- 31 May 2023
- Managing the Future of Work
Assessing AI-boosted background checks; fair-chance hiring
Joe Fuller: Can hiring be brought into line with the modern labor market? Many processes are either outmoded or susceptible to error and bias at scale without proper controls. Consider the background check. Verifying credentials and screening for events like criminal convictions has historically been a laborious, largely manual procedure. Automation and AI offer speed and accuracy, allowing employers to extend offers quickly and with confidence. But such technology can also reinforce long-standing patterns of discrimination. And increased government scrutiny of AI use in hiring processes promises to put the background check business under a microscope.
Welcome to the Managing the Future of Work podcast from Harvard Business School. I’m your host, Harvard Business School professor and nonresident senior fellow at the American Enterprise Institute, Joe Fuller. My guest today is Daniel Yanisse, co-founder and CEO of Checkr, the leading provider of background checks. Daniel’s experience as an engineer and entrepreneur convinced him that this key element of the hiring process was ripe for disruption. Early in Checkr’s history, Daniel and his colleagues became aware of just how massive the population of those with a criminal conviction is. A combination of empathy and pragmatism led them to take up the cause of second-chance, or fair-chance, hiring—extending employment opportunities to those individuals with criminal records. Checkr itself became a second-chance employer. We’ll talk about recent advances in background checking, how to guard against the shortcomings of AI, and the business case for fair-chance hiring. We’ll also discuss the policy landscape and how to tailor employment services to make the market for low-wage and frontline workers fairer and more efficient. Daniel, welcome to the Managing the Future of Work podcast.
Daniel Yanisse: Thanks, Joe. Thanks for having me.
Fuller: So, Daniel, tell us a little about yourself and how it is you came to found Checkr.
Yanisse: Yes. So my co-founder and best friend, Jonathan, and I in 2014, we were both working as engineers in an on-demand delivery start-up in the Bay Area. And as we were building the business with our co-workers, one of the challenges was hiring and onboarding delivery drivers fast enough. And as part of this vetting process, we identified the background check step as the bottleneck. It was taking a lot of time, we were losing candidates on the way. And by looking at the different options back then, we couldn’t find an automated solution that met the needs of our last business. And so we decided to explore this as an idea for a start-up to build something better, more streamlined, faster, a better product for the needs we had as a growing business.
Fuller: What were causing the delays? Just inaccessibility of the data? Or it seems like a problem others would’ve bumped into. So there must have been some real obstacles to solving it if it had not already been addressed.
Yanisse: Yeah, I think the pain was more acute for us, because we were creating a marketplace with thousands and thousands of jobs and drivers. So we had to really hire thousands of people. So it’s quite different from the regular HR process where you interview people, and you hire them one by one. So we had much more need for automation and speed. The traditional background check process is pretty slow. It requires paperwork and emails and PDFs. So there’s also, for the candidate side of the equation, there’s quite a lot of friction, and it takes time. So I think we try to really make it very easy, with one click for drivers to apply and apply to the job and get onboarded. And then also was trying to streamline the fulfillment of background checks that back then could take weeks sometimes.
Fuller: Sounds like a classic market, with pain points at both ends that nobody had addressed. So tell us about the evolution of the business.
Yanisse: Yes. So we didn’t know anything about the space, but we knew that was a pain point for the business we were working at. And then we did some research and realized that we talked to early companies back then, like DoorDash, Uber, Lyft, who were also hiring delivery drivers at scale, and they had the same problem. So there was enough validation for us to leave our jobs and focus on this full-time as our first start-up. And so we then started researching how to get access to background check consumer data, how to make sure the background checks are compliant, and how to really fulfill them. And every step of the way we learned the legacy and traditional way of doing it. And then as software engineers, we immediately looked at solutions to automate the process. We are also focused on building an API first as a product—so application program interface. So really to allow our customers to build the background check process directly in their app, because that was not available at the time. And so we invented the first API to run background checks.
