Kennisbank

AI and the marketing department: what changes in the work?

Marketing is one of the departments where AI tools are becoming visible fastest. Writing text, planning, and processing data are tasks that language models and automation tools have been trained on for a longer time. Yet that doesn't mean the entire department is up for grabs. In our taxonomy, 54 tasks fall within the marketing domain, and these divide clearly into three categories: tasks AI can take over, tasks that must remain under human oversight, and tasks that remain human work. This article shows the pattern, not the full list.

Where AI takes over the work

The tasks that are most fully suited for AI to take over have a few things in common: a clear input, a recognizable format, and little dependence on current context outside the text itself. Writing a social media post based on a topic, platform, and tone of voice is a good example of this. The pattern is familiar, the output is short, and deviations can be corrected quickly. Scheduling and publishing social media posts often falls into this category too: once content and planning are set, executing it is a matter of connecting systems to each other.

Writing ad copy follows the same logic. The structure of an advertisement on a search engine or social platform is tight, the length is limited, and there are large numbers of examples on which a model can be fine-tuned. That makes this type of text work suitable for full takeover, provided someone assesses the final message and brand consistency before the text goes live.

What connects these tasks is that quality can be tested well against a format. There is little room for difference in interpretation about what constitutes a good social post or ad copy within the set parameters.

Where oversight remains necessary

The largest group of tasks in marketing falls into the middle category: AI does the preparatory work, a human approves or rejects it, for a reason. Writing a blog article is a telling example of this. AI can produce a draft based on topic, target audience, and tone of voice, but factual accuracy, nuance, and brand voice require assessment by someone who knows the context and the reader. An error or the wrong tone in an article that is publicly visible carries a different weight than an error in an internal memo.

The same applies to drafting a campaign brief. AI can create an initial setup of goal, target audience, and message based on previous briefs and available data, but the budget, strategic choices, and coordination with other departments require a human who bears responsibility for the outcome. Responding to reactions and messages on social media often falls into this category as well: AI can formulate a draft response, but in the case of complaints, sensitive topics, or unusual situations, a human must judge whether that response is actually appropriate before it is sent.

The reason these tasks retain oversight is not that AI can't handle the language. It's that the consequences of incorrect output lie outside the text itself: reputational damage, a misallocated budget, a dissatisfied customer who doesn't feel heard. Once the outcome of a task affects something beyond the document, an approval step is needed for good reason.

Where human work remains

A smaller but recognizable group of tasks remains human work, even in the longer term. Executing and monitoring a campaign along the way is an example of this. Adjusting course during an ongoing campaign requires combining figures with market feel, timing, and sometimes political considerations within the organization — factors that are not all captured in data and that weigh differently depending on the situation.

Segmenting an email list based on behavior, interest, and customer characteristics may look like a data task at first glance, but the choice of which segmentation is strategically valuable for the coming period is tied to commercial priorities that lie outside the system. AI can organize the data; the judgment of which groups are relevant for which purpose remains with a human.

Drafting a content calendar for the coming period also remains, at its core, human work. It's not just about which content appears when, but about coordination with campaigns, seasons, and sometimes current events that are difficult to predict. This is precisely why the same work turns out differently for one organization than for another: the classification into categories is a starting point, not a fixed given for every organization.

What this means for your situation

This classification is generic: based on the nature of the task, not on the specific situation in your team. How many hours this actually frees up depends on how your marketing department currently works, which tools are already in use, and how much of the work is already standardized. We therefore prefer not to name a percentage without explanation; read why we calculate in hours and not in people on this page about our calculation method, and what a bandwidth does and does not say on this page about bandwidths.

If this topic moves toward personnel decisions, separate legal requirements apply that this page does not address.

If you want to know how this plays out for your own role or team, you can fill in the free quickscan from ftetoai: twelve questions, no account needed, with an indication of what portion of the hours in your profile can be taken over by AI today. The full work scan, which goes deeper into your specific tasks, is still under construction — we'd rather report that honestly than offer it before it's ready.