave the instrument new relevance, for two reasons. First, AI changes almost every role, so the target state must be redefined. Second, Article 4 of the EU AI Act requires role-specific competence, and a matrix is an obvious way to document that link. The shift is from a general HR exercise to a tool that supports planning and evidence at once. That dual role is what makes the AI skills gap matrix valuable today.

The matrix works through a simple per-cell comparison. For each role and each relevant AI skill you enter a current value and a target value, usually on a scale of 0 to 3. The difference is the gap, and the sum of gaps shows where measures have the greatest leverage.

                Prompting   Data handling   Checking   Automation
  Marketing       2/3          1/2            1/2          0/1
  Sales           1/2          1/2            1/3          0/1
  HR              1/2          0/2            1/3          0/2
  Finance         1/2          2/3            1/3          1/3
   (current / target)  gap = target minus current, larger = higher priority

The value of the matrix is not in filling it in but in the prioritisation afterwards. A cell with current 0 and target 2 in a critical role is more urgent than a small gap in a non-critical one. A diffuse feeling (we must do something with AI) becomes an ordered list (these three cells first). Crucially, the target value is derived from role and risk, not from a wish to reach the top score everywhere.

An illustrative model, not a client figure. An agency fills the matrix for four roles and four skills, scale 0 to 3.

The largest single gap is not spread wide but concentrated in one skill. Instead of booking a general AI course for all, a targeted build sprint on automation pays off, because 6 of the 14 points sit there. The numbers are chosen for structure. The point is that the matrix focuses the measure and avoids scattering effort.

The value rises with the number of roles and the density of regulation. In marketing agencies, our first target market, the matrix helps direct scarce training time to the skills with the greatest effect. In finance, insurance and healthcare it also serves as documentation, because role-specific competence must be demonstrable. In industry and engineering it makes visible where automation knowledge is missing. In IT and SaaS it orders the many possible AI skills by need. The common thread is that the matrix prepares a decision, it is diagnosis, not therapy.

In short, the matrix is the diagnostic instrument. It says where the shortfall is, then hands off to a measure that closes the gap.

It helps when several roles are involved, when training time is scarce and must be prioritised, and when proof of role-specific competence is needed. It is redundant when only a small group has an obvious gap, because then the effort of the matrix exceeds its value. It harms when it becomes an end in itself and filling it in replaces the measure. A matrix without a following measure is a nice table with no effect.

Article 4 requires role and context specific AI competence. A matrix that derives the target state per role with reasons and documents the current state can structure this evidence and justify the choice of measures. It is not an official certificate and does not guarantee automatic conformity. Whether the derived requirements are adequate is for the company to assess, with qualified counsel where needed.

Article 4 scope note: this page describes a diagnostic instrument and how it can support competence planning and documentation. It is not legal advice and not proof of automatic conformity.

AI upskilling · AI literacy · AI enablement · AI adoption · AI certification

Two ways, depending on where you are.