Will Artificial Intelligence Replace Many Federal Workers?

The short answer is yes. But there is a lot more to it than that, so let’s take a look at what Artificial Intelligence (AI) may do to and for the federal workforce. There is a lot of talk about the coming wave of AI (like the previous coming waves of AI) and how it will finally revolutionize virtually everything. Lots of AI hype frequently means lots of gloom and doom about a dystopian future, or glowing stories about an idyllic dream world driven by technology. I prefer to look at what the reality of today is telling us, and for the most part, at least with respect to the federal workforce, it is not bad.

As our use of technology has grown, and the capabilities of technology have accelerated, we have started to see some of the benefits of AI in many types of work. Technology can allow knowledge workers to devote more time to the substance of their work and less to administrative details. For example, an ICF analysis of data for the state of Colorado found that “Of the 15 task categories studied, time study participants spent the highest percentage of time (38 percent) on documentation and administration …” The obvious outcome of the study was the need for more people, but it also identifies areas that are ripe for better use of technology. Other areas where we are already seeing use of basic AI are help desks. Most of us have probably called a service provider and had the call answered by a computer rather than a human. A few years ago, such calls could be maddening. Now they are becoming more usable and more useful. I made such a call yesterday, and although my immediate reaction to talking with a bot was to get irritated, the system quickly identified what I needed and I was off the phone faster than I might have been otherwise. In that case, it was most likely a good rules-based bot. As AI-based bots become more common, we should expect more and more routine customer service interactions to be with technology rather than people.

Other uses of AI include law enforcement. Facial recognition, analysis of financial data from multiple sources, increased use of sensors to detect movement in remote areas, and use of sensors in general have the potential to revolutionize law enforcement. Imagine walking through a screening line in an airport, bags in hand, and stopping only if the AI detects an anomaly. While that sounds like science fiction (the movie, “Total Recall” to be specific), it is actually a combination of sensors, biometrics (such as facial recognition) and AI that is finally beginning to come to market. There are countless other ways that AI can be used in government, including border security (a “wall” of sensors and AI rather than a conventional barrier), medicine, research, and reducing the need for human interaction to obtain information.

So what does that mean for the federal workforce? Should we automate as much of the federal workforce as possible out of existence? That is a potential outcome, but it is not the most likely or the best. If we look at the history of adoption of technology, we find the disappearance of one type of job leads to the emergence of others. For example, if we go back 50 or 60 years and look at a typical federal office, we would find row after row of clerks. In fact, during the 1950s, the federal government employed more than a million of them. Those jobs were mostly automated out of existence. By 2014, the number was down to 123,000. In 2018, the number is down to 106,000. So — a million federal jobs went away and the government workforce shrank by a million? No. In the 1950s we had about 1.8 million federal employees. Today we have 2.08 million. We lost over 900,000 jobs, mostly due to technology, but the workforce did not shrink.

What happened is a combination of forces that kept the workforce stable as technology changes might have enabled it to shrink, First, federal jobs are important to members of Congress. They do not want to be the ones who did away with jobs in their districts. Second, the population served by the government grew by 86 percent. Third, demand for government services increased. During the 1950s there was no Medicare. There was no Environmental Protection Agency, or Department of Energy, or Education, or many other federal programs that people have come to expect. Clerical workers were replaced with knowledge workers. Now we are seeing knowledge workers beginning to be supplemented with or replaced by AI.

So what will be the effect of the coming wave of AI? Shrinking government? Overly intrusive and powerful government? Or a better government that is able to accomplish far more than it can today? Maybe all of those? Let’s take a look at a few of those options.

Cut, Cut, Cut. Nope. The size of the federal workforce is remarkably stable. The Vietnam War did not cause tremendous growth in the workforce. Ending the war did not cause big drops. The Defense buildup during President Reagan’s Administration was offset by reductions in the nondefense workforce. The opposite happened after the end of the Cold War. Even though we read a lot about the Trump Administration shrinking the government, the reality is that at the end of the Obama Administration we had 2,093,868 federal employees and at the end of March 2018 (the most recent data available from OPM’s FedScope) we had 2,075,006. That is a reduction of only 0.9 percent. Surveys of the U.S. population show widespread support for most federal agencies and the work they do. The only mission area that gets less than a 50 percent approval rating is foreign aid. Politicians may like to criticize government, but they have not shown a big appetite for cutting it. Public approval of government programs and agencies, combined with the jobs they create, make wholesale AI-driven reductions in the workforce unlikely. What is more likely is that money will be shifted to other uses and the workforce will stay within a few percent of where it is now (and has been for 60 years).

