Product Descriptions at the Touch of a Button
Maximum Efficiency in Content Production: How AI-Generated Text Is a Game Changer
A part of the ROTHENBERGER Group, ROTHENBERGER Werkzeuge GmbH is one of the world’s leading manufacturers of pipe tools and pipe processing equipment. Quality is a core value that extends to both products and processes, and ROTHENBERGER considers its focus on data a crucial competitive advantage. The Group’s high data quality, especially in the PIM system, has enabled ROTHENBERGER to optimize the creation of product descriptions: in collaboration with Parsionate the company introduced an AI-based content generation solution, automated processes and saved both costs and resources.
Bastian Seib, Head of Product Data Management and Marketing Technology, and Adrian Flock, Content Manager and Product Data Manager at
ROTHENBERGER Werkzeuge GmbH, explain how AI-based text automation at ROTHENBERGER works and how the company and employees benefit.
Mr Seib, can you start by explaining what text attributes we are talking about?
Seib: There are three relevant text attributes: marketing copy, use cases and product descriptions. High-quality content is extremely important for
ROTHENBERGER. Marketing copy drives sales and needs to explain our products in detail. Use cases are crucial because they tell customers what the product is suitable for. Lastly, specific product descriptions improve customers’ understanding and build their trust in the brand.
How was this copy created before?
Seib: Most texts were created manually. For instance, any product-specific texts, including maintenance of the technical data for the respective products, were the responsibility of our in-house product management team. We set up dedicated workflows in our Contentserv PIM system to handle these tasks. The main issue, however, was that this approach led to inconsistent terminology. So text had to be edited and rewritten manually, which was a very time-consuming process.
How did you come up with the idea to look into automated text creation?
Flock: We wanted to simplify our processes for quite a while, mainly due to the challenges mentioned and the time needed to write all these texts ourselves. ChatGPT suddenly opened up exciting opportunities that inspired us to break new ground. Together with Parsionate, our long-standing consulting partner, we managed to test and implement AI in this area quickly and efficiently.
What was the project basis?
Flock: Firstly, of course, our high data quality and secondly, our Content Definition Guidelines, which had been in place since early last year. Initially, these were intended as a support tool for product managers and incorporated in our PIM system. We defined each field according to certain specifications, for example the number of characters. We had already applied these rules to articles we entered and started to create a database. However, it all depended on people, and that remained the challenging factor.
In what ways were people the challenging factor?
Flock: Even if there are clear and uniform rules, it is human nature to interpret these differently, which happened at our company as well. These different interpretations of Content Definition Guidelines led to copy being still somewhat inconsistent, especially in external communications.
Seib: Basically, we had two options: to proofread every single text every single time, which would have been very time-consuming, or to take a new approach and go with AI-generated texts based on our high-quality, structured data. With the latter option, we only had to do a final quality check and make sure the technical data, as maintained, is correct. The decision was a no-brainer for the team, as long as we could check the technical
data ourselves before copy goes out.
„At this point, it no longer matters whether we have five new product descriptions or fifty.“
Did you have specific KPIs at the start of the project?
Seib: ROTHENBERGER is very process-driven and KPI-oriented. That’s why we discussed in advance what we would have to achieve to consider the project a success. With this completely new field, we initially found it hard to define specific KPIs. But we had the key KPIs in place, including how many products had to be revised. Also, the text creation process, progress and time savings were analyzed, which allowed us to pull a multitude of small levers and continue to optimize the entire process.
What were your priorities in terms of prompt engineering?
Flock: To ensure effective and efficient prompt engineering, I devised the Golden Sample and provided almost 400 texts created according to our guidelines. These served as the initial database. We wanted to include examples for almost all products to minimize the risk of new articles unknown to the AI. This explains the large number of reference texts. Then we revised the prompt, and it only took us a few adjustments to come
up with texts that met our expectations.
In which languages does your system create copy?
Seib: We use our PIM’s integrated translation management system. First, we create all texts in German. Following approval, they are machine-translated into English and automatically forwarded to our international teams for release. AI content generation was easy to integrate into this workflow.
What is your verdict? Has the use of AI paid off?
Seib: Definitely! Automation has brought two major benefits: first of all, the new process helps us to clearly track any causes of errors and better assess where we stand. If we are informed of an incorrect product description, we know with certainty now that it has not yet run through our Content Definition Guidelines. We can give feedback that the text is being revised and will soon be at the desired quality standard.
The second big benefit is probably the time savings, isn’t it?
Seib: That’s right. Several departments benefit from the time savings. For example, product managers can now focus on their core tasks. The same is true when new products are introduced and new content needs to be created. At this point, creating five or fifty new product descriptions is about the same in terms of time and resources.
How does the quality of AI content compare?
Seib: The texts meet our desired standard and we are very satisfied. Despite the high quality of the automated copy, it still undergoes a final check where a person proofreads the text. Other than that, new content is generated via simple selection fields. The text attribute can be entered at the beginning of the workflow, and the copy is created at the touch of a button, reducing the amount of work we now need to create a text to a minimum.
Was it also worth it in terms of cost?
Seib: Absolutely. People who previously dedicated a lot of time to content creation now have capacity for other tasks, which increases productivity overall. Plus, the costs will amortize quickly, as text automation is now used across all brands of the ROTHENBERGER Group. All the important settings in the workflow are adapted for the respective brand, which means it doesn’t require any new implementation.
Do your B2B customers, the retailers, also benefit from automated content?
Flock: Our customers benefit from the fact that we now provide high-quality and, most importantly, consistent information. Should the copy include any errors, we can respond very quickly and correct them. What’s more, we can also respond to customer inquiries much faster. If a customer wants to list a new product not yet maintained in our system, we can now provide this information much faster, which of course benefits both the retailer and the manufacturer.
How was the project received? Did your employees have a lot of questions?
Seib: No, not at all. Our entire team recognized the benefits immediately and is very happy with the solution. It has brought nothing but benefits for the individual team members: responsibilities have shifted, and a burden has been lifted from areas that previously bore the brunt. The result: increased efficiency and a better team-spirit. The other brands in the ROTHENBERGER Group were also on board immediately.
Do you have any tips for other companies facing similar challenges?
Flock: Our process was very structured. Our Content Definition Guidelines allowed us to develop examples and prepare the data. Results were useable after only a few adjustments and learning processes. I would also recommend this kind of logic to other companies as well. Having a clear roadmap and a set of rules right from the start is key to make sure procedures and requirements are clearly defined. This is important for the AI tool. Everything that is to be processed must be subject to rules, otherwise the result will not be up to par. Another important aspect is structured data. Without it, implementation takes much longer and requires more resources, as everything has to be revised and adapted.
Seib: Another important point is also to have a project owner on the team who drives its progress. This also includes a management team that supports the topic and trusts you. Successful implementation fully depends on the management having an open mind.
What are your next plans for AI and automation?
Seib: Content creation was the first important step, and a very positive experience. We want to take a holistic approach to AI and data management, especially in digital marketing. We held a cross-departmental workshop to identify further topics where we see great potential in using AI and automation. The next step is the integration of analytics. We intend to analyze the published texts for their effectiveness, target group approach and conversion rates and then apply the information gathered in the content generation process. Our goal is to make better and more intentional use of this feedback.
