Building AI literacy in the team sounds like a training task. You book a workshop, show a few tools and hand out a collection of good prompts. What it is really about is something else: the ability to assess the results of a machine correctly. Whoever cannot judge whether a text, an image or an analysis is good will not achieve good results even with the best tool.
This shift is often underestimated. AI makes many work steps faster. In doing so, it shifts the weight from production to evaluation. In a marketing department where content is created in large numbers, judgement becomes the scarcest resource. Whoever does not deliberately strengthen this ability in the team produces more and understands less. In the long run, not only the quality of the work suffers, but also the self image of the people who do it.
- AI literacy above all means judgement, not operating knowledge.
- The EU AI Act obliges companies to ensure sufficient knowledge in dealing with AI.
- One off trainings are not enough, literacy develops in everyday work.
- Leaders have to work with AI in practice themselves in order to make good decisions.
- Expertise in brand and craft remains the basis of every good evaluation.
Table of contents
What AI literacy in the team really means
AI literacy has several layers. The top one is operating knowledge: which tools exist, how do you phrase a task, how do you refine a result? This layer is visible and easy to train. But it also becomes outdated quickly, because the tools change constantly. What counts as best practice today may look different again in a few months.
Below that lies an understanding of how these systems work and where their limits are. Whoever knows that a language model produces plausible rather than true sentences handles the results differently. Whoever understands that images emerge from patterns checks more carefully whether details are right. This knowledge protects against naive trust, but also against exaggerated scepticism that rashly dismisses useful possibilities.
The deepest layer is professional judgement. It decides whether a result fits the brand, whether a statement holds up and whether an image shows the right attitude. This ability does not come from AI, but from experience in one’s own craft. It grows slowly, through many successful and failed attempts, and cannot be shortened by any tool. Good AI literacy connects all three layers and weights them correctly: judgement first, operating knowledge last.
How judgement develops in a team I learned at the agency oddity. There, between 2013 and 2017, I built the creative team for dm-drogerie markt from 7 to 22 art directors and was responsible for hiring and talent development. In the process one principle became firm for me: software can be learned in a few weeks, a secure judgement of quality takes years. That is why it pays to develop talent above all in their eye for good work. This weighting applies to AI today in the same way. Tools change, judgement remains the core of the competence.
Legislators see the issue too. The EU Artificial Intelligence Act requires in Article 4 that providers and deployers of AI systems take measures so that their staff have sufficient AI literacy in dealing with these systems. Building this competence is therefore no longer a voluntary extra.
Where building AI literacy fails in practice
The most common mistake is the one off training. A workshop can spark curiosity, but it rarely changes daily work. A few weeks later some people use the tools intensively, others not at all, and the knowledge in the team drifts apart. A common starting point thus turns, in a short time, into very different ways of working that can hardly be compared with each other anymore.
A second mistake is focusing on tricks. Collections of supposedly perfect prompts promise quick results. But they train copying rather than thinking. Whoever does not understand why a phrasing works cannot transfer it to new tasks. As soon as the tool changes, the knowledge gathered with effort is then often worthless.
Third, the example set by leadership is often missing. When leaders talk about AI but do not use it themselves, a strange distance arises. They can then neither assess what is realistic nor ask good questions. I have often seen teams pay very close attention to whether their leadership is learning itself.
And finally, fear is underestimated. Many people wonder what AI means for their own role. Whoever does not address this question openly reaps restraint instead of curiosity. Honesty helps more here than reassurance, even if not every answer is settled yet.
What AI literacy means for a brand
For a brand, AI literacy is a question of quality assurance. Every team member who works with AI becomes a kind of gatekeeper. They decide which results are passed on and which are not. The better this judgement is, the more consistent the brand remains, even as the volume of content grows.
This means that brand knowledge and AI knowledge belong together. Whoever does not know the brand voice cannot judge whether a generated text hits it. Whoever has not internalised the visual rulebook does not recognise when an image deviates from it. The experience with AI image generation in fashion shows this very clearly.
It also means that the criteria for good work must be clear. Whoever is supposed to evaluate creative work in the team needs a common yardstick. This yardstick applies to content made by people just as much as to content created with AI.
Finally, AI literacy creates independence. A team that understands the technology can talk better with agencies, service providers and software vendors and depends less on promises. It can buy more purposefully, plan more realistically and recognise better when an offer promises more than it can deliver.
What I would do differently
I would not treat AI literacy as a project, but as part of normal work. Small, regular formats help more than big events: a short exchange in which someone shows what worked and what did not, or a joint review of results in which the team learns to look closely.
I would also deliberately bring together people with different experience. Younger colleagues are often faster with new tools. Experienced people have the better judgement of quality. When both learn from each other, exactly the mix a team needs emerges. This principle applies anyway when building an in-house agency.
And I would create clear rules without stifling curiosity. Which data may be used, which results need approval, where does responsibility lie? Within this framework the team should be allowed to experiment. AI literacy grows where people feel safe enough to make mistakes and are wise enough to learn from them.
Most important to me would be making evaluation itself the learning material. Instead of only showing how to generate a result, I would regularly talk with the team about why a result is good or bad. What is wrong with this text? Why does this image feel foreign? Such conversations sharpen the eye faster than any manual. They make visible that AI literacy is in the end a form of care, and care can be practised.