By Peter Weddle, CEO TAtech
The last year or so has seen been considerable positive commentary about using generative AI to write job postings. In some respects, this enthusiasm is due to the simple fact that the quality of online ads has no place to go but up. In fact, according to TLNT, fewer than half of employers – 48.6 percent – think their ads are effective. So, with all the ballyhoo about artificial intelligence, it’s no surprise that many see this latest version of the technology as the way to make job postings right.
There’s just one problem with this view … no, actually, there are two. The first problem has long been recognized in the development of large language models in general and generative AI in particular. It’s the data trap. The second problem has drawn less attention, but is no less impactful, especially in writing job postings. I call it the terminator trap. Let’s take a look at each of these issues in a bit more detail.
The Data Trap
Like every application of AI, generative AI is trained with data. In other words, what makes the software intelligent – at least in the beginning – are the text files and images it consumes during its development. That data is the source of the problem.
First, there are accurate, relevant and useful data, and there is junk. The quality control on what is fed to the model varies from developer-to-developer, and in those instances where there’s less control, the model is less intelligent than we need it to be (or think it is). In those cases, the quality of the resulting job posting is likely to be little better and potentially even worse than what gets posted today. It’s the GIGO principle jacked up with software.
Second, much of the data input is done by the model itself. To use an old-fashioned industrial term, it happens automatically. And, there’s the rub. As was noted in a recent article published by McKinsey, “Managing data remains one of the main barriers to value creation from gen AI. In fact, 70 percent of top performers in a recent McKinsey survey said they have experienced difficulties integrating data into AI models, ranging from issues with data quality, defining processes for data governance, and having sufficient training data.” And, that aspect of the data trap can also produce lousy job postings.
Get worldclass content and business-building connections at TAtech Europe & The EMEA Job Board Forum. Avoid FOMO, register right now for the only conference in the North American market that is specifically designed both for job board and talent technology company CEOs, their direct reports and rising stars and to accelerate the bottom-line growth of their enterprises. Why the rush? Because the conference will be held June 4-6, 2024 at The International Spy Museum which has a limited seating capacity. So, register today and accelerate the success your company achieves, today, tomorrow and into the future!
The Terminator Trap
Let’s imagine a brave new world where all of the above data problems are fixed. Only the right kind of data are used to train new gen AI products and that data is well understood, calibrated and managed by humans who know what they’re doing. In short, we actually get artificially intelligent models that recruiters can use to write decent job postings. Problem solved, right? Unfortunately, no.
An ad can be accurate and complete and still not motivate top talent to apply. AI can ensure an ad has all the right words for SEO; all the details on a job’s requirements and responsibilities, and all of the specifics on its salary and the company’s benefits and still be totally ineffective. Why? Because job postings are not position descriptions; they are expressions of opportunity. To be effective, they must make both an intellectual and an emotional connection with prospective candidates.
AI models are like passive terminators. They don’t harm people, but they also don’t touch them emotionally. The job postings they write are soulless. They don’t inspire the reader or stimulate their aspirations. That’s not to say that job postings composed by humans are any better. Sadly, most aren’t. And, that’s why so many of them fail to touch job seekers, let alone the top talent in the workforce. In most cases, high performers are well cared for by their current employer, so they have to be enticed away, and only an ad with a fulsome expression of humanity can do that.
Generative AI is a powerful new resource in recruiting. But, let’s be squinty eyed about its capabilities. In the future, it will undoubtedly be a wondrous partner in the work of a recruiting team. But for now, it faces two traps that limit what it can do and how well it can do it, and that includes the composition of job postings.
Food for Thought,
Peter
Peter Weddle has authored or edited over two dozen books and been a columnist for The Wall Street Journal. He is the founder and CEO of TAtech: The Association for Talent Acquisition Solutions.