AI in Marketing: Why Content Creation Is No Longer the Problem

The Bitkom study from March 2026 contains a figure we often cite. 41 percent of German companies with 20 or more employees use AI. A year ago, the figure was 17 percent. That’s a doubling in twelve months.

A second figure from the same study is more important to us. 68 percent of AI-using companies deploy the tools for text creation. Emails, documents, marketing copy. This is by far the largest use case.

And this is exactly where we see a flaw in thinking in our work. Many CEOs we speak with still believe that AI-powered content creation is their competitive advantage. That hasn’t been true for about a year now. In this post, we explain why, and what we believe the next step is.

Content creation has become a commodity

In the spring of 2024, ChatGPT was still an experiment. Anyone who started creating product descriptions or blog articles with AI back then had a real head start. The market wasn’t familiar with the tools yet.

In 2026, the situation looks different. A McKinsey study from March 2026 shows that 38 percent of employees in Germany use AI tools regularly. A year ago, the figure was 19 percent (Wirtschaftswoche). One in six Germans uses AI daily at work.

That means: If you have ChatGPT write a product description today, your competitor is doing the same. Both get similar results because both use similar prompts. Text creation as mere drudgery is no longer a differentiator.

We’re seeing this with our clients. Those who proudly announced in 2024, “We’re now having AI write our product descriptions,” are realizing in 2026 that their competitors are doing the same. Visibility in search engines doesn’t improve automatically. The tone comes across as interchangeable.

What the numbers reveal beneath the surface

The Bitkom data has a flip side that is rarely cited. 53 percent of companies cite a lack of expertise as an obstacle to AI adoption. 53 percent struggle with legal uncertainty. 39 percent fail due to integration into existing processes (Onelake).

The last figure is the most telling. “Integration into existing processes” means, in plain terms: Employees use ChatGPT individually at their workstations, but it doesn’t change the workflow. The tool runs in parallel, not within the system.

Here’s an example from our own experience. An industrial company in the region asked us for help a few months ago. They had introduced ChatGPT Pro; three employees in marketing were thrilled, and it generated product descriptions, social media posts, and trade show flyers. After six months, the reality set in. The employees reported a time savings of about two hours per week. This aligns with the Chip study, which cites 1.7 hours per week (Chip.de). That’s nice, but not transformative.

The time savings weren’t the problem. The problem was that no one knew what the freed-up time was for. There was no pipeline of new ideas, no structured content planning, no clear goal-setting process. The AI sped up the old work. It didn’t make new work possible.

What we believe the next step is

We don’t believe that AI tools are the end of the story. They are just the beginning. The next two to three years won’t be decided by the question “do we use AI?” but by the question “how do we restructure our processes so that AI really makes a difference?”

Three points we see time and again in client projects:

1. The content inventory is missing

Before AI can create meaningful text, it must be clear what content exists. Most mid-sized companies we work with have their texts scattered all over the place. Product descriptions in the ERP, marketing texts on the network drive, trade show brochures in PDF format with the graphic designer, website texts in the CMS, customer emails in Outlook.

If AI is to strike a consistent tone for this company, it needs to know these sources. Without an inventory, you’ll get generic results. The first step toward truly using AI is simple: Where are our texts, and who is responsible for them?

2. Approval processes are still manual

In many companies, text approval works like this: An employee writes, AI may assist, the text is emailed to marketing management, who corrects it, sends it back, and the employee edits and uploads it. This process takes more time than the actual writing.

If text creation now takes minutes instead of hours, the approval process becomes the bottleneck. We see in projects that companies save two hours a week on writing but lose three hours on coordination. The net result is an extra hour of work.

3. Measurability is often forgotten

What does the text achieve? Almost no one asks this question. We often ask managing directors in initial meetings: Show us which texts on your website have led to inquiries. In nine out of ten cases, this cannot be evaluated. There is no tracking structure, no link between content and the contact form, and no A/B testing.

If you don’t know which texts are effective, AI-assisted writing is just a game of chance with slightly less effort. That’s not a competitive advantage.

What we recommend as a first concrete step

We usually tell clients the same sequence. It’s unspectacular, but it works.

First: Create a content inventory. A simple table. One row per text fragment that your company communicates externally. Product descriptions, blog articles, social media posts, email templates, trade show flyers. Columns: Title, Location, Last Updated, Person in Charge. This can be done in a week and immediately reveals where the gaps are.

Second: Select three core text types. Not ten. Three. For a mechanical engineering firm, for example, these would be product descriptions, blog articles on use cases, and proposal emails. For these three core text types, establish a clear process: who writes, who approves, who evaluates.

Third: Only use AI in these three processes. Not before. Once the process is in place, AI becomes an accelerator. If the process is missing, AI becomes a chaos amplifier.

That might not sound like much. But after three months, most of our clients save more time this way than they would with five ChatGPT licenses. Because the friction in the process disappears—not just in content creation.

What this means for your business

If you’re thinking about using AI in marketing in 2026, you’re not late to the game. You’re in the middle of the pack. 41 percent are ahead of you, and 48 percent are right where you are—in the planning phase. The question isn’t whether you’ll start, but how.

Our advice: Don’t start with the tool. Start with the process. The tool isn’t the problem. The tool is solved. The problem is that old structures can’t handle faster content creation without creating other bottlenecks.

If you’d like to discuss what your content process looks like today and where AI can make a real difference for you, reach out via /contact/. We’ll work with you in a conversation to identify your bottlenecks and determine what a pragmatic first step might be.

Frequently Asked Questions

As a mid-sized company, do we have to start using AI in marketing right away?

No. But you do need to make a decision. Either you establish a strategy for how to use AI, or you let your employees use it privately without any guidelines. The second approach is already happening in many companies anyway. The Chip study shows that 59 percent of German employees use AI in their personal lives, but only 46 percent at work. That gap is the risk.

Which AI tools are suitable for B2B SMEs?

ChatGPT Plus, Claude Pro, or Microsoft Copilot are sufficient to get started. All three cost around 20 euros per month per user. That’s all you need for the first six months. The more expensive enterprise solutions are only worth it once your process is established and you’re rolling out dozens of users. Before that, you’re paying for features that nobody uses.

How much time does AI really save in marketing?

Current studies cite 1.5 to 2 hours per week per user. That’s about 90 hours a year. In our experience, we often see less, because the time saved is lost again during the approval process. After a clean process overhaul, we see actual time savings of three to five hours per week. The difference comes from the process, not the tool.

Should we label texts if they were created with AI?

Currently, there is no legal requirement for this in Germany. The EU AI Act requires transparency for generated media, primarily images and videos. For standard marketing texts, labeling is not required. We nevertheless recommend documenting internally which texts were created with AI support. This helps with future audits and quality control.

Sources

1. Bitkom: Digitalization of the Economy, Artificial Intelligence 2026 (March 11, 2026)
2. Onelake: Bitkom Study 2026, AI Use Doubles (April 14, 2026)
3. McKinsey Study in Wirtschaftswoche: AI in the Workplace (March 18, 2026)
4. Chip.de: German Employees Use AI More in Their Personal Lives Than at Work (March 26, 2026)