AI tools in the workplace: three writing aids put to the test

We recently received an inquiry at the agency: "How do you actually use AI in PR?" The quick answer: "With caution." It's foolish to use artificial intelligence just because it's trendy or innovative and you want to be seen as a pioneer. Instead, it makes sense to use AI applications in a targeted manner to make processes more efficient, increase productivity and thus free up capacity. AI can provide valuable support for complex tasks in particular, such as analyzing large volumes of data.

As the creation of texts makes up a large part of our daily agency work, I have tested three AI tools that are designed to help with precisely this: ChatGPT, DeepL Write and Jenni AI. F.A.S. editor Patrick Bernau has compiled an overview of other current tools for journalists here. But before we take a closer look at the potential applications and risks of the tools in question, let's first take a brief look at generative AI in general.

Generative AI refers to AI systems that can independently create texts, images or audio and video products. Based on provided or freely accessible information, these systems learn to recognize patterns and create new content from them. The results are sometimes almost indistinguishable from human-made content. On the one hand, this offers great opportunities, but on the other hand, there are various risks, from copyright infringements and increased cyber risks to deep fakes – lifelike video forgeries.

If I use GenAI tools professionally, I first assess the level of confidentiality of the data and information to be entered. If it is sensitive data, it has no place in a corresponding input field. This is because every request usually trains the tools further – so the data could reappear at any time in another dialog with third parties. In addition, the legal situation regarding copyright has still not been conclusively clarified. This is another reason why it is advisable not to use AI products thoughtlessly and unaltered.

But without further ado, here is my assessment of ChatGPT, DeepL Write and JenniAI.

Anyone who has never heard the name ChatGPT in this country in 2023 probably lives somewhere deep in the woods without electricity, a daily newspaper or an internet connection. I haven't seen a comparable media triumph even with the introduction of the iPhone. But how good is the AI chatbot from OpenAI really?

The first thing that became apparent during the test was that the free version is regularly very busy and offers a very limited quota for requests. Without registration, there are just three per day. The dialog works in different languages, including English and German – although the latter sometimes sounds a little clumsy. If you need accurate information quickly, you should also stay away from ChatGPT. The chatbot regularly spits out misinformation.

My main use for ChatGPT is the creative support. Finding it difficult to start a text? The headline just won't fit on paper? Then ChatGPT's text suggestions may be able to break the mental block. If you want to summarize or rephrase a text, ChatGPT can also help – if the prompt is carefully designed. The tool can also be a valuable resource for concepts if you want to briefly summarize proven approaches. One example:

The prompt: 
Draft five headlines for a blog post about generative AI!

The Result:

1. "The future of creativity: how generative AI is revolutionizing the art world"
2. "Of algorithms and inspiration: how generative AI opens up new creative opportunities"
3. "Artificial intelligence as a creative partner: how generative AI enriches our artistic work"
4. "From Picasso to GAN: a journey through the history of generative AI in art"
5. "The power of chance: how generative AI produces unexpected and fascinating results"

Interesting, but inappropriate for this article. So let's do some fine-tuning:

The refined prompt: Imagine you are a communications consultant writing a blog post on the topic of generative AI. You compare three different tools in terms of their quality and potential to make work easier. Draft five headlines for this post!

The result now:

1. "Generative AI: the future of communication? A comparison of three tools"
2. "Communicating more efficiently with generative AI: a look at three innovative tools"
3. "From text generation to speech analysis: which tool offers the greatest potential for communication professionals?"
4. " Facilitating work through generative AI: An analysis of the best tools on the market"
5. "Communication consulting in the digital age: how generative AI tools can revolutionize the workflow"

We can work with that. What you don't get, however, is anything really new or original. The results are usually a bit too generic and over-positive to use directly. I asked ChatGPT to write the blog post mentioned in the prompt above. That was the introduction:

Would you have kept reading? No, me neither. So let's get straight to the conclusion of ChatGPT's blog post: "With further advances in Generative AI, we will undoubtedly see even more amazing applications and innovations. Stay tuned for the exciting future of Generative AI tools!" Thanks, that's enough.

