Think first, then scale

Write faster and publish more - anyone can do this today with AI. Do you want it to be not only fast, but also good? Then talk to me.

Only a lot does not help much

With AI, every single influencer can now publish more faster than your team of professionals could a year ago. However, the results are often questionable.
But it can be done better - much better.
If we lay the right foundations. Together, we can ensure that your AI adheres to the exact style you want to write in for your brand. And produce content that is exciting and factually correct - and that makes you stand out from the crowd.

Three tools in three steps

3 steps to the highest quality:

  • C-01/ Consulting
  • C-02/ Creation
  • C-03/ Collaboration

3 tools for speed and scale:

  • P-01/ Prototypes
  • P-02/ Prompts
  • P-03/ Processes


Strategy that pays off

Let's take a thorough look at what you've been doing so far. In a structured audit, we'll find out what you can tell a credible story about. We define topics, tonality and style that best represent your brand. And build a strategy from there - for authentic content that you can scale day by day.


The foundation for scaling

With me you can create what you are still missing. A name, a claim or a story for your brand? Individual creatives such as social media posts or ads? With the right P-01/ Prototypes In your collection, you set the benchmark for quality and can ensure that you scale sensibly. No matter who does it in the end - people or machines.



Human-in-the-loop. You often hear this demand. And rightly so: humans have to make decisions because AI alone cannot be trusted. Today, however, it is also true that humans alone are not fast enough. So bring on the machines-in-the-loop. With me you get: The P-02/ Promptsthat make the most of what you can expect from AI. And which P-03/ Processes - how people and machines work together.

If we can imagine it, we can do it

I have been working with artificial intelligence since 2018. What I see time and again is that humans and machines together can do what each of them cannot do alone. Here are five areas of application where this is most beneficial:

  • Names for companies and products. Two worlds, two word lists, and Markov chains at work: within hours, thousands of melodious names are created from this combination. A quick check with the registrar reveals that 2,500 of them are available as CH domains. We are spoiled for choice when it comes to the final selection - and in the end it's always the people who decide.
  • Claims & Headlines. Write a brief, give it to your creatives and wait for them to come back with a result. It's quicker with GPT & Co. But you can't do it without patience. Because language models think stochastically. The first ten headlines will be correct - but boring. If you play the Numbers Game long enough, you're sure to find some real gems among the 1,000 suggestions.
  • One content, all channels. Your blog becomes a podcast. Your podcast becomes 26 tweets. Generative AI may be boring to generate, but it's great at transforming. Use it and you can focus on the essentials: Tell what you have to say with heart and soul in just one single channel. GPT & Co can fill all the other channels for you.
  • FAQ for the Helpcenter. For years, your customers have been hearing in the call center: "This call may be recorded for training purposes." Use these recordings and train your machines. Let AI transcribe and classify. Let it suggest which FAQ you should write - and write it right away in the next step. (This doesn't just work with calls, but also with your emails, of course).
  • Unique texts for interchangeable products. You have 50 black T-shirts in your range - and need a text for each one that Google finds unique? And in 13 languages to boot? Some things no copywriter wants to do, no matter how well you pay them. Template-based natural language generation can do this with ease. Provided a human writes the right prototypes first.

The right AI for the right purpose

Of the five examples above, only three are based on large language models, and even then only partially. Language models from the field of machine learning are currently in high demand. They are indeed an excellent choice for claims & headlines and for transforming content. But which model solves your task best - GPT-4, Bard, Llama or Claude?

However, language models alone are not enough. First of all, your content has to enter the machine as speech. This is where Whisper & Co help with transcription. Then the AI has to remember all the data. This is where the decision has to be made: Are you primarily interested in extensive content? Then rely on embeddings and perhaps a vector database. Do you need to be able to quote verbatim? Then choose a model with a long context window.

When it comes to naming and product texts, old-school algorithms still solve some tasks better. When you work with me, you don't have to worry about that. Because together we will not only find the ideal division of labor between man and machine, but also the right software for your task - beyond the hype du jour.

Where do you benefit most from artificial intelligence?

I'm happy to take 15 minutes and talk to you about ways you can start implementing next week.