18F: GSA's In-House Digital Services Consultancy

There is a small office inside the General Services Administration whose entire job is to be hired by other parts of the federal government. It does not sell anything in the conventional sense, it does not run a program of its own that citizens interact with directly, and most Americans have never heard of it. It is called 18F, after the cross street where its first staff sat, and on a good day it functions as the federal government's in-house digital consultancy.

Origins and the slightly improbable premise

18F was established within GSA in March 2014, in the wake of the HealthCare.gov launch troubles, on a premise that struck a number of people at the time as either obvious or faintly heretical depending on where they sat. The premise was this: agencies building public-facing software might benefit from access to designers, engineers, and product managers who actually do that work for a living, organized as a team that could be brought in for a project and then leave again.

The arrangement is unusual mostly because the federal government does not, as a rule, run consultancies. It runs programs. Programs have appropriations and missions and authorizing statutes, and they tend to last a long time. A consultancy, by contrast, lives or dies by its next engagement, and 18F was set up to live or die that way on purpose. It sits inside GSA's Technology Transformation Services, the same umbrella that houses Login.gov, Cloud.gov, and a handful of other shared platforms.

The headcount has wandered around the 150-person mark over the years, give or take depending on hiring cycles and the political weather. The work is project-based, and clients are other federal agencies — the Department of Agriculture, the Department of the Interior, the Office of Personnel Management, and a long list of others who have, at some point, needed help shipping a piece of software and decided to call across town rather than out to a contractor.

How it gets paid, which is the interesting part

Most federal offices are funded by an annual appropriation. 18F is not, or at least not primarily. It operates on a cost-recoverable basis, meaning that when an agency engages 18F, that agency pays 18F's costs out of its own budget through an interagency agreement. The mechanism is established by the Economy Act and related authorities, which permit one federal agency to acquire goods or services from another rather than going to the open market.

This is, on inspection, an oddly elegant arrangement and also a somewhat precarious one. It means 18F has to be useful enough that other agencies will voluntarily choose to spend their own money on it, which imposes a kind of market discipline rare in government. It also means that when those agencies decide not to, 18F has a problem, and there have been periods in 18F's history when the problem was acute. The financial recovery rate has been a recurring subject of internal review and Inspector General attention.

The cost-recoverable model is worth pausing on, because it is the structural feature that most distinguishes 18F from what one might call ordinary government work. An engagement begins with a discovery sprint, typically a few weeks long, during which 18F staff sit with the partner agency and figure out what is actually being asked for. This is followed, if both sides agree, by a longer engagement to build, prototype, or advise. The interagency agreement specifies the scope, the duration, and the fee.

The open-source default

18F's working repositories are, by default, public. Code is written in the open on GitHub, design documents are published, and the team maintains a methods handbook describing how it does what it does. The default is not absolute — there are sensitivities around procurement-sensitive material and personally identifiable information — but the presumption runs toward openness rather than away from it.

This has two practical consequences. The first is that other agencies, and other governments, can read the code and the documentation and, in many cases, fork it. The State of California's digital services team and a number of municipal civic-tech offices have borrowed liberally from 18F's published patterns. The second consequence is that the work is, in a quiet way, auditable. A curious citizen with a free afternoon can read what was built, how it was built, and roughly what it cost.

The open-source default is partly philosophical and partly pragmatic. Government code paid for by the public, the argument runs, ought to belong to the public unless there is a specific reason it should not. Pragmatically, it also means that 18F's work product does not vanish into a contractor's archive at the end of an engagement.

Sister organizations and the small federal digital diaspora

18F is not alone, although it is the oldest and largest of its kind. Two sister organizations operate under different parents and slightly different rules. The United States Digital Service, USDS, was established the same year as 18F but sits in the Executive Office of the President rather than in GSA, which gives it a different convening authority and a different relationship to political appointments. The Defense Digital Service, DDS, performs an analogous function inside the Department of Defense, where the security posture and the mission set are, predictably, distinct.

The three organizations are sometimes treated in the press as interchangeable, which is roughly the same error as treating a city planner, an architect, and a structural engineer as interchangeable. They overlap in worldview and frequently in personnel — engineers move among them — but the engagements look different, the funding looks different, and the kinds of problems they take on look different. USDS tends toward the cross-cutting and the politically visible. DDS tends toward the operationally classified. 18F tends toward the long civilian build, the agency that needs help with a public-facing form or a backend modernization.

