Skip to content

Agency people are usually more pirates than navy officers

For Sale is an owner-managed communications and digital agency with offices in Munich and Hamburg. With over 50 employees, they design in all disciplines of communication, online and offline: Strategy, creation, production, programming, online marketing and media. For Sale has been in business for over 25 years. Their clients come from the B2B, B2C and retail sectors. These include some of the biggest brands in their sectors, such as the furniture stores XXXLutz, MediaMarkt, the Liebherr Group, Volvo Trucks, Wacker Chemie, ANWR with Quick Schuh, the Müller drugstores, the Goethe Institute and hagebaumarkt. We spoke to owner and Managing Director Christian Rechmann about the sensible use of artificial intelligence in his agency, the hurdles and opportunities.

Image Alt Text

Christian Rechmann, Managing Director and Owner, For Sale, Germany. Interpreted by runway.ai

How and in which areas do you use artificial intelligence (AI) at For Sale?

In text, correspondence, image design and animation, and our colleagues in Hamburg get help with coding.

Which tools do you use exactly?

ChatGTP, of course, previously neuroflash and currently also Copilot. For image generation, we prefer midjourney, runway and Dall-E and of course the AI offerings from well-known providers such as Adobe and Microsoft. In programming, we are currently testing the JetBrainsAI-Assistant.

Does AI help you mainly with writing or even more with image processing?

Depending on who you ask:

In consulting or text, Chat is used a lot to pre-structure articles or for quick research on a topic; for automatic transcripts or summarizing texts.

In graphics, it is used for image and video editing tools.

And software developers mainly use the option of documenting our code and writing unit tests for it.

Is the use of AI still rather "experimental" or is it already bringing measurable added value to increase efficiency in the company?

A bit of both. We are trying to at least keep an eye on and test the many AIs that are currently flooding into the market. This is certainly still rather experimental.

But topics such as an automatic transcript of a meeting, the summary of a text or a function such as generative-fill in Photoshop are already indispensable for agencies.

Can personnel costs be saved through the use of AI?

Yes and no. In the medium term, certain tasks will no longer exist in our environment.

We already do some illustrations ourselves with the help of AI rather than outsourcing them to an illustrator. That will increase. And AI will probably soon be used to support the molding of approved artwork, for example for use on various social platforms. It remains to be seen whether this will save on personnel costs or rather close the gap in staff shortages.

What hurdles and obstacles are there to the use of AI?

First and foremost, I see legal problems on three levels:

  1. how is the AI trained, i.e. what exactly is the source material and who owns this.

  2. the legal issues that follow on from this, i.e. who owns the generated image and to what extent can we pass it on to our customers?

  3. and the question of protecting our own data. Customers often entrust us with information that is not yet common knowledge in the market. Of course, we can't simply upload this to an AI cloud - even if we were to anonymize it beforehand, we would still be making the knowledge available to a general audience of AI users.

As far as text is concerned, it seems to me that the drop has largely been sucked, because here nobody can trace where which text comes from and generic texts can hardly be protected anyway. But for the generation of images, for example ... well, if only someone would answer that properly! There are a few landmark lawsuits underway both here and in the USA, which will hopefully give us legal certainty soon.

Until then, we are trying to deal with the uncertainty in a very transparent way, i.e. to clearly identify where something comes from and, if necessary, to decide together with our customers to what extent we want to use it.

And some providers, such as Adobe or shutterstock, are currently taking risks for us by promising to generate only rights-cleared material and even assume liability for it.

Are customers worried that internal knowledge will leak out?

There certainly are, and rightly so in my opinion. We have added the restriction "do not upload confidential information to freely accessible AIs" to the topic of NDAs and confidentiality.

Are there any reservations about the use of AI in your organization?

I don't see any general reservations in our organization; agency people are usually more like pirates than marine officers, they actually like to try everything out first. We take the concerns arising from your question about protecting our own creations and how individual tasks will change very seriously.

How do you think work in agencies will develop with AI? What do agency employees need to prepare for?

This crystal ball, if only I had it ... I have two opposing scenarios in mind here: on the one hand, we already see many areas in an agency on which the development will have a strong influence and, on the other hand, we have already lived through many changes: from the end of manual typesetting (on slides by Letraset letters) to the processes of digitization or the introduction of penny-stock picture agencies - the end of our industry was always feared and there were always new tasks for us agencies afterwards.

In essence, it will continue to be the creative process that makes an agency necessary. And for me, creativity in this context is more than just an idea, it's primarily about concepts and selection skills.

If you could wish for something in regard to AI? What should it do better for your company?

Clarified rights would be very important and at the moment we also have to make sure that the costs don't get out of hand. Microsoft and Adobe are already paying handsomely for the new functions and then there's an app here, an extension there and another license on top. An AI flat rate from a few providers or pay-per-use models would be desirable.