Decoding Talent: How AI and Big Data Can Solve Your Company's People Puzzle

By Eric Sydell, Mike Hudy and Michael Ashley

Harness the power of artificial intelligence in hiring

The typical hiring process is fraught with complexity, inefficiency, and bias and often shuts out the most talented candidates. Decoding Talent: How AI and Big Data Can Solve Your Company’s People Puzzle makes the case for using complex advanced technologies to move past these problems toward effortless optimal candidate decisions.

AI experts Eric Sydell, Mike Hudy, and Michael Ashley explain why the traditional resume-based process is out of date, why hiring is difficult, the cost of bad people decisions, how bias interferes in hiring practices, and how AI can address these problems.

Decoding Talent reveals that using AI in hiring doesn’t require your human resource professionals to unlearn and relearn their craft; rather, machine learning can complement their skills by consolidating and analyzing data to recommend actions. Imagine a world in which you didn’t have to wonder:

• Who is the best candidate for the job?
• What is the return on investment of our hiring process?
• Is our hiring process fair and equitable?
• Is our human talent deployed optimally across our organization?
• What can human resources do to better drive business outcomes for our company?
• Is our candidate experience adding value to our brand?

Incorporating scientifically based hiring can make this world a reality, benefiting both your company and the candidates for hire.

• Language: English
• Publisher: Greenleaf Book Group
• Release date: Mar 15, 2022
• ISBN: 9781639080106

Book preview

INTRODUCTION

Subsumed as they are by the hustle and bustle of our daily lives, all the seismic technological changes afoot can be hard to see. But since the advent of the artificial intelligence (AI) technique called deep learning near the start of this millennium, machines have been developing truly stunning abilities to make sense of our world. Deep learning is essentially a data analysis technique, a paradigm-shifting statistical tool allowing us to process and make sense of all kinds of data faster and more thoroughly. Some of the amazing (and entertaining) things deep learning has produced include:

•Cars that can nearly drive themselves

•Virtual assistants that understand increasingly more of what we ask them

•Software that can read medical scans as well as or more accurately than trained physicians

•A stunningly de-aged Robert De Niro in the movie The Irishman

•Illicit websites that can instantly write a term paper for you

•Multiple vaccines for COVID-19 in less than a year, when the previous world record was four years for the mumps vaccine

But while this technology is amazing and sometimes frightening, tremendous structural barriers come with trying to roll it out quickly and widely. These include the following:

•Big data is required but often hard to obtain. Tremendous amounts of data are recorded and stored, but the data is not always high quality.

•Laws, regulations, and guidelines are woefully slow to evolve. It takes time for lawmakers to reach consensus on how to handle new technologies, but AI is advancing at a rate far beyond our ability to civilize it.

•Access to data is often restricted, intentionally or unintentionally. Many times, organizations house data for safekeeping without any reasonable way to access and analyze it. They store the data, but we can’t see the data. It’s sort of like keeping your fine jewelry in a safe deposit box—a good way to keep it safe but not very convenient to access.

•Concerns related to privacy and bias, while addressable, lead to confusion and loss of momentum. It is a massive under-statement to say that AI is powerful, and with that power comes great potential for benefit and harm.

Do these challenges mean we should oppose AI and work to limit its adoption? We don’t think so, because it would cede technological leadership to those who do not have such scruples (and there are many). Instead, we elect to take a strong stance on AI: It should be harnessed for the benefit of humanity. This principle means individuals should benefit, not just corporate entities. If AI is used to continually make organizations more efficient, it will eventually do so at the expense of the people who make up organizations. It will ultimately dehumanize us.

AI does not cause bias—it reveals it. It’s the humans using AI incorrectly who cause or exacerbate existing biases. In many ways we are entering a golden age of understanding and identifying bias. Big data and powerful algorithms are uncovering it in places we never knew it existed. As difficult as it might be to face, AI being able to locate more bias in all the different places it lurks is a positive development—because it also comes with the knowledge to see bias and address it.

Many AI vendors talk about trusted AI and bias-free AI, but these phrases involve some wishful thinking. Any tool that is in any way opaque in how it operates—which a lot of AI technology is—can never be fully trusted or left to its own devices. Nor can we ever assume it is bias-free, even if an analysis at one point showed it was. As more data arrives and algorithms evolve, bias will be reintroduced. The solution is to continually monitor the structure and functioning of AI and, when bias is found, to fix it.

Even so, the biggest barrier to the realization of the potential of deep learning is us. We humans struggle to envision and comprehend possible futures, especially ones involving complex advanced technologies. We must see concrete examples and results for those futures to feel real and immediate to us. And we are fearful—fearful of the unknown, fearful of change. Of course, this natural intransigence allows us the time to civilize and tame new technologies, thus limiting the potential for negative and often unforeseen downsides. But it does so only as a side effect. A much better path involves collectively understanding new capabilities and designing our future around them with our eyes wide open to both the benefits and hazards.

