AI Will NEVER Replace Recruiters

Artificial intelligence remains the talk of the town in 2024 with experts regaling us with its endless applications to business and life. As a career headhunter, I had concerns at first. Would AI replace me before I could retire? After digging in a bit, I think that will NEVER happen, because of the data model loop that drives AI functionality. Between the amount of data it needs to work, the pace at which data changes, and the nature of external factors make it impossible.

?

BIG Data Need

AI runs on big data sets to work, but how much? This is a hotly contested question, but everyone agrees it’s a very large amount. Depending on the complexity of the model and the number of features/columns, it could be 100x to 1000x of data rows! What does that mean?

If you are tracking 10 features/columns in a dataset for recruitment purposes, let’s say for candidate attraction. It might look something like these:

1.????? Job Location

2.????? Candidate Previous Employer

3.????? Candidate Education Level

4.????? Were They a Previous Applicant?

5.????? Applicant Source (LinkedIn, recruiter outreach, etc)

6.????? Candidate Job Title

7.????? Salary Offered

8.????? Candidate Engaged with Brand?

9.????? Application Date

10.? Hybrid Policy

If you wanted to get some solid analysis from this grouping out of an AI model, you would likely need 1,000 to 10,000 actual examples (i.e. applicants/hires). There’s also no guarantee that would give you anything of value. The basic premise is the more datapoints you are tracking (the 10 above), the more real-world examples you need to train the model appropriately. Many of the Fortune 500 are big enough data wise to do this easily, but I would argue the appropriate recruiting data set is infinitely larger than 10 fields for any human influenced topic like recruitment. ?

?

External Factors in Recruitment

Selecting a career is a huge life decision. Hundreds if not thousands of variables influence that decision for every person that makes it. Now I want you to think about how many of those variables are trackable and trustworthy from a data perspective?

Take an applicant to a job for example. We can track everything on their resume and application, other online interactions with our company brand, social media engagement/activity, amongst some others. The goal at the end of the day is trying to pinpoint the factor(s) that drove the decision. What about the untraceable factors? What about their family life? Do they have children? Do they take care of aging parents? What's your company’s current Glassdoor rating? What’s the candidate’s current financial situation? Are they 2 or 25 years from retirement? Do they have a negative opinion of your entire industry? Is your office right next to their daughter’s school? Is the candidate hugely influenced by the benefit package instead of the job? We could ask 1,000 more. The larger point is with all the hundreds of data points we can track, even if done perfectly, they will never fully represent the true influencing variables in totality.

?

Constant Market Change

I started my recruiting career in 2006. To say the job titles and business needs have changed since then would be an understatement. We don’t see Domino Developers much anymore and we certainly didn’t see machine learning or blockchain jobs back then. The business world changes, constantly and rapidly. If the business world is changing rapidly, then the data representing its titles, numbers, and movements certainly are as well.

Why does this matter? Well, whenever our executives asked for data when I was in corporate, it was always regarding the new, hot, hard to fill skillset of the moment. The one that we likely had very little data on at all, let alone to train an AI model on. Now apply this thinking to the hardest job to fill in your organization. How long would it take you to get to 1,000 or 10,000 hires in that title? If you built an AI model then, not now, would it still be valuable at that point? Perhaps.

?

Let’s revisit what we’ve established so far:

1.????? AI models require lots and lots of data to work correctly, which grows with complexity.

2.????? Recruitment decisions are influenced by lots and lots of data points, which makes AI models more complex.

3.????? The business ecosystem doesn’t sit in a vacuum. As new technologies are utilized in business, they are demystified over time to later be replaced again by newer tech and the cycle continues, also creating new datasets.

4.????? Hiring leaders seek to understand the problem of the moment via data, while the data may not be reliable enough at that point.

?

AI Isn’t Replacing Recruiters

Artificial intelligence has many applications in the recruiting world. Its an excellent tool that can be applied to nearly every use case that exists. That being said, it seems to be a circular issue of new market jobs changing the data environment enough to make employment data sets too diverse to compare over time. With the addition of world environmental factors having heavy influence and being largely untraceable, I don’t see how this gets remedied. Rest easy recruiters, we are safe another day.