Staatsanzeiger: Artificial intelligence can be used to analyse large amounts of text, for example those of CVs and letters of motivation. How well does that work already?
Alexander Brem: A lot is happening right now and the development steps are very large. No one can seriously predict what it will look like in one or two years. My impression is that OpenAI, a US company that researches AI, is way ahead and that Google and Meta are just catching up. More and more providers are coming in, especially start-ups. The company Rezi, for example, specialises in AI-generated CVs. The tools are easily accessible and some of them can be used free of charge. So you can say: for applicants and HR professionals, AI tools already work well.
If so, applicants could create the perfect motivation letter and the perfect CV. How can recruiters find out whether AI tools have been used to cheat?
That is almost impossible. We also have these problems in teaching: How can I check that a student didn't write the report himself? I think the tools to detect this will never be as good as the tools that create the texts. In the future, HR managers should just assume that applicants use such tools.
Sounds like a bad message.
It doesn't have to be: maybe it's even a qualification feature. You also have to be able to create a CV with AI. It takes a certain know-how and training to make these results really good. This is called "prompt engineering", i.e. the knowledge of how to write user input in such a way that the results become individual and high-calibre.
But a perfect cover letter still doesn't make a perfect applicant. So what do recruiters get out of it?
If applicants can summarise their CVs better with AI, present their strengths, weaknesses and competences well, the better for me as a recruiter. I can also see it as an advantage and say that the applicants will simply write better applications. Personnel managers can then more easily select the people they like. In a personal interview, at the latest, I can see in a few minutes whether the person is who they say they are. AI also helps on the applicant side, because many people find it incredibly difficult to create a CV and a covering letter. For example, if an applicant has dyslexia, an AI tool can help him express himself professionally. He might then apply for jobs he wouldn't have applied for without AI. In this way, HR managers can meet suitable candidates whom they would otherwise never have invited. In times of staff shortages everywhere, I think this is a good development. In times of staff shortages everywhere, a good development, in my opinion.
So is the personal interview gaining in importance?
Yes, absolutely. But it is just as elementary to choose the filters for the selection process wisely so that it remains fair. Basically, you can say that at the end of the day, AI is only as good as the people who created it. So if they have any conscious or unconscious biases, these may of course be reflected - consciously or unconsciously - in the AI tools.
Nevertheless, isn't there a danger that we rely too much on technology and less on our gut feeling when it comes to applicants?
That can happen. But of course it is a high risk for the company to hire someone only on the basis of their CV. That's why the face-to-face is so important. This can also be done via video call to get a first impression. You need a good concept of who you are actually looking for, what is special about your company, how you want to be presented to the outside world - more than ever, you need a recruiting strategy. This is how you avoid AI-generated job advertisements, for example, not standing out from other companies. This makes it more difficult to find suitable candidates because you become more interchangeable.
And how can AI help in recruiting employees?
For example, when creating job advertisements. But the idea of "I type something quickly and then have a perfect job ad" is too short-sighted.You need some know-how in typing the prompts.In the worst case scenario, company x has exactly the same job advertisement as company y. This is because AI is fed and trained with past-based data from other users. This is where it will be important to do prompts well - that is, telling the AI to "go to my company website, analyse my other job postings and generate a posting on topic X or Y based on that." The human hand is not going to become less important, both in terms of input and output.You definitely have to read over it later. Is this right, is this what I want, does it strike the right tone, does it also fit me as a company?
It sounds like the possibilities with AI are endless. Are there also tools for HR professionals that are not text-based?
There are many tools and more are being added all the time. For example, you could use AI to create a promotional video for your company. But here again it depends on the prompts and you have to pay a lot of attention to your own target group so that the videos are also well received.
What do you recommend to companies that are not yet working with AI?
Just try it out, get a feel for it. Or find someone from your team who has a bit of an affinity. You can take the first steps that way because it's self-explanatory. There are also free tools and trial versions.
Can you really do nothing wrong?
Unfortunately, there are also many scammers. I've seen in the app store that there are also ChatGPT tools that are not from the original provider. Often you pay money for a worse version. IT security also takes on a different significance. You should be aware that if you feed the AI with information, the data will end up somewhere. And if it's sensitive company data, that can be dangerous. In terms of data protection, I would rather prefer companies that are based in Germany. Because the penalties are really draconian. So the probability that a German will violate them is lower than with American or Chinese companies. You also have to be careful not to be naïve and feed the AI with any applicant data, for example. And that's why ChatGPT, for example, is not used by state institutions or by many companies, because the data is stored somewhere abroad. It must also be clear to you that if you upload a video with any advertising slogans, these could also be tapped by all the AI tools in the world and used again for others.
That is why the state administration is using F 13 in the pilot project. The AI programme summarises long documents in seconds and creates notes. Would you say that public administration is facing a turning point in terms of digitalisation?
That is exactly the direction to go. Administrative regulations can be explained in such a way that everyone understands them. The new AI tools can mean a quantum leap here. But digitising anything in the administration is really incredibly time-consuming. At the University of Stuttgart, we have an Agility Lab that deals with exactly this kind of thing. The most important thing is not to digitise a bad process. A digital travel expense application is no use if you have to print it out and scan it in again. The country is now also working more closely with start-ups. I think that's also an important step. That means you also increase the likelihood that the administration can simply try things out. Not only private companies, but also public companies can benefit massively from AI tools to become more attractive as employers.