Why Recruiters Need to Start Writing Job Descriptions That an AI Can Understand

By: Wade & Wendy Team

Aug 06, 2019

For decades, HR professionals have been demanding that candidates make their resumes machine-friendly. They have been warning them about the perils of the ATS black hole, and how to avoid it.

And now, it is the recruiter’s turn to operate in a way that is friendly to machine-scanning technologies.

Currently, job descriptions have a tendency to be:

  • Wildly varied in terms of structure and organization.
  • Often filled with buzzwords.
  • Often filled with ambiguous adjectives.
  • Vague in terms of firm requirements.

 

Once you bring an AI recruiting partner into the equation, all of this will cause big problems.

The job description is the foundational data point of the search, and when an AI is processing that data point, it needs information which is unambiguous, and quantifiable.

Human psychology is such that we have a natural and instinctive sense of what certain sorts of language mean. We are good with association, implication, and the cultural connotations of words. We know what it means when a job description describes a “hungry go-getter who likes to own projects,” or says it seeks “solid experience” or “excellent know-how.”

But an AI? These are cultural, colloquial sorts of speech, and it’s very hard to program an AI with this sort of interpretation-heavy diction. An AI would need to be told that the hiring manager wants someone with minimum five years’ experience, who has previously worked as a senior project manager, and possesses a graduate degree in the humanities or social sciences.

In this new era, recruiters need to start composing job descriptions in a semantic register that an AI can grasp. The requirements need to be precise and explicit. We need to be clear when something is a bonafide requirement, or when it is a rough guide. You’ve said you want five years experience — but would you take four a half years, if the candidate fulfilled all your other criteria? You need to tell an AI exactly what the threshold is. They won’t assume, like a human might, that there’s some wiggle room. They might reject the perfect candidate because they only have four years and eleven months of experience.

Information is all that matters. All the descriptive padding that currently fills up job descriptions; an AI won’t know what to do with this. It won’t want buzzwords; it will want straightforward descriptors of tools, experiences, role features, and so on.

We also need to start being careful with how we structure the descriptions. As an industry, the more consistency that we have in how we structure the components of a job description, the easier it will be for machines to understand. Right now, you often see a bunch of paragraphs vaguely thrown together. But an AI needs to know the framing of the section it is engaging with. Is this paragraph describing a role description, or the candidate being sought, or potential future opportunities, or something else? Again, an AI can’t assume. As we increasingly recruit AI assistants, we’ll need to arrange the information in a way that it is hard to misinterpret.

In the era of AI recruiting assistance, how we write job descriptions has to change. Recruiters that thrive will be those that learn to work with a tool like Wendy, rather than leave their habits unchanged.