Fuller: So it sounds like a background check requires expertise in mining data from completely discreet and separate data sources, which should have everything from different levels of accuracy and accessibility. And also, presumably on the demand side, your customers want to put greater or less emphasis on some elements of the background check, want the things as simple as the data formatted differently. Tell us about the various scenarios, if you will, when someone goes through a background check. Where are the really key data, which of the data that might cause a company to make a mistake of either type—either proceeding with a candidate that they ultimately don’t want or excluding a candidate that ultimately would be a productive worker.
Yanisse: Yes, to put together a background check, first you have to decide as a customer, an employer, what’s interesting to you. There’s many types of checks and verifications that can be offered. The most common are to doing a criminal background verification, potentially a DMV license and driving record verification if the person’s driving. But you can also verify identity, employment history. You can do drug screening, credit checks. There’s a whole suite of information that can be verified from the candidates. And then, yes, we have to go as a background check company to mine data from thousands of different data sources. Many of them are public records. Sometimes we need to reach out to credit bureaus to previous employers to compile this information and deliver it to the employer, to our customer, to then make a decision to move forward or not with a candidate. I think in the past, the challenge was it’s a lot of data to uncover, especially the criminal data. The legal data comes directly from the justice system. So it’s legal language, it’s very unstructured data, it’s different in every single county in the U.S. So just being able to read even a 20-page PDF with background information is quite challenging. And so we applied software to make this information more structured, clean it up, package it in an easier way for customers so that they can then apply their different rules in order to make the right decision for their business.
Fuller: That’s an amazing amount of complexity. So how has this evolved? How have background checks evolved in recent times? Obviously, you’re describing highly personal information. There’s been such a focus on rights to privacy and data controls, obviously, that are even more advanced in foreign jurisdictions. How has it evolved since you started? And also, what kind of effects did Covid have on background checks?
Yanisse: To do a background check for employment purposes, it’s a regulated process in the U.S. It’s regulated by the Fair Credit Reporting Act. So we focused on making those steps very easy and seamless for candidates so that they can in one click understand what information they agree to share and also get a copy of that report at the same time as the employer gets it. Then we’ve streamlined the speed and accuracy of the fulfillment with technology. So we’ve reduced the time from weeks to hours to run a background check, which is great for the candidate to get a job faster and get to income. It’s also great for businesses to be able to compete and hire quickly in a tough labor market.
Fuller: Daniel, I’ve noticed a phrase as it relates to background checks: “continuous background check.” Could you elaborate on what that is, and is that an important development that we ought to be keeping our eyes on?
Yanisse: With the automation and technology, it’s relatively easy and cost effective to have an ongoing verification and to flag discrepancies and changes real time. It’s also very helpful when you are working remotely, or you’re not in the office every day. For example, if you have a delivery driver, and they get in a bad accident or a DUI or get arrested, that’s something that you might see if you’re in person but you might miss if it’s an online work situation. It’s an easy way to stay in compliance and have ongoing updates rather than waiting years to having that surprise. There’s, again, transparency and disclosure to the worker.
Fuller: You know, that’s particularly interesting to me, because we’re seeing movements by companies or confederations of small and medium companies to try to get solutions for workers to avoid problems or events. It could be an eviction. It could be an uncontested lien. It could be a criminal conviction. So giving employers more ability to get a grasp of indicators that might lead to a problem where, whether it’s internal resources at the company or external resources—state and local agencies, not-for-profits—that could be mobilized to help interdict otherwise what might become a dangerous cycle.
Yanisse: And then I would say lately we’ve been focused a lot on fairness, on bias. The way background information was used in the past created a lot of barriers and friction for many people with some background records to get access to employment. We care a lot about enabling and helping people to move on to get second chances. And we believe there’s also a lot of benefits for businesses to have a more nuanced decision-making process and not over reject thousands of very, quite qualified candidates. So we put quite a lot of effort into this filtering, education, risk assessments part of the equation to allow more candidates to get good jobs and better jobs, and also to allow our customers to hire more talent in order to grow their business.