There is also a fourth piece of this picture, the GSA Centers of Excellence, which is a separate TTS program focused on agency-wide IT modernization rather than on individual digital products. 18F and the Centers of Excellence often partner on engagements where a single agency needs both a strategic modernization plan and the team to build the first deliverables under it. The boundary between the two is blurry, and people who work at GSA will cheerfully admit as much.

What an engagement actually looks like

The texture of 18F's work tends to surprise people who expect either a Silicon Valley software shop or a traditional government program office, because it is neither. A typical engagement might involve a small team — say, four to seven people — embedded with an agency partner for somewhere between a few months and a couple of years. The team usually includes a product manager, one or two designers, two or three engineers, and sometimes a content strategist or a procurement specialist.

The procurement specialists are worth flagging, because procurement is one of the harder problems 18F has tried to address. A meaningful share of federal software trouble traces back not to engineers but to the contracting documents that engineers were handed. 18F has, at various points, run an Acquisition Lab and embedded with agency contracting offices to help them write better statements of work, structure modular contracts, and run agile-friendly procurements. This work is less glamorous than building software but arguably more consequential, because a well-structured contract can produce good software for a decade while a poorly structured one cannot.

The deliverables vary. Sometimes 18F builds a working application and hands it over. Sometimes it builds a prototype that informs a later procurement. Sometimes it advises and writes no code at all. The methods handbook published on the 18F site lays out the standard playbook in considerable detail, including the user-research practices, the agile cadence, and the human-centered design framing that the team has settled on.

The strange position of being inside the bureaucracy and of it

One of the recurring observations made by people who have worked at 18F is that the organization occupies an awkward middle position. It is staffed largely by people from the private technology sector, hired under special hiring authorities that allow GSA to bring in technical talent on term-limited appointments. It works inside agencies that have, in many cases, been doing things a particular way for forty years. The cultural translation problem is real and is not always solved gracefully.

The awkwardness cuts both directions. 18F staff sometimes underestimate the legitimate reasons an agency cannot simply ship a feature on Friday afternoon — reasons involving statute, regulation, the Paperwork Reduction Act, accessibility requirements under Section 508, and the long memory of agencies that have been burned by previous modernization efforts. Agency staff sometimes underestimate the genuine value of having someone in the room who has shipped consumer software and knows that the form on page 4 does not have to be there.

Where the model works, it works because both sides do the translation patiently. Where it does not work, the engagements tend to end early or quietly, and the postmortems are sometimes published on the 18F blog in a register that, by federal standards, is unusually candid.

The relationship to the rest of GSA

It is worth situating 18F within the larger GSA, because the parent organization is sprawling and 18F is a small piece of it. GSA, per its own organizational materials, is principally in the business of acquisition, real estate, and shared services for the federal government — the Multiple Award Schedule, the Public Buildings Service, the SmartPay charge card program, the per diem rates that determine what a federal employee can expense for dinner in Albuquerque. The Technology Transformation Services, where 18F lives, is the part of GSA that runs technology-flavored shared services rather than physical or contractual ones.

This placement matters because GSA, as the government's quartermaster, has an unusual cross-cutting view. It serves nearly every federal agency in some capacity, which means 18F inherits, for free, a degree of convening power that a digital services unit housed inside any single mission agency would not have. It can talk to the Department of Veterans Affairs in the morning and the Bureau of Land Management in the afternoon without crossing a turf line, because GSA is in the business of crossing turf lines on purpose.

A modest closing observation

The notion that the federal government should keep a small, project-based digital team on staff, available to other agencies on a cost-recoverable basis, sounded slightly strange in 2014 and sounds slightly less strange now. Whether 18F survives in its current form, gets folded into something else, or eventually outgrows the consultancy model is a question that depends on appropriations, administrations, and the slow-moving politics of federal IT, none of which can be predicted with confidence. What is clear is that the experiment has lasted more than a decade, has produced a substantial body of working software and published methods, and has trained a generation of public-interest technologists who are now scattered across federal, state, and municipal governments doing similar work under different banners.