AI requires recruitment teams to rethink skills and their value in the hiring process

As AI tools are leveraged by more and more employers and HR pros rush to understand what skills are needed for the workforce of the future, the AI tools themselves are beginning to offer talent acquisition (TA) teams—and their learning and development (L&D) colleagues—a new opportunity to reimagine how skills are understood and how skills competency is demonstrated and grows.

According to Tigran Sloyan, cofounder and CEO of AI-powered skills platform CodeSignal, AI poses both a challenge and a solution to an aging talent problem, especially when it comes to skills.

“There’s not a job you can’t name where there isn’t a certain set of inputs that the person starts with, set of work activities or work tasks that they complete, and then a certain set of outputs that they’re expected to deliver,” he said. To succeed in the rapidly evolving AI-enabled future of business, companies need to rethink and use AI to grow skills.

Cancel keywords

Instead of building recruitment and hiring strategies around static lists of required skills or résumé keywords, organizations can reverse-engineer roles from actual work.

He told HR Brew that the existing hiring process winnows candidates down by searching for skills keywords in résumés as a filter. But this process doesn’t serve recruiters or applicants. The tools are just detecting which applicants spelled out specific skills on their résumés, not whether they have any actual proficiency in them.

There’s no way to capture a candidate’s ability. Sloyan’s CodeSignal is helping TA teams understand what sets of tasks and skills are essential for performing a role.

“First, you’ve got to start from defining skills,” he said. “The right way to think about it is, skills are psychological attributes to allow you to perform certain work activities in the context of work.”

Teams can then better assess which applicants actually have the skills required for the job.

Bundles of skills

Skills are derived from what action needs to be conducted to accomplish work tasks. Sloyan pointed to a three-tiered model—essential, foundational, and tool skills—as a useful way of structuring skills in the AI-enabled era.

Essential skills are the core capabilities tied directly to job activities. They deliver the expected outputs—like the writing skill of a reporter or negotiation skill of a salesperson. Foundational skills sit underneath those essentials; writing skills include foundational spelling and grammar skills, for example. These are tool agnostic.

Tool skills relate to specific platforms or technologies, and Sloyan said his view separates out tool-specific skills from other foundational skills. Video editing on Adobe tools looks a lot different than CapCut, for instance.

Before the widespread deployment of AI, many organizations relied on long, static lists of skills inside their skills taxonomy framework, often facing challenges assessing and developing skills, and evolving them with the business. Technology is a major change here, too.

AI-powered

Historically, robust skill assessment at scale has been limited to technical roles, according to Sloyan. It was far harder to evaluate non-technical skills in businesses in a consistent way. But AI-powered simulation tools can now recreate real-world work scenarios and evaluate how candidates or employees perform against defined rubrics. Demonstrating understanding of skills can be consistently measured at scale with the help of AI.

There’s a future on the horizon in which recruiters may ditch résumés all together and deploy tools that can assess skills in action. That unlock could allow HR teams to continuously perform skills mapping, and L&D pros options for the technology to help train employees on the job, in the middle of the flow of work.

“It cannot be done without AI,” he said, “It changes too fast. “Humans can’t process that much data, that much information, that fast, and turn it around and say, ‘All right, our skills architecture completely changed over the last six months.’ We can’t. AI has to do it for us. AI has to be observing what happens, how it’s evolving, and being able to extract those insights.”