Fuller: I first became aware of Checkr talking to large companies about their second-chance initiatives. And quite frankly, you just kept coming up. And I finally said, “Well, I’d need to know more about Checkr.” In some ways I think about such initiatives as really first-chance initiatives, because I don’t think a lot of the people who have become in the custody of the criminal justice system really had much of a chance at all. But putting that thought aside, how is it that you became aware of the centrality of people with criminal convictions being frozen out of the labor market or having very limited choices? And when you became aware of that, how did you set about understanding what it would take to reassure employers that extending such a chance to someone would be not just a worthy thing to do, but a smart business decision?
Yanisse: I agree with you; the majority of people who end up in the justice system didn’t have a first chance early on to have access to the opportunities and educations, which leads many people to crime, potentially. Initially when we started, as we started to run our first hundreds of background checks, hearing the personal stories of many job applicants who were sharing the struggles, the mistakes they’ve made in the past, but also the strong motivation and desire they had to turn their life around and how important that next job was for them. And so we started building empathy with the consumers who were overly rejected again and again despite potentially sometimes having relatively minor crimes or very long time ago. And so we started to better understand the justice system, the high level of incarceration in the U.S., and also the lack of training and opportunities. It’s very hard to reenter the workforce for those millions of Americans who had to go through the justice system. We went on a few different prison visits and prison education programs to help inmates train and get ready for their jobs as they get out of prison. And we met amazing individuals, extremely talented people with so much emotional intelligence and motivation, that we were really impressed and thought, “Hey, those people are amazing. It would be great to hire them in our company at Checkr and help more people get successful in jobs.” So that’s really what sparked the interest. I think for me, it also connected with my background as an immigrant. My parents are coming from quite poor countries. I was lucky to be born in France and then move to the U.S. But I was able to connect with the struggle of not having the same opportunities as the local population and having to work harder to get there. And so what did we then, initially, it was quite counterintuitive. We were not sure that, as a background check company, we could change behaviors. We were not sure businesses were really open to second chances, but we had to try, since that was so core to what we do. And so we started with our own business, our own company. We started by hiring one, two, three amazing people from those programs that we met in prison as they were getting released and got some great candidates. We helped them train and get working in a tech company with us, and that was very successful. And once we started to see traction and success, then that was the beginning of our big focus on fair chance, internally, and then really advocating for it and helping customers and other businesses to get the benefits themselves.
Fuller: We’ve done research about the hiring process and the use of artificial intelligence in applicant tracking systems that have the inadvertent effect of walling off or hiding a large number of workers from an employer’s consideration—what we call our “Hidden Workers” framework. Here it is, it’s spring 2023. Probably because of the giant shadow being cast by OpenAI and ChatGPT, especially version 4, the whole question of artificial intelligence and its efficacy and its fairness has come to the fore. Certain jurisdictions are now beginning to place a mandate on employers to demonstrate that their AI and hiring isn’t somehow prejudiced or biased. What are your thoughts about that? How does it affect both what you offer your clients, but also, how does it affect your own artificial intelligence, your own decision-making rules, as you gather this very sensitive data from all these different sources that are integrated into a candidate profile?
Yanisse: We’ve been using AI, the traditional AI, for a few years now in order to solve very specific problems as it relates to the background checks. So we’ve used it, in conjunction with humans, to improve accuracy. AI is very good at finding patterns and similarities and differences. And so, for us, it’s very helpful in terms of making sure that some background information belongs to the right person. We use it to be able to really improve that matching accuracy. And we feed it with data trained by experts, and we check it and audit it by data trained with human experts. AI can be an augmentation of our productivity from employees. And the second use case is, as I explained, is the 3,000 counties, a very fragmented criminal and data information on the legal side. And so we use AI to classify and categorize information so that it’s more readable by our customers. And I think when it can become bias, it’s when the AI is used to make a decision on employment to say, “Should I hire this worker or not? Is this worker qualified or not?” I think that’s the use cases where you have to be more careful. And at the end of the day, it’s important for the recruiter or the hiring manager to have their independent assessment, even if potentially they’ve been helped by AI. I think for a hiring decision, it’s important to have a human review. And definitely in our space for background checks, for example, a human has to do an individualized assessment before making a decision, and our product enables this.
