What were the key challenges when using AI?
First of all, please note that many AI issues have not yet been legally clarified; all these are personal assessments. The authors of this text are not lawyers. This is no legal advice. All information comes without a guarantee.
- Copyright and Rights of Use
In German law, copyrights are inalienable and eternally bound to the author. Only the Rights of Use may be transferred. An „author“ can only be a human being. The product of a machine or algorithms is not protected by copyright. Therefore, everybody can use AI output without taking care of a possible rights violation. For you, this may not seem relevant in practice. But: In every standard contract framework, the agency transfers the „exclusive right of use“ to his client. From a legal perspective, you are not allowed to pass on rights that you don’t own. So, this is a serious legal risk. There are two solutions for this challenge: a) Do not transfer exclusive rights in your contracts. Or b) let humans rework and adapt the output from the AI, until it becomes a separate creative achievement of your agency.
- NDAs
Many standard NDAs we were asked to sign during our PR career prohibited to pass on confidential data under all circumstances. Even within our traditional working process we could hardly keep to that – the definition of “confidential information” even included the facts we should promote to the public. But even the more sensibly formulated NDAs represent a major hurdle, as servers outside the EU are not considered secure in the European legal area. As a result, uploading confidential information to an AI tool based in the USA can be interpreted as a breach of confidentiality regulations. This even applies if you have concluded a premium contract that excludes the use of the entered content for training purposes.
As a solution, you may get in negotiations about every single NDA you receive and adapt it to the use of AI. Or, until further notice, don‘t enter confidential data into AI systems!
- Personal rights / GDPR (EU)
GDPR is an EU law for the protection of personal data. The preconditions to process and have personal data processed, for example the name of a client, his address, portrait photo etc., is: your AI tool provider needs to sign an official GDPR order processing contract. And additionally, the data subject personally consents. The solution here is to get the contract from the AI tool provider, and then ask your client, whether he is fine with his name getting entered into the machine – if you don’t want to do without AI in this case.
- Hallucinations and hidden bias
LLMs don‘t understand, what they deliver – they calculate the most likely variant, regardless of their truthfulness. AI tools are therefore based on the material with which they have been trained and are dependent on it. So, they may have prejudices and a bias. For example, visual recognition may work better on white men, and symbolic images for skilled workers preferably show the typical genders. To solve this, agencies should introduce a second loop for human content checking into the creative process, apart from the mere proofreading. The editor must be aware that there may be errors in the content of the delivered material.
Basic stance for working with AI
Before using AI tools on a large scale, you should clarify and formulate your fundamental position on them. Practical measures can be derived from such a strategic basis. At evercom, for example, we take this official stance on AI:
“The emergence of AI in society is irreversible, and we are pleased that it is so, as we consider it to be quite useful for our work. We must surely use it to expand our capabilities, increase our productivity and resize our knowledge boundaries. It allows us to exponentially increase the possibilities of adding value for our clients.
In order to fully understand human emotions, successful communication requires empathy. Connecting and weaving bonds is part of our natural condition as social beings. Our analytical capacity, critical spirit and creative vision are the essence of our profession and shape evercom’s identity. Their invariable human nature demonstrates that the human role in our industry, and especially in our agency, is irreplaceable.
Ultimately, we will be able to do more and improve by combining our natural abilities and experiences with the proper use of AI and other tools and technologies to embrace in the future.”
Applications of conversational AI in our workflow
To get to the practice, conversational AI is getting used for many different tasks at evercom, as these examples demonstrate:
► Documenting and browsing
“Tell me key dates on which sustainability is commemorated in some way around the world.”
► Summarizing texts, translating, or understanding them
“Translate and summarize this text with layman’s terms.”
► Text based on ideas or concepts
“Write a post of about 900 words on the need to legislate fixed pricing to tackle inflation.”
► Word games
“Write an acronym for e.v.e.r.c.o.m. with digital marketing words.”
► Buyer persona
Tilo Timmermann, Co-founder and Managing Director of the technology PR agency TDUB Kommunikationsberatung in Hamburg, Germany.
Ander Serrano, Partner and Director Tech & B2B Communications of evercom in Madrid, Spain.