Wisdom from the Ancient Greeks for Procurement Reform: First, Do No Harm (Mikey Dickerson)

By the 17th century, technological progress had equipped would-be doctors with a bewildering array of new tools. Armed with exotic chemicals, sharp instruments, and a can-do attitude, they set out to cure every disease and discomfort ailing humans, an organism whose complexity far exceeded their understanding. Their patients died by the thousands. As the practice fumbled towards respectability, they adopted a piece of wisdom from the ancient Greeks: Primum non nocere. First, do no harm.

Modern-day custodians of government services are in a similar predicament. Sprawling, human-machine hybrid systems typically span thousands of employees, dozens of distinct software products, at least one mainframe, and hundreds of connective macros and scripts. The way we build these systems allows us to add complexity regardless of whether we understand previous iterations or not. Over time, the result is a system that far exceeds the grasp of any single person.

I worked on a number of such systems, starting with Google, then healthcare.gov (which, contrary to reports, was much simpler than the government norm). In regards to the U.S. Digital Service, we did our best to improve core functions in Social Security, the IRS, FBI, Medicare, VA, and Department of Defense. I have studied similar overly-complex systems in California, Canada, Washington, Montenegro, and at a variety of nonprofits and corporations. The current state of business process automation follows a predictable pattern.

To take just one example, Medicare is governed, in theory, by tens of thousands of pages of regulations. These regulations are written by a complex rule-making process, and are not organized in any one place that can be read front to back. In practice, Medicare decisions are the emergent results of a thousand or so administrative staff and about 5,500 COBOL scripts that run each night. The system's complexity is no longer limited to what humans can understand, because machines keep repeating processes when their purpose has been forgotten. In time, the knowledge designed into automated systems is fragmented, diffused, and ultimately destroyed.

Updating or repairing such a system is analogous to medieval—and maybe even modern—surgery. We show up with a mixed bag of rusty old tools, shiny new tools that we partly understand, and superstitious rituals like “agility” that we carry because they worked once. Tactical interventions are possible. When the problem is a clogged tube, we have a good chance of finding and unclogging it. Usually, we can even prevent the metaphorical patient from metaphorically bleeding to death. But when we set out to redesign or “modernize” the whole thing, we are out of our depth. It’s like asking a surgeon to fix your body so it is like a teenager’s again.

From my experience, I’ve learned that:

• Most failed projects would have benefited from a smaller budget.
• Rigidly specified launch timelines and multi-fiscal year schedules for prescribed expenditures are never helpful and often lead to waste.
• Contract structures should not leave vendors unsupervised for long periods.
• Deliverables should be plain-English citizen-based outcomes, rather than defined in terms of arcane technical requirements.
• Agile contracts are not a magic bullet. When they work, it is because they break projects into increments that are small enough that they can be committed and built upon, or abandoned to be tried again.

However, the most important lesson that I can suggest to the people that have custody of such systems is to start from “first, do no harm.”

This should be obvious, but it isn’t. Multi-year, multibillion, stem-to-stern overhaul or “modernization” projects have no meaningful chance of success. They grind to a halt in the requirements gathering stage. As the years and the appropriated dollars evaporate, the harm to service delivery accumulates, as routine upgrades and maintenance are blocked by the “modernization” project. The two most common outcomes are that the modernization project is abandoned altogether, as when VA canceled Health-e-Vet in 2010 after spending $600 million, or that the status quo becomes so unlivable that the agency is ready to accept any amount of risk to cut over to the new thing. This is how the FBI replaced its National Instant Criminal Background Check system. Policymakers can avoid these outcomes by injecting a little humility early in the process. (There should be ample supply lying around from past projects.) Maybe after a bit more trial and error, we will possess the organizational science to design a big bang replacement for Medicare billing or California unemployment insurance. But we have to survive this medieval period first and a little ancient Greek wisdom can guide us.