Fuller: So it sounds like not only you were developing best practices inside the four walls or the four virtual walls of Checkr, but you’re also proselytizing those with your customers and helping them build compatible approaches that take the power of the data you’re offering, but also insulate them from these types of algorithmic errors or biases that we’re also concerned about.
Yanisse: And I think there’s many positive use cases. For example, if you look at the status quo of background checks, maybe five, 10 years ago, the decisions were very binary. Most people with a flag or discrepancy were rejected. Most people with no information were moving forward. And so we really used it to reduce the amount of people who are rejected and making sure that we’re able to show a 360[-degree] view of the candidates. Yes, maybe there’s some information or flags or information that’s not perfect, but we also complement it with the positive story of the candidate writing themselves what they learned from the experience, what they’ve done to improve, to rehabilitate. And so that leads to millions of candidates who used to be rejected in the past now moving forward, getting to that interview, and getting to job opportunities. So I think, again, it all depends on how you decide to use the technology, for what purpose, as a business.
Fuller: Let’s pivot a little bit to your commercial customers, the companies that rely on your technology to do the background checks. And it sounds like you can do them very, very quickly, which would certainly help in a job market, where a number of people effectively expect an offer at the end of their last interview, on their first day of visiting a company—certainly people with good data skills and good social skills. What’s the scope of your customer base today, and is it from all sorts of positions or still skewed more toward entry-level positions, lower-paying positions, middle-skills jobs?
Yanisse: The background check process is very much the same for any industry. Whether you are a delivery driver or a plane pilot or a doctor or a professor, there’s a similar type of record information that’s reviewed. There are some additional potential verifications per industry. But that allowed us to scale the product to all industries and all business sizes. After having a strong start in the tech community with tech companies, we then realized lots of demand from small and medium-sized businesses who want to hire a few people, have a simple product to move fast and get an offer out quickly. Because you’re right, these days, the expectation for any job is the same day or a few days, and it used to be a process that take weeks. And so now we have tens of thousands of customers from very small businesses of two employees all the way to some of the largest Fortune 500 with millions of workers.
Fuller: Well, that’s certainly a classic story of entrepreneurship, particularly in a software-as-a-service business, where, as you get bigger, you can afford to reach smaller companies and you can price offers that they can afford to buy, which is important. Could you kind of walk us through, if I were a prospective customer, but I’m the CFO, I’m the Chief Financial Officer. I want you to explain to me why this is a good business decision. I’ll donate personal funds to a charity if they help offenders or other hidden workers, but you have to prove it to me in dollars and cents. What are the highlights of that pitch?
Yanisse: I would say that the fairness aspect is an additional value add, but the main reason customers want to come to us is because they need to have an efficient hiring process that is cost-effective and allows them to build their business. So the value prop for us is, number one, the speed. The speed of hiring talent is very important, because the longer it takes, the more candidates have to go through, the more competition there is. So everyday matters, and that has an impact on, overall, the total cost to hire, which these days is thousands of dollars to hire one worker. Secondarily, managing the process can be relatively manual. And so we automate this process. So compared to the hourly wages of having experts, workers doing this job in-house, outsourcing it and automating it to a software solution is actually very cost-effective from a financial standpoint.
Fuller: It’s interesting, because that’s a little bit of an inversion of history. Our analysis has shown consistently that the drive by companies to make that hiring process as efficient as possible has historically led to just excluding any candidate that’s remotely problematic and that the model is essentially go and find somebody who’s willing to apply for this job, who has something that’s very close to this job already. So my pre-qualification is that the worker is doing either the one job down in a competitive organization or the exact same job in another employer, but never to expand the pool. And it sounds like you’re convincing companies to actually think about efficiency differently.
