AI in marketing has arrived in many companies, but rarely in everyday work. There are pilot projects, test accounts and enthusiastic individuals. What is missing is the step in which AI is no longer a project but part of how a team works. I call it the step from experiment to operating system.
This step is less technical than it sounds. Above all it is a leadership task, and it demands decisions that a pilot project lets you postpone comfortably.
- AI in marketing rarely fails on technology, but on the move from pilot to daily work.
- If you want impact, redesign workflows instead of attaching tools to old processes.
- Every use needs an accountable person, quality rules and a measurement.
- Human judgement stays with brand, visual language and tone.
Table of contents
Why AI in marketing gets stuck in the pilot stage
Pilots are comfortable. They cost little, show quick results and commit you to nothing. That is exactly their problem. A pilot proves that something is possible in principle. It does not prove that it holds up in daily work, when deadlines press, approvals are pending and many people work at the same time.
The current McKinsey survey shows this pattern clearly: almost all companies use AI in at least one function, but far fewer have embedded it across the whole organisation. The difference is made by companies that have redesigned their workflows around AI (see here: McKinsey, The state of AI).
From tool to workflow
The most common mistake is to attach a new tool to an old process. The AI then writes a draft faster, but the alignment takes as long as before. Little is gained.
It makes more sense to look at a workflow as a whole: from briefing through creation to approval and measurement. Where do waiting times occur? Where is the same information prepared three times? Where does nobody check whether the result fits the brand? That is where AI comes in, and that is where its value is created. I have described how to set this order in AI strategy for the marketing department.
What an operating system for AI in marketing looks like
An operating system has rules that apply to everyone. Applied to AI in marketing, this means: every use has a person who is accountable. There are quality rules, for example on tone, image rights and sources. And there is a measurement that shows whether the use saves time or improves results.
It also means being honest about where AI does not belong. In brand work, for me this mainly concerns the judgement on visual language, tone and attitude. Machines can deliver variations, but the decision about what a brand expresses stays with the team.
What this means for the team
When AI becomes everyday work, roles change. Less time goes into making, more into selecting, checking and justifying. That calls for a different skill: judgement instead of routine. I have seen teams experience this as relief and others see it as a devaluation of their work. Both are understandable, and both need to be discussed openly. More on this in AI literacy in the team.
What I would do differently
I would start fewer pilots and take each one more seriously. Better two workflows that are truly changed than ten tests running on the side. And I would define from the start how we decide after three months whether a use stays.
I am unsure how quickly the tools will keep changing. What is a good workflow today may be outdated in a year. That is exactly why I consider clear accountability and simple rules more important than choosing the one right tool.