General partner at Benchmark, Benchmark
Streamlining hiring with AI: Benchmark backs AI start-up Mercor
Brendan Foody, Mercor co-founder and CEO, and Bill Gurley, Benchmark general partner, join 'Squawk Box' to discuss using AI to streamline hiring, impact of AI technology on company HR departments, and more.
About Bill Gurley
Bill Gurley, a general partner at Benchmark, has been promoting his new book “Runnin’ Down a Dream” through a series of public appearances, including a TED Talk, a fireside chat with Malcolm Gladwell at NYU Stern, and interviews on multiple podcasts. In these appearances, he has argued that career excellence is driven by “fascination” rather than passion, and that obsessive, continuous learning is the key to long-term fulfillment. He has also discussed the importance of building strong peer groups early in one’s career and avoiding “boldness regrets” by taking risks. Gurley has also spoken extensively about regulatory capture, particularly in the technology and AI sectors. He has argued that regulation often benefits incumbent companies rather than promoting competition, citing the Telecommunications Act of 1996 and FDA approval of COVID-19 saliva tests as examples. He has expressed concern that leading AI companies are lobbying for regulation in ways that could stifle innovation, and has warned that the U.S. risks building a “cage” around its own AI industry while China advances. Gurley has also criticized the U.S. financial system for not adopting instant digital transfers, and has expressed interest in stablecoins as an alternative.
Transcript (15 segments)
✨ AI-enhanced transcript with speaker attribution
H
Host0:03
Meantime, talked about this before the break. An AI hiring platform backed by investors including Jack Dorsey, Peter Thiel, Larry Summers, trying to fix some flaws in the hiring process, including reducing human bias. First on CNBC, the company's co-founder and CEO, Brendan Foody. Benchmark general partner, great to see you both. Brendan, explain this. Right before the break I teased this: I said, is it possible AI can hire people better than a human can?
B
Brendan Foody0:45
Yeah. So thanks so much for having us on, and I think it certainly is. The premise is there's a legacy services industry where people will manually review resumes, manually conduct interviews, and manually match people with roles. All of these processes can be done far more effectively with our lens. We update the resume, read through work experiences, projects, and papers written, and ask dynamic follow-up questions with a model that can predict what jobs they'll perform well at and match them with opportunities.
H
Host1:23
So this is really more of a matching platform rather than... I mean, can an employer use this directly, though?
B
Brendan Foody1:30
Exactly. Yeah. Also we facilitate all hiring compliance and payroll associated with this. People come to us. They say they want to hire a software engineer or lawyer for a particular job they need to do, and then they can press a button and hire that person immediately with confidence in how they'll perform on the job.
H
Host1:53
And, Bill, you know, people don't love HR departments. They don't. Hate to say that. I love our HR department. What do you think happens to HR departments if this succeeds?
B
Bill Gurley2:06
I think you have this situation in every single functional area of a company right now where we're trying to figure out where AI can add value and do things arguably better than a human. I don't know that anyone out there loves extended interview processes; those tend to touch multiple people in the organization well beyond HR. If we have the ability to accelerate that and lower the error rate where you hire people that don't work out, those things are super powerful. I think the big thing all the venture capitalists are trying to figure out is where does AI have the ability to have the highest impact right out of the gate, and when we came upon Brendan and Mercor and what they were doing, we got super excited and believe they're one of those types of companies.
H
Host2:59
Now, when you look at this investment, for example, I think a big question for all investors now, given just how AI is moving so fast, how do you think about the moat around Brendan's business and around other AI businesses? Because at some point, if, for example, OpenAI is what's powering part of this and becomes so powerful and so successful, won't it be able to actually re-create all of these companies?
B
Bill Gurley3:26
Well, I think that's a great question, and you probably know that I spent the majority of my career seeking out companies that were two-sided marketplaces and had network effects where the more time you have in market, the more customers you have, the better the product gets. I might turn it over to Brendan to explain exactly how that happens inside of Mercor. One of the reasons I'm here today and the reason Brendan and I hit it off so much is precisely because of the question you're asking.
B
Brendan Foody4:00
Of course. Obviously there's a network associated with a marketplace. We have people that will join the platform as candidates, taking interviews and looking for jobs, and companies hiring on the other side. But perhaps even more interestingly, as we're collecting data on every bonus an employer issues, every raise they'll give to someone, the reasons they'll dismiss someone from a job, and training models on how people will perform on their next opportunity. We know people at top labs well and bet the models will be incredibly capable but not able to solve the problem of effectively predicting human performance. With enough data and a usage data flywheel, we'll be able to build the best model in the world to predict human ability.
H
Host4:46
Therefore, where do you get the data from?
B
Brendan Foody4:49
It's from our all-in platform. Because we're the ones processing all of the payouts and billing to customers and candidates, we're able to see who's getting raises for what reasons, dismissed for what reasons. It's useful to customers on an individual level. They want to make better predictions about who's going to perform well on the kinds of opportunities that they have, so that we can set them up for success. That's one of the most powerful parts of the platform.
H
Host5:23
Dare I ask, do you worry, or maybe it's an opportunity to sell to them eventually, a Workday or a company like Ripple or somebody like that? Clearly lots of employer and employment data.
B
Brendan Foody5:36
Yes. A great question. I'm constantly amazed at how slow those companies are moving, and it's clear that nonetheless we should still be thoughtful about it. I see it manifesting itself largely through partnerships and collaborations with them. But at the end of the day, I think whoever builds and ships great technology extremely quickly is going to win the market, and I really believe we can do that.
H
Host6:06
Bill, as you know, so many in the Valley, we discussed it earlier, are trying to figure out where the opportunity lies in AI, specifically on the software end. I think we understand the hardware side of it, at least chips in that respect. But when you look at businesses, how much of it is the data piece? Here's a marketplace with a two-sided sort of social effect, but how much of it is folks who have proprietary data they can marry to an OpenAI, for example?
B
Bill Gurley6:39
If you have a problem that is deterministic, in some ways the pull-up driving is that way, you can use traditional AI models, allow them to run competitions with each other and get to really amazing places. A lot of the business process problems aren't that way. I think the question then becomes, how can LLMs be most effective in things that involve language? You know, I'm really excited by the new voice products in OpenAI. I think that voice could become a new frontier, and you have to really get into the market with these solutions, see where they have an impact fast. Customer service is one of those, we believe strongly, and interviewing and HR recruiting can be one of those, and I think that's what the whole world is starting to see.