Yanisse: Yeah, I mean I think it’s a bit counterintuitive. I think that there’s two things. Number one—and I fell into that trap myself initially when I hired the first people in my company. I was thinking, we need to hire the people with the best education, with the best company experience. You want that perfect resume that that’s really exciting. And then as you build recruiting experience, you realize that some of the best candidates and best employees are the people who might not have the perfect pedigree. Maybe they don’t have the best schools, maybe they don’t work for your competitor. Actually, I learned that, if you give a chance to someone who is more driven, wants to prove themselves, maybe has a chip on their shoulder, maybe doesn’t have the perfect background, actually those are going to be your best coworkers, who are extremely motivated to win in the business. And then the second dimension is, there’s just so much competition for labor, even today, even with the current macro, there are so many skilled jobs where there’s just not enough qualified talent in the U.S. If you look at the healthcare industry, we don’t have enough nurses. Even in engineering, we don’t have enough programmers. We don’t have enough truck drivers in the U.S. There’s just a big shortage of labor. And so, if you want to compete as a business, you have to expand your pool, maybe change some of your criteria, give some chances to candidates to be able to hire, and actually those could become your best workers. And we’ve seen it in our company, whether it’s hiring, giving second chances or not hiring that perfect resume, it leads to the best performance, and it’s great.
Fuller: Well, certainly, in the absence of a remotely sensible immigration policy in the United States, everything we can do to improve the rate of good matches of people with jobs, but also to remove barriers that have contributed to America’s appallingly low workforce participation rate, I think, should be a focus not just of corporate activity and innovation like Checkr’s brought to those issues, but also of public policy debate that we have to understand what’s walling people off from the workforce. It seems as if you’ve grown with your clients and extended your operations internationally. Could we talk a little bit about that? I have to imagine that that was a decision that was pretty daunting.
Yanisse: Expanding internationally, I think, for any business is a huge project. We are really focused on the U.S. market first and more recently, because of the customer demands, we have a lot of medium and large U.S.-based customers who also have international operations. They have remote offices in different countries. We’ve been ready to be thoughtful on how fast we expand, because not only it’s hard to go in other countries because of language difference, GDPR [General Data Protection Regulation], and different regulations, but for us, it’s a different product we have to build in every country. The data sources are different, so we have to build different data, the pipeline in every country. The employment laws and regulations are different, and even the culture and the hiring process is different in different countries. We have launched a product that covers over 200 countries. We are mostly working with U.S.-based customers who have international operations.
Fuller: When you think about what you’re learning about these other markets, what do you think are the commonalities, the issues, or problems that a solution like Checker solves that are really applicable everywhere, versus what are the biggest differences that really distinguish certain types of labor markets that once again influence both how you serve your customers, but also almost how candidates are judged?
Yanisse: Yes, I think things that are similar are really all of the workforce trends that we’ve been seeing. Especially since Covid, the whole world had to change and move work a lot more online and remotely. That’s a huge trend. That means that different tools we use for interviewing, recruiting, working together, have to move from in-person and paper and email to better solutions. We have Zoom for video conferencing and then a lot of the HR processes have to move online. So I think that’s a trend that’s global and that drives demand for HR automation solutions like what Checker provides. And then, you’re right, every country is different in terms of compliance rules, requirements, type of licenses, and verifications that are needed. But in general, there are the pain points of making sure the person I am working with is the person who says who they are, because maybe we’ve never met in person. We want to verify and build that trust, since we don’t build trust in person, maybe, as much within the interview process, verifying employment history, or some references or identity are pretty common ways to build that trust to then start that remote work relationship.
Fuller: Well, I was earlier this morning with one of the leading scientists and professors here at Harvard in the artificial intelligence and deep learning space, and from what he was saying about the capacity to create deep fakes, whether they’re visual, verbal, completely fictional identities online. I imagine that this is going to become a more challenging environment and probably one that’s even more valuable for your customers to rely on people of your deep expertise and competence. Why don’t we go over to the public sector, because you’ve got a lot of exposure now with different regimes about privacy, about control of data, about access to data, the data that should be in the public record, and that that is more problematic, data you can strip out of unofficial sources. How do you think this market for this intimate important, potentially life-changing data—because it is going to influence people’s career paths, incomes, what their work history looks like—as you think about the domain of public policy, what do you think are the important questions that public policy makers at least have to be considering?
