Skip to content

Generative AI in HCM: Innovation and the Double-Edged Sword

Featured Image

A vivid personal memory from the late 1990’s was when a very talented and energetic member of my HR technology team based in Zurich (where I worked and lived at the time) literally ran into my office and said: “Steve, please hurry, you have to see this.”

He then ushered me into a small room that housed a computer terminal and printer and simply said: “Type any question or a topic or even the line to your favorite song.

That was my very first experience with the internet and using a revolutionary new product just introduced to the market. It was called “Google Search” and the year was 1998.

I’ll embarrassingly admit that I typed my favorite sports team at the time, the New York Mets; and lo and behold, the results that came back included story after story on the 1969 team that won the World Series that year after having a mediocre (or worse) team for many years in a row.

I was absolutely spellbound. That said, a short year later, my HR tech team was developing HR decision support tools for the web.

That was twenty-five years ago

and there are few who would debate the statement that there has been nothing invented and made available since, whether related to technology, science and medicine, manufacturing, transportation, education and learning or any other area of our lives, that approaches the profound impact the internet has had.

Note that I’m purposely not saying “positive” but profound impact. Why? Well, it’s simply because there might be just as many people lining up to say “the web has advanced societies and humankind in general” as there would be folks on the opposite line highlighting its deleterious effects on various aspects of society, such as with social media.

And therein lies the paradox if not conundrum of innovation, or in simpler parlance, the double-edged sword.

Now let’s fast forward to November 2022 and the public launch of ChatGPT

The AI chatbot from OpenAI, and a proliferation of largely competing products that have emerged (Bard, Bing, etc.) is clearly an innovation that is so dominating nearly every technology – and so many business discussions, we can see parallels with the excitement caused by the internet years ago.

As most already know, the ChatGPT platform includes natural language understanding (NLU) capabilities, automated search and response features, and integrations with a wide range of enterprise and personal computer apps and systems, as well as communication tools; and although not getting much attention so far, integration with ChatGPT alternatives.

In essence, we’re now talking about a veritable AI chatbot community, and a diverse one at that!

We will explore the wide range of use cases afforded by this new technology and assess the various impact on organizations in the 2nd blog of the series.

Let’s highlight two ‘Gen AI’ capabilities or uses garnering the early streams of attention in HR / HCM circles

Content suggestions or assisted authoringSummarization to ease information consumption

Examples of content suggestions getting the early coverage include policy writing, job descriptions, interview questions and possibly employee communications. The latter, employee communications, seems straightforward in terms of which situations lend themselves to Gen AI.

Clearly one that likely doesn’t is when real empathy and authenticity needs to come through, as with a communication about a workforce reduction or change in business direction which is sure to impact some more than others.

Of course, the premise here is that we might not be delegating these content writing tasks in their entirety to Generative AI, at least initially.

While the new technology, powered by very large machine learning models pre-trained on trillions of words, will be able to factor-in many aspects of context when writing a job description, it is likely that experienced HR professionals will still want to make the tool earn the trust of its human colleagues.

And regarding the summarization use case, this could turn out to be one of the major productivity drivers delivered by Generative AI, although the notion of proving itself to its beneficiary or user might apply here even more.

One area of concern is clearly the potential for bias, even in summarizing information or content. If the tool was not adequately trained on content from diverse sources and also writing styles, its take on the content will reflect the biases and predilections of those (narrow) sources and styles.

The good news, however, is that one of the beauties of Gen AI in HCM is that in the prompting process of the tool, one could seek to mitigate some of the potential for bias or inadequate training data (especially in the case of esoteric topics) by specifying for example that “the job description should appeal to and resonate with a very diverse audience.”

And over time, we can assume that users of this technology will know where the trust level should be higher or lower, with the consequence being more-or-less human partnering, as well as knowing where primary ownership of the end-product should lie.

In addition to the ‘human’ learning process that will occur relative to Gen AI usage scenarios, other key questions will start to get answered as well, although some might take a few years. Such questions include:

  • Will Gen AI systems and tools readily integrate with some HR technology assets and adjacent systems more than others?
  • In which job types and/or which personality types or approaches will Gen AI systems be more likely to positively – or adversely – affect employee productivity, job satisfaction, perhaps even retention?
  • When Gen AI does impact employee productivity, will it be by substituting for worker effort or by complementing worker skills and effort?
  • Do certain software development skills lend themselves to achieving Gen AI mastery more than others?
  • What capabilities might versions 2.0 or 3.0 of Gen AI include that earlier versions did not?
  • As in the case of sensitive employee communications being required, do we foresee a time when manifestations of empathy from Gen AI systems will be evident, and is that because empathy is being learned by the system?

In conclusion, we can say the word “learning” is unquestionably at the heart of this discussion, machine learning of all available knowledge, biased or “uncertified” or not, and human learning around the optimal ways to leverage Gen AI while minimizing potentially negative consequences.

One might look back on the last 25 years of the internet permeating our lives and presume that another way it will have benefited us is that we are now much more alert and attuned about potential negative consequences of misuse (whether intentional or not), while also having an opportunity collectively and individually to minimize those types of outcomes.

Moreover, let’s not let our concerns about potential bias, misinformation, harmful effects or mistakes diminish the justifiable enthusiasm this huge technological advance has generated.

Steve Goldberg
Steve Goldberg
HR Process & Tech Leader | HCM Analyst & Advisor

Steve Goldberg's 30+ year career on all sides of HR process & technology includes HR exec roles on 3 continents, serving as HCM product strategy leader and spokesperson at PeopleSoft, and co-founding boutique Recruiting Tech and Change Management firms. Steve’s uniquely diverse perspectives have been leveraged by both HCM solution vendors and corporate HR teams, and in practice leader roles at Bersin and Ventana Research. He holds an MBA in HR, is widely published and is a feature speaker around the globe. He’s been recognized as a Top 100 HRTech Influencer.

Related Insights