Vorteile of ChatGPT

• Can provide ideas for texts and create outlines or headlines as a basis
• Can process large amounts of data, for example to summarize long texts
• Can complete simple text-based tasks

Disadvantages of ChatGPT

• Provides neither sources nor a plausibility check for its results. "Hallucinations", i.e. false information, occur regularly.
• Always has an overly positive tenor when writing texts – everything is great and innovative
• Cannot recognize biases rooted in training data
• Cannot ensure the protection of sensitive data

Many people value DeepL highly as a translation tool. In the free version, you can create short texts of solid quality in 31 languages. As a basis, they usually work really well. The glossary also offers the option of specifying your own translations for certain terms, which is particularly useful for technical terms or company-specific expressions.

What is less well known is that the DeepL AI can do more than that. DeepL Write has functions with which you can improve the style, spelling and grammar of texts. For example, you can take the sometimes somewhat rough initial translations and test whether DeepL Write can raise the text to a higher level.

At DeepL, this paragraph sounds like this after the check: "What is less well known is that the DeepL AI can do even more. DeepL Write has functions that can be used to improve the style, spelling and grammar of texts. So you can, for example, take the first, sometimes somewhat wooden translations and test whether DeepL Write can raise the text to a higher level."

Not bad, is it? DeepL Write is available for German and English texts, among others, and even the free version has no major usage restrictions. The tool highlights errors in the text and suggests synonyms on request. The highlight: you can choose whether the text you want should be of a simple, business, academic or technical level. It's definitely worth playing with the tool a little, for example to formulate emails or write an outline. This way you can give a text a polish or at least get some creative impulses. Of course, DeepL Write can't make a silk purse out of a sow's ear either.

Advantages of DeepL Write

• Can check texts in various languages for style, grammar and spelling
• Is intuitive to use
• Offers the option of checking texts at different levels

Disadvantages of DeepL Write

• Is a specialized tool and therefore only offers limited functionality

Unlike ChatGPT and DeepL, Jenni AI often flies under the radar. And quite wrongly so. The tool itself is free, but requires registration, e.g. via a Google account. The user interface can be a little overwhelming at first, as it is more complex than the previously mentioned applications. However, once you have found your way around, Jenni AI really comes into its own. And it's so good that Patrick Bernau describes the tool as "scary" in his list above.

Based on a short prompt, Jenni AI can, for example, create a structure for a text, check an existing one and provide it with counter-suggestions - or write the entire text. You want to check the language of a text, paraphrase, simplify, shorten or add technical depth? No problem!

German and English work equally well here. A translation function is also built in - but the results read somewhat unnaturally. An AI auto-complete function helps if you get stuck on a sentence. If you use quotations in your text, you can automatically mark them according to various scientific standards. Even PDFs can be used as direct input for citations. Once you understand how Jenni AI works, the tool mutates into a real efficiency monster.

To make a direct comparison: Jenni suggested this introduction for this blog post: "As communication professionals, we are constantly on the lookout for new tools and technologies that can help us make our daily work more efficient and successful. One emerging technology that is gaining momentum in this area is Artificial Intelligence. Artificial intelligence enables us to automate processes, create predictive models, optimize logistics processes and support decision-making."

With minimal changes, I could have used that without batting an eyelid.

Advantages of Jenni AI

• Can create outlines, structures and texts at a high level using short prompts in various languages
• Offers the possibility to create texts in different styles
• Provides the ability to include quotations and references

Disadvantages of Jenni AI

• Is quite complex in its possibilities – therefore requires a familiarization and learning phase
• The translations are sometimes bumpy – so it makes sense to work within one language

Generative AI can do much more than create funny pictures and bad jokes. For example, the three applications tested can certainly make everyday work easier for communicators and PR professionals. They help to complete tasks faster and more efficiently – especially when your mind is blank. Important to note: technology is not a cure-all and you should never use the results of machine-generated texts without checking them. Copyright law is still unclear, so any text that you use unaltered moves you into a legal gray area.

My conclusion: when it comes to text and language, the need for human professionals is not going to disappear any time soon. AI tools can sometimes provide good support, but for high-quality texts I rely on my sense of language and the expertise of my colleagues.