Yanisse: I think there’s lots of problems and no easy solutions. But a few things that come to mind is, I think, first, we have to embrace that a lot of data is available on the internet. So I think it’s about, it’s good to create regulations and rules that are modern to then help people make sure that they have control over the data, that what they’re sharing is not going to be used against them in the future, that people have the rights to make mistakes or they will be forgotten or having control. I think those are some of the hard challenges that we all have to work on, whether it’s public or businesses. But if we create frameworks and tools and technologies to produce people’s data, to educate people on what you can do and maybe what’s not a good idea to do, and then maybe use some of those regulations to—like CCPA [California Consumer Privacy Act] in California or GDPR in Europe—I think those are good regulations to have consent, transparency, disclosure, and some agency from consumers about their personal data, especially the private data and information that businesses might have on them. So I think, yeah, there’s lots of opportunity and has to be a partnership between the latest technology and businesses and then the public sectors and lawmakers to keep up with this extremely fast change.
Fuller: Are there any industries or industry groups that are on the cutting edge of putting technology to work to help expand the labor market and minimize mismatches?
Yanisse: I come back to what we are very focused on, which is fair chance and helping getting access to second chances. I think in order to reopen those opportunities, we have to start with the job descriptions, the job marketplaces, job postings, to say, “Hey, we are welcoming people who don’t have a perfect background. We are an equal opportunity employer. We are a fair-chance employer. We will evaluate all applications.” We used to have the criminal conviction box, right? Like, “If you have ever been arrested or convicted, check this box,” and then automatically, you get rejected. That’s changing this. And I think what companies like Indeed or LinkedIn, for example, are doing—and we have been partnering with them—is they’re starting to build in their project, which are the biggest job sites in the world, start to make it easier for employers to say, “I’m open. I will accept fair-chance applications,” and for people with the records, being able to then find some of those employers. I think those are some of the exciting innovations that open up the job market that have happened over the last year.
Fuller: Certainly, one of the things we were impressed by at various times when we’ve published papers that some of those vendors of the type you mentioned might have taken as criticisms, we found them actually to be quite open to the idea that there are issues with the way the labor market works now, that they want to work to solve, and in fact, that they would like more effort to go behind third parties—in this case, it might be Checkr or Harvard Business School Managing the Future of Work Project—to raise the right issues, so that they can have a different conversation with their customers. So, well, tell me just a little bit, as we come to the conclusion of our conversation. What do you anticipate the future holds for Checkr? What are your ambitions? And when we check in with you in two or three years, what will you be talking about?
Yanisse: Yeah, so we are very excited to continue to champion this fair-chance movement, bring the benefits of giving second chances to hundreds, thousands, of other businesses. Beyond background checks, we’re just very excited about the opportunities in the workforce and the remote workforce to build great products, to remove that friction, to bring more fairness in the process. We’re working on future innovations around payment of workers, payment transparency, allowing instant pay payments. We have a product we’re launching now called “Checkr Pay” to allow gig workers, hourly workers, to get paid instantly without any fees being taken out of their paycheck. We’re working on streamlining that application process and helping customers hire the same day. So continuing to compress that time and improve the client experience.
Fuller: Well, Daniel, those are great directions. I’m particularly excited to hear about your payment product, because, of course, at the lower end of the workforce, so many workers are unbanked. And if you can create products that make it affordable for them to be banked and also begin to start developing a banking history, which will help their credit rating and many other things, that will just feed a virtuous cycle. Well, Daniel, so thanks so much for joining us. We’ll look forward to keeping track of the developments at Checkr. And thank you for all your good and important work as it relates to second-chance initiatives.
Yanisse: Thank you, Joe. It was great to be here.
Fuller: We hope you enjoy the Managing the Future of Work podcast. If you haven’t already, please subscribe and rate the show wherever you get your podcasts. You can find out more about the Managing the Future of Work Project at our website hbs.edu/managingthefutureofwork. While you’re there, sign up for our newsletter.