Six Million Dollar Men: Policy, Technology, and Talent Management
In 2010, the Defense Integration Military Human Resources System — a joint service solution to provide more accurate, timely pay to servicemembers — was cancelled after spending over $1 billion. The program failed because the services, in a successful attempt to avoid reforming and consolidating their personnel policies, made so many modifications to procured off-the-shelf technology that it no longer worked.
In a recent War on the Rocks article entitled “Good Will Hunting: The Strategic Threat of Poor Talent Management,” several authors correctly identify the Defense Department’s inability to manage talent. They convincingly argue that artificial intelligence and machine learning-enabled systems should be incorporated into the personnel system to better match talented people with the job for which they are best suited. But to achieve these laudable goals, one must first understand personnel policy and why talent is currently mismanaged. The largest inhibitor to talent management across the Department of Defense is the structure of the personnel system. Lack of data and bandwidth does inhibit the process, but is merely symptomatic of a larger problem.
To unlock technological power without doing personal or institutional harm requires a tandem approach of technical innovation and policy reform. Consider an analogy from science fiction: the Six Million Dollar Man. Scientists evaluated Steve Austin’s broken body, incorporated technological enhancements to enable new functionality, and utilized the new man/machine system in an innovative way to address a range of previously unsolvable problems. Similarly, the military must identify and reform the broken elements of the personnel system to enable new technologies to make a “better, stronger, faster” whole. Policies that are especially ripe for change include artificial timelines for promotion, up-or-out promotions, and a board-based advancement process. More broadly, to enable effective talent management, the military must also develop new ways of conceiving of people and jobs.
The Department of Defense is incapable of managing talent because personnel policies preclude identifying, developing, and utilizing talent outside of narrowly defined career fields. The result is an intentional failure to match easily identifiable, qualified people jobs corresponding to their skills.
In 2015, I used internal data to conduct a review of all Navy jobs requiring higher level education. I found that only 31 percent were occupied by people who met the job requirements. Another 18 percent were partial matches. This is not a matter of having enough qualified people, since my review also found that only an average of 16 percent of warfare-designated officers (pilots, submariners, etc.) had been tapped to fill corresponding jobs. The only places where talented people and jobs aligned were to meet the needs of a specific career field. This indicates unwillingness to place personnel outside of career paths despite documented Navy requirements, even when it leads to violations of explicit Department of Defense and Navy policy. To explain this, one must understand the personnel system’s dynamics.
Three Personnel Problems
The job base across the Department of Defense resembles a large pyramid with many entry-level positions at the bottom and an ever-decreasing number toward the top. Servicemembers enter at the bottom, but predefined intervals for promotion opportunities force continual movement up the narrowing layers. With no meaningful way for warfare-qualified personnel to enter laterally at a higher layer, there is an insatiable demand for new bodies to refill the base every year. One either promotes along an artificial timeline or is forced out of the military, a system known as up-or-out. Each career field manages its own milestones and promotes internally to meet manning needs, allowing it to disregard any external institutional priorities.
The result is that servicemembers have finite time to accrue a promotable record within their career field. There is no mechanism for recouping a lost career opportunity, or for staying at a level commensurate with an individual’s skill or preference. Everyone is forced into a leadership track requiring a broad knowledge base, creating irresolvable tension with career fields that require high levels of training and expertise.
Continuing in the system requires approval of a series of promotion and milestone selection boards, which can only devote mere minutes per person. Limited board bandwidth requires that certain shortcuts be taken, such as utilizing only the most basic information from performance evaluations or selecting only those who have held jobs predefined by the career field managers. These jobs are designated as more important because former incumbents were promoted, and when available, the most upwardly mobile personnel are placed in them. This self-reinforcing cycle of making a small subset of jobs more promotable, thereby decreasing the value of all others, is commonly known as the “golden path.” The common reprise that “promotion boards tend to self-select, tapping officers who look like them” should come as no surprise.
Together, these elements of the promotion process explain why timing has a disproportionate effect on one’s career prospects. “Success” becomes defined as continued upward viability instead of job performance, so evaluations do not provide accurate performance-based information. Certainly not all evaluations dismiss individual performance. But the systemic incentives to not “waste” a top evaluation on those less upwardly mobile are clear, and people respond to incentives.
Further complicating the situation, there is no discerning why an individual was ranked a particular way. If a pilot, did they receive the top mark because of their skill in the cockpit, exceptional leadership, or both? Were they the most innovative thinker, hardest worker, or best teacher? Or were they merely the most senior person to depart the unit at that time? Each qualifier likely predicts success in a different subset of jobs, so addressing these questions are critically important when defining “successful performance” or developing predictive matching technologies.
The personnel system’s structure creates a series of irresolvable dilemmas that prioritize career progression over talent matching. Much work in the Department of Defense happens beyond the tactical-level focus of career fields, such as employing forces across a theater or crafting strategic priorities and policies. But developing expertise in these areas, however necessary to the military, often ends careers because they are, by definition, outside the defined career fields in which one promotes. In this way, constant upward career progression leads to less overall expertise and enormous personnel training costs for new accessions. Up-or-out can unnecessarily waste useable talent if people quit because a narrow career field cannot meet their needs, or are promoted beyond their capability with no mechanism to return them to a role where they previously demonstrated exceptional performance. There 2019 National Defense Authorization Act gives new statutory authorities to the services, but it is not clear that the substance will live up to the hype.