Big Brother is Watching. The jury is still out on this one. The tension between advocates for civil liberty and privacy on one hand, and advocates for security on the other, is growing. Do we want facial recognition tech to be able to detect potential terrorists? Or criminals? Of course we do. Do we want the same tech following us around and amassing data on every move we make? Probably not. This one will take years to reach some level of balance that is acceptable to the American people. It is ironic is that we generally do not want the government intruding into our business, but we routinely post our every move on Facebook, Instagram, Twitter and other social media platforms. And those platforms use that information to sell us advertising and market the data to others for advertising, political campaigns, and more.

The Sky’s the Limit. I believe the more desirable outcome is to use AI to greatly improve services to the people. If we use the example of understaffed organizations who employees spend too much of their time on administrative tasks, the goal is not to cut jobs, but to enable that staff to accomplish the mission by letting the systems handle the routine or administrative work. It is not a sexy way of looking at AI, but it is a way that we can enhance the capabilities of government without increasing the cost. There is far more that can be done. Analysis of more and more data by AI may mean the Transportation Security Adminstration can offer superb security with virtually no intrusion or delays. It may mean that Customs and Border Protection can provide a far less permeable border without resorting to physical barriers and with fewer Border Patrol Agents put in harm’s way. It may mean safer food and medicine though analysis of data from various sources and in ways that we cannot do with people. It might mean consulting an AI virtual physician for a diagnosis rather than driving to a VA hospital. It might mean combining AI with Virtual Reality to deliver training that is responsive to the learner in ways conventional training is not. We could spend a month imagining the possibilities and only scratch the surface. It might even mean that Siri could dial a phone number for me without saying “I’m sorry Jeff, I don’t understand.” But maybe I should not hold out hope for that one …

So how do we get to that version of the future? We could start by viewing AI as an opportunity to improve capabilities and services, rather than to cut costs. One of the obstacles to the adoption of technology has always been the fear that it might cost someone a job. We should just accept reality and move in that direction in a way that helps workers learn new skills for the changing world of work. The truth is that work has been evolving from the beginning of civilization and it will continue to evolve. Will some jobs go away? Absolutely. But they can and will be replaced with other jobs.

We also have to break our ties to the way things are today. The blinders placed on us by what we see today are among the greatest enemies of progress. We look at how work is done and view the future through that lens and let our decision-making be bound by it. The better approach is to look at what we want to accomplish, rather than how we might change what we are doing today. That means we should not delegate the design of the future to the stovepipes of today. Self-perpetuating stovepipes typically try to preserve their power, their numbers, and their existence.

The answer to the question “Will AI replace many federal workers?” is definitely yes, it will. But it will also open up more opportunities for work in government, and it is likely to make government far more effective and able to serve the American people in a more personal way.


Tagged ,

HR — The $5 Billion Question

A recent Government Accountability Office report on Department of Defense (DOD) efforts to reduce overhead costs highlights the difficulty in achieving savings in administrative services. DOD was tasked by Congress to reduce overhead costs by $10 billion per year. HR is one of those overhead services that can consume a surprising amount of money. In DOD or any agency, the cost of operating HR offices is a significant line item.

How significant? Let’s take a look at only those jobs that are in the two primary HR job series (GS-201 for HR Specialists and GS-203 for HR Clerks and Assistants). At the end of March (the most recent available data) the federal government had 29,830 GS-201s, with an average salary of $90,648. There were 10,168 GS-203s, with an average salary of $46,135. Without fringe benefits, the GS-201s cost $2.7 billion per year and GS-203s cost $469 million. Add 30 percent for fringe benefits (health and life insurance, retirement, etc.) and those costs go up to $3.5 billion and $610 million (a total of $4.1 billion). We know that not every person in HR is a GS-201 or GS-203, so the actual cost for HR office labor is actually more than $4.1 billion. A reasonable estimate of non-HR series labor costs in HR is about 10 percent, so if we add another $400 million to cover those folks, we have total HR labor costs of about $4.5 billion per year.

