When you last saw an interesting job opening and browsed through the list of “desired qualifications”, did you wonder who came up with those? No matter if it’s a sales, engineering, or operations role, there are likely 10-12 seemingly related yet uniquely phrased bullet points outlining the ideal candidate. Was it an insightful hiring manager? An experienced recruiter? Or an algorithm supported by a database? Chances are, it’s a combination of all three and beyond. The follow-up questions then became: How did humans and machines work together to compose the skills list? Whose input takes precedence? Is this approach truly effective in attracting talent that fits not only the particular role but also the company culture and long-term growth plan?
Workday has been trying to address many of these questions and revolutionize the concept of jobs through its Skills Cloud. It uses the concept of “skills ontology”, which I think is better explained as “skills taxonomy”, to decompose each job as a bundle of tasks – give a presentation, conduct web analysis, develop a team of people, etc. Each task then demands for relevant knowledge, skills, and abilities.
“After all, AI is so powerful it can tell us which route to drive to work, can’t it tell us what skills we need?” In Josh Bersin’s recent analysis of Workday Skills Cloud, he used the analogy of Google Maps vs. Waze. The former tells you how the roads are supposed to work, whereas the latter shows how people are actually navigating them. While traditionally jobs have been configured through top-down business requirements, Workday Skills Cloud is trying the Waze approach. It’s building a system that “self-describes the skills, capabilities, and experience one needs to succeed”.
The sought-after skills are synthesized based on how the top performers in existing roles are executing their tasks and qualifications that make them effective leaders or individual contributors. With its inherent advantage as part of the Human Capital Management and even Finance ERP ecosystem, the Skills Cloud sources data from performance metrics, talent profiles, learning & development records, and recruiting information across industries.
Similar to other people analytics solutions that offer predictive recommendations based on existing “success examples”, Workday Skills Cloud could be perpetuating issues of representation. It brings to question the differences between correlation and causation in the context of talent management: What are possible explanations people excelling in their roles share certain traits? While a degree from a prestigious private institution could be a commonality among existing top performers at a Fortune 100 company, is it a necessary qualification for someone to have in order to succeed in the role? As new technologies and tools emerge, instead of seeking candidates proficient in the exact same coding languages as the company’s star developers, would it be more valuable to identify individuals who can pick up new tools quickly to complement the existing stack of tech skills?
Another challenging metric the solution is trying to measure is the “strength level of skills”. If someone identifies “project management” as a skill, how strong or transferable is their skill across contexts? How does it change over time? While Bersin discussed his belief in “capabilities as skills plus experience”, he didn’t quite explain how “experience” should be quantified, how to factor in and weigh measures such as tenure, number of related projects, support vs. leadership roles taken. Another approach to objectify the strength of skills is adopted by networking and learning platforms such as LinkedIn and Edcast as they popularized human “tagging” – the more people who are willing to endorse you on a certain skill (especially if they’re also good at the same thing), the higher the perceived strength and credibility in that domain. Workday only revealed considerations such as the depreciation of a skill over time as it could become rusty, yet much of the strength rating algorithm remains a black box.
The opaque nature of AI-enabled skill evaluation makes me hesitant to support its use in hiring and promotion decisions, yet I agree with Bersin on its potential to yield aggregated insights for companies to understand their collective skill gaps or shortage in market competition. Personally, I’m most excited about how Workday Skills Cloud could promote a culture of upskilling. By highlighting the skills in demand within one’s organization, it could encourage employees to pursue development resources or project opportunities that hone in on that skill. Essentially, I think the current analysis of skills data is most powerful when used to inform employees of their growth potentials and how their future could be aligned with the direction that their organization is headed.
To answer the question of how machine and humans work together in the realm of skill identification and job mapping: Workday’s stance is to let the algorithm do the grunt work for defining “success” before humans perform the check and balances and make judgment calls on how the data is used in talent decisions. At the end of the day, both Google Maps and Waze are created to serve the commuters who opted to share their data. My hope is that similarly, the cloud of skills can provide forecasts for employees navigating their career journeys on how to amplify their impact.