Of course, when we sat down to write this quarter’s blog about AI’s impact on recruiting, we immediately went to ChatGPT and asked it to write the blog for us. AI was quite the cheerleader for itself, saying how it has “rapidly emerged as a crucial tool” in recruiting, and is “offering innovative solutions to age-old recruitment challenges”. In reality, while AI will assist in helping with more mundane tasks, almost everything it mentioned has already existed for years, and the majority of the recruiting process seems like it will be untouched by AI for years to come. Below we’ll break down our thinking on AI’s impact across various stages of the recruiting process, as well as HR.
Sourcing
Currently, AI has a hard time assisting with sourcing due to two things: (1) the need for real time data, and (2) the lack of access to LinkedIn. The data used to train the current ChatGPT model is almost two years old, which is a lifetime in recruiting terms. And since it won’t return any results from LinkedIn, even prompts like “Who are currently the top Machine Learning researchers” will only result in short lists of people who are generally famous, retired, or both, which are not useful for recruiting purposes. That said, LinkedIn does seem to have a golden opportunity to utilize its data for real-time models and create a premium product that could substantially streamline the sourcing process.
Another issue with AI assisted sourcing is that the signal-to-noise ratio could ratchet up with an automated sourcing system. This is because candidates could litter their resumes with keywords prone to AI prioritization (“Hey ChatGPT, here’s my resume, rewrite it to make it more attractive to AI screening tools”). Candidates could also use AI tools to customize their resumes to be more attractive to specific companies, such as Google, Meta, or Apple. All of which could make identifying the best candidates more challenging.
Another key aspect of sourcing that won’t change with AI is employee referrals. Typically, around 50% of Google’s hires were through employee referrals, and the early stage startups that we work with hire from within their networks more often than not.
One area where AI is definitely adding value is with the drafting of job descriptions. In the past, you would need to survey similar posts from several other company websites, and then draft a customized description to meet your specific needs. Now, all it takes is a few trial-and-error prompts to generate a highly passable job description for just about any role. Essentially, this means AI has taken about 90% of the work out of the job description writing process.
Automated Outreach
Recruiters have been using various drip email software solutions for years to automate early candidate reach-outs. These existing tools, and any new AI-enhanced developments in this area, may work well for low-level candidates, but we find that highly skilled or very in-demand candidates (such as AI engineers) need more of a human touch to stand out. Even with impressive advancements in AI email drafting, savvy candidates can spot an automated email pretty easily, so we always recommend that founders reach out to potential candidates with highly personalized notes for their most critical hires.
Scheduling
Automation at the scheduling level has existed for over a decade now, and most applicant tracking systems have built-in efficiency tools that will align skill sets between potential interviewers and candidates, and automatically find available days and time slots to pull together the earliest available interview panel. Considering the widespread existence of this technology, AI will likely be used to enhance these efforts, but not in a groundbreaking way.
Interviewing and Coding Assessments
Online candidate assessments have existed for at least a couple of decades now, even longer than automated scheduling tools. Google experimented with them back in 2004 and ultimately decided to keep the in-person interview process for everything so there could be more of a back-and-forth collaboration during problem solving exercises or when a candidate got stuck, which was key for assessing culture fit. Removing the human element with automated interviews and assessments greatly increases the risk of poor candidates making it to onsite interviews, which is a huge waste of time for companies.
Virtual interviews, a much more common occurrence today, present their own challenges. There will always be those who attempt to game the system by having a friend in the room feeding them answers or trying to discreetly ask ChatGPT for answers in real time. There are a few strategies that companies can use to combat this. For example, one of our portfolio companies, CodeSignal, has implemented a “Suspicion Score” on their coding assessments to help companies gauge whether a candidate may have used ChatGPT or something similar.
AI can definitely help with building out a set of interview questions for various technologies, but there’s an obvious risk to using these questions verbatim. Smart candidates will be doing the same, asking what questions are typical for those same topics. Companies may need to “AI-proof” their interview process, using non-traditional or non-obvious types of questions, avoiding questions that show up with basic AI prompting, doing more 1:1 coding interviews, and finding other creative ways to assess a candidate’s true skill level.
Looking forward, we feel that the main interview process will not be significantly changed by AI despite what tools may come out in the future. When someone will be managing or sitting next to this person every day, they will want to be deeply involved at every decision point.
Offers and Closing
It goes without saying that this stage of the process will always be personal and human-driven and will reflect how personal and human-driven things were throughout the entire process. This is especially true for in-demand candidates with competing offers. The back-and-forth negotiations, making one last strident pitch for a candidate to join, having founders or well-known advisors personally call candidates to try to convince them…the success of these efforts can hinge on how deeply connected the candidate feels to a company and their potential future coworkers. The more successful a company is at building this connection, the better chance a candidate will accept their offer.
Human Resources
This is an area where we think AI will have the most impact. HR teams essentially do the same cycle of tasks every single year. It’s also a highly templatized part of an organization, with potentially hundreds of policies and procedures, which will make AI highly useful as teams are building things out. For a small company without a lot of HR experience, AI will be incredibly useful for things such as building lists of potential questions for an employee survey, establishing an annual performance review process, or creating basic policies and employee handbooks. As always, having a human in place with HR experience to ensure that anything in the legal realm is accurate and up to date will continue to be important.
Summary
While AI is affecting the fringes of recruiting, there is still a ways to go for it to have a meaningful impact. At companies like Notion, HuggingFace, and Repl.it, which are all highly AI-focused and doing the most hiring in our portfolio right now, recruiting teams still look a lot like they did 20 years ago, with sourcers, coordinators, and recruiters putting candidates through an interview process that is largely unchanged.
Switching jobs can be a huge life change requiring a human touch. Candidates need to get a strong feel for the types of people that work at the companies they’re considering, and companies want to get the best sense of culture fit possible throughout the entire interview process. Candidates will also want a human to answer their questions so they can dig deeper and understand the nuanced differences between companies. AI can and will increase efficiency across aspects of recruiting and HR, but beyond the basics, these areas will continue to be highly personal and people-driven.
