“Hey Siri – complete my performance appraisals!” When I tried this approach or when I asked ChatGPT to do the same, the AI algorithm did not understand the request. While I was relieved to know that humans might still be in charge of some of the HR processes, I do know that AI is being leveraged in a number of HR areas already. Ethical and responsible leverage of AI in business has been a bit of a debate with me and my colleagues.

While many researchers and managers in organisations are leveraging AI tools in areas related to customer experience, business operations, competitive insight, and product enhancement, there have been a number of increasing concerns regarding decision-making and ethics. In addition, there is growing concern about the transformation of jobs (even that of a university professor) across all sectors. 

Human resource management is no exception. For decades, HR professionals have been concerned with the overly administrative nature of functional roles as this can often limit the capacity for more strategic, value-added HR services. AI promises to help reduce some of these administrative tasks and may be able to provide more responsive insight to the business. As we know, AI is already taking hold in several HR processes. Examples include:

Employee Self-Service Chat Bots – providing support for employees regarding the basics of address changes, benefit elections, life events, etc.

Recruitment – finding potential candidates on social media, screening candidates/resumes, analysis of video-based interviews, skill fit analysis, job fit with criteria, etc.

Development and Careers – AI can track learning progress and map it to career interests to help individuals map out their future plans. Some universities are providing these services to students and some businesses see the value in long-term workforce development.

Motivation and Engagement – While this is not yet a common area, some firms are using AI to help measure the social climate and engagement. By asking key questions of employees through pulse surveys and linking to other work tools, patterns can be established to understand potential turnover risks and general commitment levels.

Productivity Support – AI can help support worker efficiency by pointing out meeting scheduling patterns, use of time, email tips, application usage, and other areas to suggest improvements toward productivity.

While these examples may seem somewhat standard, new questions emerge when we consider areas such as selecting candidates for open positions, gathering input for performance evaluation, suggesting career development, evaluating pay practices, monitoring employee actions, suggesting cost cutting measures, and other areas where we typically rely on a combination of quantitative data and human judgement. 

As we consider more advanced AI applications in the areas of human resource management, three important factors seem to emerge: neural network models, ground truth, and large language models. 

Neural Network Models – We can think of this as basically the patterns by which the machines learn using a series of data inputs in layers over a series or stages. When we decide to leverage AI for a particular application, it is important to understand the starting point for the model (e.g. is it pre-trained or built specifically for this use?). We need to ensure that we understand how we selected or built the model and what types of data are used as input. For example, the machine may not understand our values related to diversity and may unintentionally perpetuate bias in the model. It is important to ask the AI experts how they analyse the model for error detection and error analysis. While we don’t all need to be AI experts in building models, it is good to consider the questions related to using the right model.

Ground Truth in Decision-Making – In these days of disinformation, it may be hard to understand the real facts vs. someone’s opinion. Professor Lebovitz at the University of Virginia and colleagues have noted that the ground truth in AI tools can be a critical element in our applications of AI. Ground truth refers to the information that is known to be true based on objective, empirical evidence. In other words, leveraging subjective comments from employees or other sources is not solid ground truth. Before deploying AI solutions, managers should be certain that the data sources are verified. Too often, managers look at the accuracy of predictions in the model as the evidence of AI quality. However, this face value may not be appropriate in the unique context of your organisation. Understanding the ground truth is critical as an anchor for HR-related AI systems in the future.

Large Language Models (LLM) – The power of the new large language models is exciting as we see people using tools like GPT-4 to write speeches, papers, and even songs. In fact, some university students have been caught using AI to submit reports for classes or even write application essays – while this is creative, they receive a failing grade since this is not allowed (yes, universities use AI to check for use of AI). Around the world, businesses are experimenting with LLMs to communicate with customers, manage vendors, and other such applications. Creative IT and HR managers have suggested that we might also manage employee communication using LLMs and even put together performance evaluations. While this is certainly possible, this begins to cross the line of machines managing people. Caution should be noted and HR leaders would be prudent to consider a proactive approach to decisions related to LLM deployment.

In the future, I might not ask, “Hey Siri – run the HR Department!” However, there are several areas where AI can really make a positive impact on HR efficiency and insight. For now, I will trust that we have human judgement in place for big HR decisions and directions, but fully expect that most will decide that detailed tasks and robust data gathering is best left to AI. One of the areas that we will continue to check is how we might eliminate bias and create more equality and rational decision-making. After all, this is one of the advantages of machine logic and it would be great to use this as an opportunity to take an evidence-based approach to management decisions. By embracing the potential with AI while also noting the risks, we might create an exciting future in the HR profession.