An AI strategy for the marketing department rarely begins with a strategy. Usually it begins with tools. Someone tries out a text generator, another person creates images, a third team tests software for translations. After a few months there are many small successes and no common direction. What it is really about is the question of which work should change and why.
The difference is fundamental. Tools come and go, and their pace is high. Whoever aligns their planning with individual programs plans against a moving target. A sustainable AI strategy, by contrast, is aligned with one’s own processes, with the brand and with the people who work with the technology. These things change more slowly and are therefore the better foundation. Whoever starts there can swap tools without having to begin from scratch every time.
- An AI strategy describes changed work, not a list of tools.
- The biggest lever lies in redesigning processes, not in isolated use.
- Responsibility belongs in leadership, not only in IT or in individual teams.
- Brand and quality need clear rules before the volume of content rises.
- Success shows in measurable goals that were set beforehand.
Table of contents
What an AI strategy in marketing really is
At its core, an AI strategy is an answer to three questions: which tasks does the technology take over, which remain with people, and how do both work together? Only when these questions are answered is it worth selecting tools. Before that, every purchase is an experiment without a hypothesis. You do learn something, but you do not know what you actually wanted to find out, and so you cannot say whether it worked.
I find it interesting what the McKinsey study on the state of AI in companies suggests at the level of principles: the clearest difference is not made by the technology itself, but by redesigning workflows and by whether top leadership takes charge of steering. That matches my experience. Where AI is only attached to existing processes, the effect remains small.
In my current role I lead a team of 22 people who look after markets in EMEA. Part of our work consists of preparing decisions better with analytics and AI supported marketing. Again and again it turns out that the bottleneck rarely lies in the tool. It lies in the question of who needs which information when, and who decides in the end. An AI that delivers an analysis in minutes helps little if the process afterwards takes weeks. That is why I start with the decisions themselves and only then ask what support the technology can provide.
For marketing this means: you should not ask where you can use AI, but what a process would look like if you were allowed to design it anew today. Content production, maintaining product texts, evaluating campaigns and preparing decisions are obvious fields for this. In each of these fields it is worth first honestly mapping today’s process before talking about improvements.
A good AI strategy is therefore also an organisational document. It describes roles, responsibilities and approvals. The technology is only one part of it, and often not even the hardest. It is harder to change habits that have settled in over years and that many people feel are right.
Where an AI strategy fails in practice
The most common mistake I have seen is confusing activity with progress. Many pilot projects create a feeling of movement. But if none of them moves into regular operation, the department has in the end only gathered experience and changed no work. At the same time fatigue grows in the team, because every new project again costs attention without everyday work noticeably improving.
A second mistake is missing responsibility. If nobody in leadership really owns the AI strategy, it gets ground down between IT, the legal department and the business units. Everyone has good reasons for caution, and nobody has the mandate to weigh the reasons against each other. The result is standstill with a clear conscience.
Third, many underestimate the data basis. AI in marketing is only as good as the information it works with. Incomplete product data, inconsistent terms and scattered image archives slow down every project. A well thought out martech stack and clean master data are therefore not a side issue, but a prerequisite.
And finally, measurement is often missing. Whoever does not define beforehand how success can be recognised can afterwards present everything as a success. That weakens credibility with management. By the next budget round at the latest, the basis for justifying further investment objectively is then missing.
What an AI strategy means for a brand
For a brand, the most important question is not speed, but quality. AI can increase the volume of content considerably. Without clear rules, however, the volume of content that does not sound or look like the brand also rises. More output is not a value in itself.
That is why every AI strategy should be linked to the brand voice and the visual rulebook. The technology has to learn what the brand says and what it would never say. That is work that must be done before use, not after. Whoever takes their content supply chain seriously builds these rules directly into the processes.
Then there is the question of approval. The more content is created, the more important a clear review process becomes. Not every text needs the same attention. But it must be clear who approves which kind of content and by which criteria. A product text for the online shop requires a different review than a statement on sustainability or a campaign visual.
Finally, AI changes the expectations of the brand itself. When content becomes cheaper, what cannot be produced at will gains value: attitude, consistency and real relationships with people. A good AI strategy protects these things instead of sacrificing them to speed.
What I would do differently
I would start with a few processes, but redesign them completely. Better two processes that really work differently afterwards than twenty experiments running side by side. That creates trust and provides a pattern that can be transferred to other areas.
From the start I would set goals that are understandable for people outside marketing too. Lead times, the quality of product data or the number of correction loops are more tangible than abstract promises of efficiency. Whoever sets up marketing metrics cleanly can later show the contribution of AI honestly.
And I would treat the AI strategy as a living document. The technology develops quickly, and some things that seem sensible today will be outdated in a year. An AI strategy that is reviewed regularly is more valuable than a perfect paper that nobody touches anymore.
Above all, though, I would involve the team earlier. The people who work with texts, images and data every day know best where time is lost and where quality suffers. An AI strategy created over their heads is politely acknowledged and then bypassed. A strategy they helped write, by contrast, is defended, even if the first steps are bumpy.