AI in HR: From Transactions to Innovation – by Katherine Jones, Ph.D
Some time ago, I wrote that the ubiquity of artificial intelligence (AI) in HR was not an “if” but a “when.” Now, “when” is today. The current uncertainty in all businesses from Covid-19 and a volatile political and economic environment has proved a catalyst for AI growth – modeling, for example, “what’s if” scenarios and amplifying the need to make predictive decisions for people and facilities planning, budgeting and forecasting — based on data.
Via adaptive learning, computers have the ability to modify output based on evidence or events fed into the program. This is commonplace today in applications like Spotify and the streaming channels (based on your recent television viewing patterns, you might like this similar movie or series.) But what makes it “intelligent”?
Previous generations of software had a lot of “if-then”—”if X occurred, do Y.” This kind of coding is ubiquitous in HR software: “if I am out of the office, then forward the approval request to my designee;” “if I hit the “hire” button, then automatically send new employee welcome kits and paperwork, add her to the payroll, and order a smartphone;” “if I give a performance rating of X, then my employee gets a salary bump-up of Y.” This isn’t intelligence—it is coded in the software, the steps are prescribed, and will be carried out in the same way until someone revises the software. This is how AI is different: it “learns.” Without anthropomorphizing too much, an AI program analyses data – ordinarily a lot of it—then “decides” what should happen next to complete a task. Thus, machine learning is a fundamental aspect of any AI program.
Web AI programs are like stalkers: if you look at shoes on one site, you are likely to be plagued with shoe ads for months to come. It learns the pattern quickly—but clearly is unable to tell that you already bought a pair of shoes.
Much as AI permeates the social platforms so widely used by kids and adults alike, similar “smart” algorithms are embedded in current HR and people management products. Many talent acquisition solutions today have algorithmic insights that can help recruiters ascertain apt candidates; other algorithms are frequently deployed to suggest career options both to candidates and existing employees within the organization. Based on history, the software is able to predict which applicants are likely to accept a job offer and which employees are presenting current flight risks. Employee education and upskilling are ripe for AI use, by suggesting learning by role, interests, and applicability to trainees. Note that in all these uses, AI just provides a new smart underbelly for applications and processes that are in widespread use in HR today.
Through machine learning, “rules” may be learned and applied, but if oversight is lacking, the results may be detrimental to the business. While these applications learn through the input of more and more data, they come “out of the box” with built-in assumptions, delivered as algorithms. And while it may seem like the path of least resistance is to believe the machine, HR leaders need to beware: algorithmic transparency is necessary. If, for example, a successful candidate in your financial services company has in the past been an Ivy-educated white male, and the data that leads to that conclusion is inherent in your talent acquisition software, what will your workforce of tomorrow look like? Will your diversity goals or your jobs-for-veterans initiatives be met given these historical assumptions? Likely not.
History bears this out in the now-familiar Amazon example: the company decommissioned its AI recruitment solution because it discriminated against women candidates – relying on what past successful workers looked like to fill the pipeline with white males. Amazon had fed the system a decade’s worth of resumes from people –mostly men– applying for jobs at Amazon who then were deemed successful in their jobs. Trained on that selection of information, the recruitment system began to favor men over women. The lesson here: one cannot always use the past to satisfactorily address the future, or in this case, even the present.
Last fall, HR.com published a study which included HR professionals’ opinions on AI use within their processes and their organizations. While most responders admit to not being very knowledgeable about AI and few saw AI-powered tools used widely in their enterprises today, they did see AI as a core part of their HR future. The AI-knowledgeable respondents have higher expectations for AI’s future in HR then those who professed to be less knowledgeable or who thought their HR departments lacked AI knowledge.
However, overall few knew what if anything the rest of their company was doing with AI, and, in other research, even fewer knew how others in their industry used or planned to use machine learning or AI-embedded solutions.
On the other hand, a great many employees today are used to AI support at home through Alexa and Siri, and are used to interacting in natural language with bots in many on-line and telephone transactions. It is barely a stretch to bot-supported common HR questions for candidates, new hires, and employees in general, and in fact, most leading vendors today provide a number of programs that use bot-based voice and text communication to interact with employees and customers alike. Many other vendors are racing to compete in this area, easing the path to widespread AI use through natural language processing and machine learning. These applications understand people’s inputs (rather than just code) and learn and ultimately improve feedback and predictions for actions based on that learning over time.
Again based on the HR.com research, is the overall view of HR professionals toward the uses of AI. HR respondents saw analytics and metrics as the areas in which AI has the greatest potential to improve the HR function, perceiving AI as a tool to further its ability to better understand larger amounts of data and better synthesize it to meaningful information for the organization—perhaps Excel on steroids. This in itself was hardly surprising. Lacking the tools and sometimes the analytic skill to accomplish this in the past, even with the many solutions available today, AI may appear as a quick fix to the long-standing lack of people analytics. Ironically, the laggards in the HR.com research saw greater use in providing actionable analytics than the leaders, who saw greater use in other categories such as talent acquisition, learning, compensation and payroll, and performance management.
Note that the ability of AI to dish up numbers fast does not alleviate the need for HR professionals to have the analytic skill to know if the right assumptions are being made, and whether the metrics delivered are meaningful. Becoming awash in data is not the goal. AI does not replace analytic intelligence and constant monitoring on the part of HR humans.
Following analytics, these HR professionals saw two benefits of AI: allowing people to focus on more value-added work, and automating tasks. Again, not surprising – if analytic data is delivered faster, and if machine-based predictions can be followed, users should have more time – and in this study, HR responders thought that innovation in their organizations would ensue because the people employed therein would have more time to become innovative. Now while this may be true, looking at AI-based applications solely from the viewpoint of office automation is to diminish its potential.
Today, AI in HR is mainly deployed to speed tasks that employees and HR staff are doing regularly: for employees, getting Bot-supplied answers in real time saves people-intervention, and that, in fact, saves time. For HR, the ability to get data faster, and to move that data into fodder for predictions is indeed useful. As an underlying technology, AI is currently in a raft of traditional human capital management products, and already proving useful in those settings where there is enough clean data from which the system can extrapolate and learn. But whence innovation? Enabling HR staff members to perform the same tasks they have always done faster or even more efficiently, is only a small step in progress. Let’s consider another question totally: how can AI be used to transform business processes, to move a transactional environment into a strategic one? Especially now, how can AI foster the agility required for our rapidly changing economic and political environment? Let’s look not only at what we can do with AI to improve current activities, but what we could or even should do with this very powerful technology for the future.
About the Author
Dr. Katherine Jones, veteran high-tech market analyst, is an independent thought leader in all areas of human capital management and the technologies that support it. She has been an analyst at Aberdeen Group, Bersin by Deloitte, and Mercer following a career that includes marketing in high-tech companies such as NetSuite and academic administration in higher education. Her master’s and doctorate degrees are from Cornell University. She can be reached at firstname.lastname@example.org or @katherine_jones.
Comments are closed.