Job Board Leaders’ Roundtable with Gerry Crispin and Ben Eubanks

This month’s gathering of leaders of job boards and related organizations had a bit of a different format from our 17 previous meetings.

Gerry Crispin of CareerXroads and Ben Eubanks of Lighthouse Research & Advisory discussed Ben’s LinkedIn post and his subsequent conversations there with Gerry about a recent academic study about the impact Freelancer.com’s LLM-assisted proposal writing tool had on the quality of hire. The study found that low performers on similar projects were hired nearly 20 percent more often, and high performers hired about 20 percent less often. Why? Because the LLM caused their proposals to look very similar, and so more employers tended to hire based on price instead of qualifications.

Ben and Gerry summarized the study, discussed where they agreed and disagreed, and many of the other attendees chimed in to share their thoughts, ask questions, and offer suggestions for how we can address the problem.

AI-generated show notes:

Meeting Introduction

Steven Rothberg mentioned that the upcoming discussion is based on a dense 129-page “article” that they believe should be discussed due to its important “learnings”. Steven Rothberg credited Matt Farrah as the founder of nurses.co.uk and niche boards in the UK, for the idea of these regular gatherings.

Gerry Crispin’s HR History and Longevity

Matt Farrah commented on Gerry Crispin’s extensive background in Human Resources dating back to September 1975, which Gerry Crispin acknowledged by joking that they aim to be “the last man standing” by outlasting Lou Adler, who is a year older. Ben Eubanks shared that their favorite stories from Gerry Crispin involve sending out books with jobs in them.

Job Board Leaders Roundtable and AI Discussion Topic

Steven Rothberg formally opened the “Job Board Leaders Monthly Roundtable” for January 2026, held on January 8th. The topic for discussion, introduced with special guests Gerry Crispin and Ben Eubanks, centered on whether AI-powered hiring systems are leading to higher or lower quality hiring, prompted by a white paper suggesting the latter.

Guest Introductions and White Paper Summary

Gerry Crispin introduced themself as having 55 years of experience in recruiting and a background in industrial organizational psychology, currently focusing on building a peer-driven community of talent acquisition leaders; Steven Rothberg referred to them as the “godfather of recruiting”. Ben Eubanks, Chief Research Officer for Lighthouse Research and Advisory, summarized the research paper, “Making Talk Cheap: Generative AI and Labor Market Signaling,” which studied a freelance matching service and found that applications, made easier by AI, led to lower quality candidates being selected more often, crowding out higher quality applicants.

Critique of the White Paper’s Focus and Methodology

Gerry Crispin argued the paper is important but misdirects attention, suggesting the focus is more on economics—differentiating “signal from noise”—than recruiting. Gerry Crispin detailed the brilliant methodology where authors measured effort and customization in applications (pre-AI) as a predictor of performance in writing gigs on freelancer.com, which was later validated by post-gig ratings. The shift to post-LLM usage resulted in double the application size, making the metric of predicted performance less valid and leading to a more “random selection” process because applications appear too similar.

AI’s Impact on Signal and Candidate Quality

Ben Eubanks clarified that the jobs studied in the paper were coding jobs, not writing jobs, but agreed with Gerry Crispin that collecting performance data only for selected individuals is a challenge. Ben Eubanks shared an anecdote of a friend at Boeing who hired someone who excelled in technical screening using an AI tool but couldn’t perform the job, highlighting the confusion AI is creating in discerning signal from noise. Gerry Crispin contended that “lower quality people have been picked because they can game the system forever” but that AI makes candidates’ efforts consistently high, destroying fluency as a valid signal.

Consequences of AI on Hiring Signals

Steven Rothberg summarized that the study showed the value of a well-written proposal declined due to AI making them look the same, forcing employers on platforms like freelancer.com to rely more heavily on the requested wage. Ethan Bloomfield challenged the idea of failure, suggesting that if candidates perform well using AI tools in tests, they should be enabled to use the same tools in the job, postulating that failure to perform is due to corporate governance prohibiting AI use.

