Feature
Technology has been helping HR practitioners parse and analyze data for more than a decade. The advent of generative AI is bringing those capabilities to a whole new level, adding speed and power to HR analytics use cases such as sentiment analysis, learning development matching and performance review synthesis.
“Analytical AI could have done all these things, but it was much more cumbersome, and the result was of lesser quality,” said Julian Kirchherr, a McKinsey partner and co-author of a McKinsey report about using generative AI in HR. “They all went under the heading of ‘HR analytics,’ but generative AI provides a big boost to many of these applications.”
HR professionals are rightly intrigued, but it’s crucial that they take the time to understand best practices for integrating these powerful tools before adopting generative AI in their workflows. Here are some of the more important best practices to consider.
Proceed Slowly With Proper Oversight
Dr. Dieter Veldsman, chief HR scientist at Academy to Innovate HR (AIHR), recommends putting in place a multidisciplinary team with oversight over any AI experiments the organization runs, whether in HR or another department.
“Start small and then scale over time as you learn lessons,” he said. “I know this is not a popular view, but go slow. I think the people who will win the AI race are those who adopt it responsibly, not necessarily the ones who adopt it first or quickly.”
Within HR, oversight should also come from the head of the department who “needs to have a clear perspective on what they are trying to achieve with all this new technology,” said Stacia Garr, cofounder and principal analyst at RedThread Research. These leaders must establish necessary processes and practices to ensure these tools are being used responsibly and usefully.
Related Article: Glitch in the Machine: Why HR Should Tread Carefully Along AI's Path
Integrate Your HR Technology
Fragmentation of HR technology is a common issue. It can prevent meaningful use of AI tools. Garr calls addressing this problem “a critical first step” in using AI to assist in HR tasks.
“Many companies are not aware of all the technology they own,” she said. “It is important to have a connected and meaningful system in place to facilitate this type of work.”
An audit of all your HR tools and the functions they play in your department’s workflows is a good place to start.
Establish Good Data Governance
Considering that HR departments handle some of an organization’s most sensitive data, it is essential that AI usage be combined with excellent data architecture and governance. It is a good idea to ring-fence HR information that is being used in data sets for any large language models.
“You have to be very careful in how you set up your data architecture if you work with AI in HR,” said Kirchherr. “Really think about where this information is being saved. There are good pathways for all of this, but it needs to be carefully managed.”
Understand Your AI Tools and Datasets
Not even the creators of large language models fully understand how they work. But people who use them to analyze specific datasets should at least have a sense of their strengths and pitfalls, as well as a solid picture of the contents of each data set.
“We ran an experiment with our own internal AI bot, and it completely hallucinated a lot of information that we knew we didn’t have in our database,” said Veldsman. “It’s important to be very cognizant of how the model learns and the controls that you want to put in place.”
A baseline control for anything produced by an AI tool is to have humans check its results against the data.
Related Article: HR Leaders: With Great AI Power Comes Great Responsibility
Get Everyone on the Same Page
Your employees should be told when you are using AI in any way that may impact them, such as in synthesizing data to create performance reviews. They are apt to be concerned and suspicious if they find out after the fact.
“Losing trust in the relationship with the employer can occur if employees are not brought along the journey,” said Veldsman.
The term “AI” is often indiscriminately used for any advanced analytics capability, whether it uses actual AI technology or not. So, part of the process is ensuring that everyone in your organization is working from the same definition of AI.
“Can you clarify the terminology used in your business when you talk about AI? What exactly are you referring to?” asked Veldsman. “Currently, it seems to be a broad term encompassing many things that it is not. It is important to use the correct terminology to ensure a common understanding within your business.”
Use AI as Your Copilot
Generative AI is powerful and has the capability to transform entire portions of HR practitioners’ jobs. However, large language models have been shown to fabricate information and confidently assert wrong answers — known as hallucinating — which makes it a professional liability to depend on these tools too much.
“Don’t delegate your job to generative AI,” Kirchherr warned. “It hallucinates. It gives wrong answers at times.”
Conference
Oct
20
Gartner IT Symposium/Xpo Orlando 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
Instead, he says, approach AI as “your copilot” — a fast and knowledgeable assistant whose work you should always check and who still has considerable limitations.
By abiding by the best practices above, you can ensure your copilot is a trustworthy partner in navigating your daily workflows and responsibly benefiting from the transformations coming to HR.