Easier said than done.

General Motors (GM) employs some very smart engineers. Many of these individuals work to refine and upgrade GM’s power trains every year—always analyzing, tweaking, improving. And as a result, each year, GM’s engines become more powerful and efficient. Within the microcosm of internal combustion engine (ICE) development, this continual improvement is awe-inspiring.

But if you stand back and look at the overall automotive landscape, you will notice electric vehicles have taken off in recent years. Tesla was founded in 2003, and as of this writing, no major legacy automaker, GM included, has a vehicle that comes close to rivaling Tesla’s sales numbers. At the same time, nearly all major manufacturers are now gearing up to produce electric vehicles themselves. Why has it taken so long? Why didn’t GM mobilize its massive talent base to crush Tesla before it could even get off the ground?

Because ICE engineers can’t just stop what they are doing and learn how to make electric power trains. GM has legacy product lines to develop and support, and it can’t just pivot its attention elsewhere. Plus, engineers who spent years in school learning how to build traditional automobiles would be hard pressed to just throw out what they know about fuel-burning power plants and start over with new electric propulsion technology. The vast majority of people do not want to unlearn and relearn, to start over at square one, to deliberately turn away the expertise that made them successful in favor of some unknown and possibly risky way forward.

And yet, as of 2021, GM has finally shown real progress toward taking on Tesla and making an earnest commitment to join the electric vehicle movement. The company has reportedly invested billions in creating its Ultium platform. No vehicles have yet been released, but the technology certainly seems promising. While we don’t have hard data to support this assertion, it’s a good bet that most of the engineers creating this electric, battery-powered future for GM probably are not legacy ICE engineers who re-specialized but, rather, younger, newer engineers who studied electric power trains in school. Ultium likely required GM to create altogether new functions, structures, and teams to generate a viable electric vehicle initiative, rather than reallocating existing resources.

And so it goes with talent management. Making accurate decisions about people requires being able to predict their future behavior, one of the hardest problems in the known universe. Psychologists, industrial and organizational psychologists in particular, have been studying how human characteristics correlate with performance on the job for decades. They have identified many traits that are predictive of success, as well as a lot of variation across different job types. Need to hire a maintenance tech? Look for reliability, technical acumen, and possibly team orientation. Need to hire a physicist to find the next undiscovered particle? Hire the smartest person you can find, even if that person is a jerk.

But hiring is not accomplished by scientists. It is done by recruiters and hiring managers who have not spent years reading esoteric technical reports, conducting intricate and tedious research, and learning how to use and apply advanced statistical techniques. Organizational hiring is done to fill seats as quickly as possible and is measured on metrics like time and cost-to-fill. It is enabled by various platforms and tools that must be integrated and learned. In short, it is kind of a mess. As a result, scientifically based hiring is often left out of the mix, and those hired are successful because they get along well with their interviewers or have an existing connection to the organization. Expediency rules the day, not predictive, validated, fair science.

Can this morass of complexity and human bias be tamed? Can employees who have worked their entire professional lives in service of a traditional process with legacy tools be expected to fundamentally rethink their methods and learn confusing new technologies—all while working day jobs and essentially rendering their own resumes obsolete?

We all know the answer to this question is a resounding no.

But what if they don’t have to? The message of this book is that there is another way—a solution for reframing and rebuilding talent management from the ground up that doesn’t require your extended team of recruiters and hiring managers to unlearn and relearn. Instead, we recommend and describe how to change the foundation of your hiring and talent management work from process (think complex applicant tracking systems with many add-on tools and packages) to insight. By consolidating data and using machine learning to analyze it and recommend actions, you would be able to move past the morass toward easy, effortless, and optimal candidate decisions.

Just imagine a talent management world in which you never have to wonder about the following:

•Who is the right candidate for a particular job?

•What is the return on investment of our hiring process?

•Is our process fair and equitable for all parties?

•Do we have an existing employee who would be successful in a different key job?

•Is our candidate experience adding value to our brand?

•Is our human talent deployed optimally across our organization?

•What can human resources (HR) do to better drive business outcomes for our company?

•How can we lead our industry by harnessing and optimally using our human talent?

All of these things are possible with current technology and are part of the vision we put forth in Decoding Talent. This book explains key ideas and actions necessary to make this future a reality for your organization, including:

•Beginning to consider data as the foundation of good decision-making. You don’t need to learn how to write AI syntax in Python, but you do need to ensure your organization is capturing the data AI requires to be effective.

•Understanding that while it is reasonable to capture all the data you can, not all data is equal in terms of usefulness (as illustrated in Figure 6.2 in Chapter 6).