These numbers do not include other costs of running HR, such as technology, contracts, office space, and other overhead costs. Most people use 80 percent as the amount of a typical budget that is direct labor, and 20 percent for those other costs. If we add that into our HR cost calculation, we end up with a conservative estimate of the total cost of running federal HR offices coming in around $5.6 billion per year. The costs are increased by duplication. Every time an agency has an abundance of HR offices, the cost of delivering services goes up. Salaries are the most obvious cost. For example, if an agency has too many HR offices, it will likely have far too many supervisors, administrative jobs, and other support jobs in HR. It will have duplicate costs for some information technology. It will have higher costs for office space. All of those add up.

One example of how much those costs add up is my experience in the Defense Logistics Agency. We reduced the number of full service HR offices in DLA from seven to two. In doing so, we reduced the need for much of the office space, eliminated duplicative IT contracts and licenses, and reduced the number of supervisors. The bottom line? A 28 percent reduction in the cost of delivering HR services, with substantially improved quality of service.

What we did at DLA was not rocket science. It was the basics of management. Reduce duplication, control overhead, get more resources focused on service delivery, and improve processes to get better results. That same approach can work in the Department of Defense or any other agency that has too many HR offices. I think that may mean most of them. Just look at the numbers. Two percent of the federal workforce are HR folks. The number should be significantly less than that. A servicing ratio goal of 1:100 is very attainable. When I left DLA we were at 1:102. The government wide cost savings for streamlined and consolidated HR services could easily exceed $1 billion annually.

How would that consolidation look? DOD is a great example. DOD has 730,000 civilian employees and more than 20,000 GS-203 and GS-201 employees. That does not include the non-HR job series folks who work in HR. Does DOD need 2.8 percent of its workforce to be in HR? At a cost of almost $2 billion per year? I think the answer is no. The entire DOD workforce could be supported by 20 – 25 HR offices that provided HR support to the Armed Services and the Defense Agencies. I would suggest that DOD consolidate all of its HR services into a Defense Field Activity, reporting to the Chief Human Capital Officer, that would be responsible for HR information technology, HR service delivery, and HR policy. If we use very conservative numbers and aim for a servicing ratio of 1:80 (about 9,125 operating HR jobs), and allow a very generous number (up to 20 percent, or 1,800 more jobs) on top of that for policy, HR IT and oversight positions, we would come up with a number of about 11,000 HR staff for the department. Even if we added an extra thousand jobs for whatever they might be needed to do, the savings would easily be in the hundreds of millions.

Would it be easy? No. It would require consolidation that DOD has resisted for many years, even after DLA proved it could be a successful approach. It would require significant changes in thinking, with the Armed Services and Defense Agencies moving to the role of customer of an HR organization rather than owner of those same organizations. It would be disruptive, and we know how much everyone loves big changes. But — the bottom line would be better HR services at a lower cost.

HR is an interesting case. Most managers and employees will tell you they are not thrilled with HR. They can almost always point to HR folks who they think are great, but they complain about timeliness and quality of HR services. While some of that is the result of federal HR rules, regulations and laws that are inflexible and outdated, much of it is the result of poorly managed and trained HR staff who are using bad processes and outdated technology. When DLA consolidated its HR services, reduced costs, and improved the quality of services, we did not do it with a new HR staff. Most of the same people who delivered poor service were able to deliver good service when they worked for a more effective and efficiently designed organization. The problem was not the workers, it was the management. I believe we might find the same thing is true in other parts of DOD and in other agencies.While we could continue lamenting the bad service, conduct wholesale outsourcing of HR work, or just do nothing, I would prefer to see what we can do with a smaller, better organized and better managed federal HR workforce. The results might surprise everyone.