The Intelligence Is Artificial
In theory, military personnel commands could increase talent-based outcomes today with no new data inputs or technology. But this does not happen because the system requires that they focus on career progression. Qualifications for jobs off the golden path are, at best, of secondary concern to filling them with personnel whose careers will be least adversely affected by taking an uncompetitive job. Artificial intelligence and machine learning are powerful tools, but introducing them into the personnel distribution system without reforming policy will damage careers by prioritizing anything other than promotability.
Beyond this major issue are two others related to systemic personnel problems that preclude teaching an algorithm about performance and success. First, artificial intelligence-enabled talent matching systems must be accurately taught to produce the desired result. They need to know who is succeeding at a particular job and correlate that with other relevant information to predict who else might succeed in said job. The problem is, as the previous section discussed, the current measure of success is not directly linked to performance. There is no existing database from which to teach what performance-based success looks like because even performance evaluations provide scant and potentially misleading data.
The second problem manifests during personnel distribution. Servicemembers are sent to a particular job within a command, but the command itself controls the individual’s work. It might be exactly what is advertised, or completely different, based on the needs of that command at the given time. A talent matching system would have to be responsive to each command’s needs in a fundamentally new way in order to remain relevant.
Some of these issues can be resolved by incorporating new ideas about talent and competencies into the process.
At its core, talent has three components. The first is learned skills, the only element the existing personnel system explicitly recognizes in training and education. The second is innate abilities or characteristics that serve organizational goals. These are personal attributes which can be developed through experience, but cannot be learned at a schoolhouse or readily changed. For example, a pilot requires more hand/eye coordination than a submariner, but a staff officer (who could be either) requires the ability to communicate ideas and network.
The last component of talent is interest. One who is passionate about something tends to devote more time in its pursuit. Personal interest will not make up for a complete lack of aptitude — a pilot still needs hand/eye coordination — but it makes one more likely to compensate for shortcomings in innate ability.
Mapping an individual along all three axes with detail and accuracy is necessary to predict their likely avenues of success. The more difficult task is encoding each job for needed skills, which requires the development of competencies.
Competencies are generally considered personal characteristics that enable superior performance in a given job or role. Each job requires different mixes of competencies, and its own set of predictors for success. One must disaggregate what makes one great at a particular job, of which there are many different types even within an existing career path. If policy reform allowed anything other than a leadership track, developing competencies would naturally map viable career alternatives.
Developing the foundations of talent and competencies should enable the outcomes envisioned by the Good Will Hunting article’s authors, but it is as critical a first step as reforming the overall personnel system. I respectfully disagree with the authors, however, that existing data will be sufficient to start. Data for performance-based success and personal aptitude are largely unavailable because the personnel system currently cannot make use of them, but much of it could be collected with ease if the demand signal was present.
We Can Rebuild It
If Steve Austin’s broken limbs had been replaced without considering how to best use his new capabilities he might have been stronger, but not smarter. Reforming the personnel system can be accomplished in more ways than I can detail in this article, but to do so without considering how man and machine can best work together to enable talent management would diminish the effort’s outcome.
The military must first address the three major problems with personnel policy to enable even the most basic tenets of talent management. Once accomplished, the laudable goals articulated by the Good Will Hunting article’s authors might well come to fruition.
The iterative nature of artificial intelligence and machine learning makes it likely that the system will help identify new variables highly correlated with success in each competency. As more data are entered, clear patterns should emerge about which are most relevant for a given job. The results could one day even provide feedback on the effectiveness of changes made in training syllabi and personnel policy, as a baseline would exist to allow comparison.
People cannot, however, be entirely removed from the process. No computer system can account for variables such as personality conflicts or the myriad of other factors that might affect one’s job performance. People will provide many of the subjective evaluations that feed the algorithmic assessment. They need to understand and directly benefit from an accurate system, or risk corrupting the process by providing bad data. And sometimes, it takes a personal touch to push someone out of their comfort zone where there is a reasonable hope they could learn and grow in ways an algorithm would not predict.
The personnel system is ineffective because it fills jobs with unqualified people, and inefficient because it needlessly causes massive personnel turnover that reduces the overall level of expertise in the force. It can cost millions of dollars to train each new accession to even a basic level of proficiency, but the return on that investment is often poor. In an era of renewed great power competition and a ballooning national debt, neither the Department of Defense nor the nation can afford the status quo.
LCDR Jacob Yanofsky is a Navy helicopter pilot. He formerly worked on the Talent Management and Sailor 2025 initiatives while on the Chief of Naval Personnel’s staff, and earned an M.A. from Catholic University of America.
The views expressed here are the author’s own, and do not reflect those of the Department of Defense or U.S. Navy.
Image: Flickr/Tom Simpson