Welcome to the new and tricky world of applying for a job in the AI era. Yes, applicants can use LLMs to fill out applications and cast a far wider net. But AI has also made recruiters far more anxious about the authenticity of the people they’re evaluating and, in some cases, it is clogging up candidate pools. In a stagnant job market, it feels like a tech-fueled race to the bottom; it’s not clear anyone is really winning—at least right now.
David Paffenholz, the cofounder and CEO of Juicebox, sits at the center of these tensions. His company runs an AI-powered recruitment firm, which provides a search engine for candidates and AI-powered candidate evaluations. As a result, he’s had a front-row seat to how artificial intelligence is transforming the way we apply for jobs.
“It’s this kind of weird middle ground where there’s been a lot of AI application tools, and a huge surge in volume of applications because of that,” Paffenholz explains. “Meanwhile, recruiting teams are a lot more constrained in where they can use AI to tooling and application review. . . . So it’s resulted in this weird paradox for application volume and surge. . . . The recruiter tooling hasn’t quite been able to keep up.”
Fast Company spoke with Paffenholz to learn more about how AI is reshaping how candidates craft job applications and how job applicants are responding. The cover letter is definitely dead, he says. And if you haven’t started playing around with ChatGPT—and you’re still looking for a job—now might be a good time to start.
This interview has been edited for length and clarity.
How would you describe the state of the AI talent wars at this point?
There’s intense competition for top talent, and I think a lot of that [is] driven by the kind of research that goes into developing LLMs. But what we’re seeing today is that all types of companies that are building with those LLMs have been, to some extent, going through a similar talent war. They are looking for the best possible talent, often technical talent, but also nontechnical talent, to build out their products, and oftentimes continue building their products using LLMs.
It’s kind of one of those things where you know it’s an entirely new skill set, and it’s hard to find people with lots of experience in it because it’s only been around for a few years. There’s pretty intense competition for people who’ve shown that they can build with LLMs and use that to develop great products.
Where do you find people who are LLM-fluent?
The interesting part of LLMs is that it’s a little bit different from previous skill sets, where there used to be a more defined way of going about learning them. I think what we often see, even for large companies, is recruiting more aggressively out of startups . . . because startups tend to have an environment where it’s easier to go and implement new tooling and build actual products with that tooling pretty quickly. That’s also something that’s kind of driving some of the acquisitions that we’re seeing in the space.
What do you think people should be doing to be positioning themselves well here? What should people be practicing, even on a consumer level?
We see this consistently both at the top of funnel, which is where we serve customers, so finding talent . . . but also during interview processes, where it’s become almost a default interview question of: How have you used AI? How have you engaged with AI? Do you have examples of that?
The cool thing is that it is something that everyone can do and can show evidence of, even if it’s outside their work environment. There are so many consumer-facing AI applications, and so much opportunity to use and test those, let alone build something with those, that I think it is actually possible for candidates to prove their AI fluency or AI skills by using the tooling that’s available.
A lot of people are using AI to fill out and populate applications. How big of a problem is that on the recruiter side?
What is the value of a cover letter if, like, 95% of cover letters are AI-generated. . . . Other things that are going to evolve to be a way for candidates to show their uniqueness and their strengths, say through a case study or take-home.
What about AI applicants—meaning applicants who aren’t real?
There are two main areas on the application side. One is kind of malicious actors, where they have an intention of deceiving the employer. That’s particularly prevalent with remote work setups, where someone can actually be taking on the job, whether they have a goal of stealing IP or whether they have a goal of actually making income through fraudulent means. . . . The second category, which is perhaps harder, is maybe it goes more in the direction of like, embellishment—at large scale—from candidates.
How would you respond to pessimism from some job applicants who feel they’re just sending stuff into the abyss? That they have to send applications at scale because they’re just dealing with AI readers. As a result, they, in turn, just need to use AI to generate everything.
It’s on the employers to make sure that if they are using AI systems, they’re doing it in a way that’s reflective of the goals of their hiring process. Even outside of the regulatory piece, [which involves] ensuring that there is no bias in any system that they use. It’s also about getting them to the goals that they actually want to have in their hiring process. I think AI tooling has potential to do that really well, but also has to be used quite carefully by an employer.
Part of that is that employees have an obligation to also inform the applicant: Hey, is AI being used in this process? So that the applicant is aware of that as well.
