Matching resumes to job postings is not how job boards should use AI

In addition to being the founder of College Recruiter job search site, I’m also the cohost of the Inside Job Boards and Recruitment Marketplaces Podcast. In my role at College Recruiter and as cohost of the Podcast, I often think, learn, and talk about the use of artificial intelligence by job boards, including to enhance employer branding and engagement with candidates before they even apply.

The industry has existed for more than three decades and, for that entire time, it has demonstrated the ability to not just encounter new technologies that could potentially help or hurt, but to absorb those technologies to make them an integral part of our products. The use of AI is no different and tremendously promising as well. Some job boards see the Holy Grail in the form of AI being used to match candidate resumes / CVs / profiles to employer job posting ads. I don’t. Yes, I can see that AI can do that well and at scale, but only when the data it is trained and then acts on is good, and the reality is that the vast majority of resumes are poorly written, backward-looking documents. 

Resumes do an okay job of allowing the reader to understand where you’ve been, but almost never where you want to go. Similarly, job posting ads are typically poorly written, forward-looking documents that allow a reader to understand what the employer wants a future employer to do, but not the employer’s requirements and preferences for the skills needed for the role. As a result, the recruiter, hiring manager, or AI is forced to infer from resumes what the candidate should be able to do well in the future and from job postings what qualifications those candidates should have. 

AI may appear to do a better job of matching resumes with postings, but that’s largely because few AI-powered systems are transparent. You feed the resume and job posting data in and receive back a black-and-white list of prioritized resumes showing you the candidates ranked from best to worst. The problem is that the AI often has no objective, rational basis for ranking one candidate above another, and those running it have no way of knowing if the ranking was accomplished, even in part, through the use of illegal factors such as whether a candidate was of a certain race, gender, or other such factors. 

That said, I do feel that AI has tremendous promise in our industry, including matching candidates with employers. I just don’t feel that matching resumes with postings is the Holy Grail. I see more promise is AI essentially acting as a career coach to help candidates better understand their past accomplishments and determine what their future career path should look like. That will create forward-looking data about the candidate’s desires and likelihood of being productive in certain roles, and then the AI can be used to better match that candidate with roles based on the brand of the employer and details about the role itself.