Alternative Selection Methods and Ethical Considerations

Gerry Crispin emphasized the need for a different method to collect data and that companies are beginning to clearly communicate acceptable AI usage to candidates. Steven Rothberg mentioned a short one-week internship pilot used by a company as an extended assessment to facilitate authentication and skills verification. Ben Eubanks recommended their podcast “We Are Only Human,” featuring Dr. Carrie Miller, an AI ethicist, for further discussion on this topic.

The Problem of Volume and the Meltdown of Current Tools

Jeff Taylor introduced the analogy of companies seeking “giraffes” but being flooded by “camels” (high volume, potentially lower quality applicants) and characterized the current situation as a “meltdown of our current set of tools”. Jeff Taylor estimated that two-thirds of job seekers are now using AI to airbrush their resumes, which, combined with the ease of applying via platforms like Indeed and LinkedIn, exacerbates the volume problem.

Proposed Solution and Industry Trends

Jeff Taylor suggested moving toward direct sourcing and microtargeting in databases to find candidates before they adjust their resumes, believing the traditional job announcement is in trouble. Steven Rothberg noted that some employers, particularly in early career hiring, are reversing the trend of being school and major agnostic, focusing on specific institutions and GPAs as a form of “authentication,” which Matt Farrah and Jeff Taylor noted as a regression against skill-based hiring. Jeff Taylor predicted a shift in important skills within five years toward character and soft skills, such as problem-solving, creativity, and critical thinking.

High-Level Hiring and Extended Assessments

Tom Chevalier mentioned an observation about a company called Sully that requires “A players” capable of four months of output equaling one year, and a 60-day pilot for ex-founders, suggesting a trend towards intense, high-stakes assessments even for high-level roles.

Discussion on Talent and Hiring Predictors

Steven Rothberg initiated the discussion by passing the baton to Jason Gorham, Talon XI, who emphasized that talent needs to be utility players, able to pivot quickly, citing their own experience in staffing and e-commerce. Steven Rothberg built upon a point made earlier by Ben Eubanks, concurring that data like GPA and school are low predictors for hiring quality, and that the previous reliance on a small number of schools and majors was largely about efficiency. Ben Eubanks’ point also extended the logic that these low predictors apply to the value of referrals, silver medalists, alumni groups, and internal hires, which Steven Rothberg acknowledged, but expressed concern that relying solely on people known creates a small, walled community that prevents hiring the best people.

Impact of AI and the Need for Better Data Collection

Gerry Crispin highlighted that AI is helping to destroy the fluency of applications, making it difficult to differentiate between better candidates. Gerry Crispin observed that the reversion is often to less predictive mechanisms, such as hiring from a few preferred colleges, which is based on comfort level rather than performance. They stressed the need to focus on the top of the funnel to collect more effective data, possibly through the use of AI agents that can interview 250 people in a few days with fairness.

New Methodology for Data Collection using AI Agents

Tom Chevalier introduced their business called Tink, which involves AI agents as part of the solution for collecting good data. Tom Chevalier argued that historical job boards relied on a person’s desired job and location, and a resume, but that resumes are now compromised due to AI. They proposed that newer career concierge businesses use a conversational environment with AI to learn about a person’s interests, constraints, and desires, suggesting this data will be much richer because candidates will be open and honest at the beginning of the conversation. Tom Chevalier concluded that AI itself is not bad, but rather the methodologies used to apply it, and that AI agents can be awesome in collecting data.

Consensus on the Problem and Future Solutions

Matt Farrah, who helped start the roundtables, agreed that there is a consensus on the problem and the general form of the solution, which involves improving how they understand and work out a candidate’s value through better data. Matt Farrah mentioned that the fundamental issue isn’t the CV but how AI reinterprets and splices it, making it irrelevant. They emphasized that the industry is at the beginning of a long journey to discover the solution and then educate employers, citing an example of a client who mistakenly valued receiving a thousand applications.

Future Roundtable Topic and Meeting Logistics

Steven Rothberg thanked the participants for the discussion, and announced the next roundtable a month away, featuring Bill Boorman. Bill will discuss CPA+ (Cost Per Application Plus) or CPQA (Cost Per Quality Application) in the context of authentication, exploring if using a quality metric to compensate job boards helps employers authenticate candidates. Steven Rothberg suggested that paying job boards when an employer interviews someone signals that both the employer and candidate are real, preventing issues like bots or foreign actors.