Many new software applications are intelligent in ways that yesterday's app developers only dreamed of. Computer-created intelligence called AI -- for artificial, adaptive or augmented intelligence, depending on which vendor you listen to -- is enabling improved corporate security using anomaly tracking, 24/7 production schedules via robotics, more flexible supply chains, improved medical diagnoses and a host of other positive business outcomes.
HR departments cannot escape AI either: Every major vendor touts an AI-based product to make HR professionals' lives easier.
Today's applications of AI in HR surpass the merely analytic capabilities that ease HR's job by gathering information that used to be difficult to collect and synthesize into actionable planning. Algorithms today support the collection and analysis of huge volumes of data and can see correlations that can lead to predictions.
Questions one could conceivably answer include the following:
- What are the best sources of candidates that will accept positions offered to them?
- What candidate traits best correlate with long-term success?
- What is the work history of managers who prove most successful at fostering promotions among diverse workers?
- What is the length of tenure in one position that triggers the most attrition in Gen Xers?
But can AI in HR measure fit?
We all know that a star player in one organization may end up being a flop in another. Why? Did the person change? Not likely. Rather, they fit into the culture of the organization.
So while AI-based talent acquisition software may well be able to match skills to jobs or link people's skills to their experience, what would an algorithm use to predict fit? Even if your app told you which people from company X were successful in your organization in the past, does that help you with the candidate in front of you? In addition, some experts question the importance of fit when working with a distributed remote team of, say, freelancers or home-based workers.
Hence, relying on a quantitative AI-based determination of the likelihood of a candidate's success in your organization may not produce the results you hoped for.
Beware the AI in HR bandwagon
Almost every vendor making HCM products today is likely to say its product is AI-based. What does that mean? Some A isn't so I.
Here are some questions to start with to ascertain how a product actually works:
- If the product makes predictions, what data does it use to do that?
- How does the algorithm weigh or use the historical data that it accesses?
- What other customers are using this product today and what is their reported range of error in prediction accuracy?
- If the algorithm uses historical data, will relying on it create an organization fit for yesterday but not for the future?
Ascertain if your organization's history logically seems suited to its future; in today's rapidly changing economic and business world, it is unlikely. This also applies to the many AI-based talent acquisition programs that look at candidates through the eyes of past hires, which could be heavily weighted with Caucasian males, hampering your diversity goals.
AI should make decision-making easier or smarter -- or both. Solution providers purport to do that. Buyers, however, should consider how they will measure AI's success in their organization: Will the intelligence provided help them make fiscal decisions more accurately? More timely? In creating the business case for using AI in HR applications, proposals will have to clearly articulate the intended results and why they surpass the traditional solutions already installed.
Goals of AI
What are the implications of adaptive intelligence for your new applications?
As desirable and useful as implementing AI in HR can prove to be, when HR makes decisions based solely on computer-derived algorithms, someone in the organization must understand what factors are used in creating the correlation or the prediction and how the algorithms work. Learning systems do just that: They learn and apply rules based on data gathered over time. What the application learns is supposed to affect how it behaves -- hence, it adapts.
While the decisions made based on AI may prove correct for your organization, you still have to make sure that the assumptions applied over time are understood and reviewed for accuracy, fairness and relevance to organizational goals.