•Getting started with the six-step process for maximizing the benefits of AI and big data. (We talk more about these steps in Chapter 6.) This process is simple and intuitive, not something requiring an advanced degree or coding skills to comprehend.

What you most need to know to get started on your path is that most modern organizations do a poor job of capturing the data necessary to make AI effective. But the key to making AI effective goes beyond just capturing and storing data to capturing the data and making it available to analyze. As we describe later in the book, these are often two different things. It has always been common practice to store data just in case it is needed for a legal issue or some other problem. But often that stored data is not exportable or analyzable in any practical way, rendering it essentially useless for AI and organizational learning.

All powerful tools can do harm if used inappropriately, and AI is no different. The last several years have produced plenty of negative press about how biased AI can result in discrimination and adverse impact.

For example, IBM’s Watson for Oncology was marketed in 2013 with the goal of curing cancer. The idea was that it would be trained on patient data to uncover treatment options, but it did not work and even recommended some dangerous treatments. One doctor commented, This product is a piece of s—.¹ It was later reported that engineers trained the system on a limited sample of hypothetical patient data.

More recently, Facebook came under fire for serving up job ads in a biased way. Researchers studying the issue reported that Facebook’s ad delivery can skew the ads by gender in a way that can’t be justified, leading to the possibility that their ad delivery algorithms might be violating anti-discrimination laws.

The field of industrial and organizational psychology has been doing research on people in organizations for around a century now, and several of this book’s authors have been doing it for over two decades. In that time, we have had the good fortune to work with many clients, both large and small, in helping them study their talent pools and use scientific methods to discover the characteristics that lead to the best organizational and individual outcomes.

One of the best objective evaluations of the success of our approach is an annual award given by the Society of Human Resource Management and the Society for Industrial and Organizational Psychology, called the Human Resource Management Impact Award. This award is given out by scientists to organizations for using evidence-based HR practices. While saying something is evidence-based does not sound like an overly high bar, the reality is that many solutions in the HR space are powered more by hype than rigor. Additionally, many awards are pay-to-play: A company has to buy a membership or sponsorship to be considered. But the HRM Impact Award is not. It is truly independent. In the past six years, the following clients have won the award for their collaborations with us:

•2015: Bank of America for its Universal Fit Pre-Hire Assessment, which in the first year of use was shown to save 4,500 hours of recruiter time and almost $7 million in reduced turnover costs.

•2017: A health-care company for a customer service assessment shown to improve training performance, hold times, call transfers, and call backs.

•2018: Amazon for a suite of assessments based on its Leadership Principles predicting new-hire job success.

•2019/2020: Walmart for its retail associate assessment, which predicts performance and turnover for its more than three million annual applicants.

•2021: Comcast for a revamped hiring process that not only predicts job performance but also allows for greater flexibility to meet varying staffing needs resulting from the pandemic.

This is not a technical book about how to program AI or deep learning tools. It is not even a book about how to run HR or talent management. It is about a way of thinking—a way of contemplating data and decision-making that starts fresh from the foundation of modern big data analytics and describes the art of the possible once you view the world through this lens.

And while you don’t need to have any deep knowledge of AI or other advanced analytical tools, you do need to have some idea of what those tools can accomplish, so you can plan for the bright future they can bring to talent management. You also need to understand the pitfalls of this powerful new technology to avoid tripping over them. AI, like any powerful tool, can be used in good and bad ways. To be ultimately beneficial for the human race, AI must be carefully harnessed and intentionally applied.

In the chapters ahead, we discuss the issues with HR that have existed since its inception, bring in industrial and organizational psychology and advances in technology such as deep learning, and then describe how these modern capabilities can be used to create the HR organization of the future. Doing so can effectively solve the problem of talent decisions forevermore, but to reach this state, you must be willing to invest the time and effort to get it right.

Those who do will decode first, and they will win the personnel game.

CHAPTER 1

WHY THE RESUME MUST DIE

If I had a gun with two bullets, and I was in a

room with Hitler, bin Laden, and Toby,

I would shoot Toby twice.

—MICHAEL SCOTT (speaking about the HR rep), The Office

A world of difference exists between the technological tools used each day by HR practitioners and this book. Productivity solutions like Slack and social media sites like LinkedIn are built on two-way communication. A book, on the other hand, is a one-way encounter. The author writes. The reader reads. Despite the unilateral nature of this relationship, we know something about you. In your heart, you suspect that how your company recruits and manages future high performers is fundamentally broken. And has been for years.

We make no claim of psychic powers. We didn’t ask you to complete a questionnaire full of questions like, Are you able to comfortably lift thirty pounds? and Do you hold a commercial driver’s license? But we can say with