The Digital Service Was Never the Point

Colorado becomes first state to commit to the product model at scale

A few weeks ago, I sort of threw down the gauntlet here, calling for bolder transformation towards a new operating model for government. I borrowed the framework of the three horizons, in which the first horizon is the status quo, the third is the new system we need to move to, and work in the second horizon can either extend the life of the status quo system (H2-) or help to end it by building what we need next (H2+). The most common question I’ve gotten since then is “Who is working in that H2+ space?” My answer today — and it’s a good one!! – is Colorado. And I’ve invited Kelly Taylor, former USDSer and co-founder of the Colorado Digital Service, to help explain.

This week, Colorado published a strategy to reset its Office of Information Technology, moving the state from a project-based operating model governed by time, scope, and budget to a product-based one governed by outcomes and customer value. This is a remarkable set of documents that everyone should read. (Deck here and narrative here.) As one friend said, “This is some nailed on the wall Martin Luther s**t.” It reminds us, in a good way, of Martha Lane Fox’s original “Revolution not Evolution” letter to the cabinet minister of the UK in 2010, which sparked the creation of the Government Digital Service (and by extension, at least in Jen’s view, of the US Digital Service).

Slide from https://oit.colorado.gov/about-us

Its diagnosis is refreshingly frank: that OIT isn’t delivering at the level Coloradans and agencies need, that it’s organized around its own internal structures instead of outcomes, and has lost sight of its statutory purpose of serving agencies. The symptoms of this include state employees doing manual work that tools should do (this is sadly common), agencies getting high cost-recovery bills disconnected from outcomes, and teams routing around OIT because going around is faster. Where there are skilled agency teams, they are being treated as a compliance problem rather than an asset. It’s not about the people involved; it’s the result of a system built for another era. And the results of that system are particularly unacceptable given the state’s $1.5B budget shortfall this year. The 17 agencies are asking for this, not being told.

We’ve seen frank assessments before. What we rarely see is a prescription/treatment commensurate with the diagnosis. Instead of more promises to expand use of various carveout authorities or increase investment in pilot programs, Colorado is essentially saying, “Look, it’s clear our status quo is failing, so let’s adopt a model that’s proven to work — proven even here.” In a certain sense, it’s very simple. The point of a limited scope experiment is to learn. If you learn that it works, and test it under sufficiently varied conditions, and it still works better than the status quo, then you roll the new way of working out broadly. That’s how it is supposed to work, but it very rarely does.

In this case, the experiment has been the Colorado Digital Service, which Kelly Taylor and Matthew McAllister founded in 2019. It did not begin with a budget line. It began the way most things in government that actually work begin: with someone who had seen what was possible somewhere else and went looking for people who might say yes. In 2018, Kelly was serving on a digital tour of service at USDS in Washington, DC, and started talking with folks back home in Colorado, with health agency leadership, with the then-Governor’s Chief Innovation Officer, and eventually with the Governor himself at a conference in DC. Matthew had been working on similar ideas. The two joined forces that June.

When the new Governor was elected, the work shifted to his administration. In March 2019, Kelly and Matthew pulled together a volunteer group of designers, engineers, and product managers from the Boulder and Denver tech corridor to run a two-week discovery sprint on the state’s troubled payroll system. The analysis gave the CIO and the Governor’s office their first piece of evidence that the supply of senior technologists willing to raise their hands for government work was real. In June, OIT submitted an interim supplemental for $1,130,848 and 6 FTE. The budget committee tabled it.

So Kelly and Matthew spent the summer doing the unglamorous work that structural change requires: one-on-ones with legislators, meetings with the Joint Budget Committee analyst, meetings with anyone who would take one. They testified before the Joint Technology Committee in August. A national kitchen cabinet of civic tech advisors helped recruit talent and lined up funding from the Beeck Center. On September 20, the JBC approved $647,000, half the ask. The Governor announced the team publicly on October 17 with a quirky video and a tweet. When Kelly left to come back to USDS, he handed off leadership to Matthew. When Matthew left, Sarah Tuneberg, who had previously led pandemic response tech for the governor, took the reins.

Colorado is not the only state to create a digital service¹. What usually happens next (and by next, we mean over the course of a few years) is that the new team does good work, and the system carves out a small space where the new way is tolerated (sometimes barely tolerated) while everything around it operates exactly as before. In the three horizons framework, this is the “pilot that doesn’t scale” trap. When someone asks Why are we spending so much money on IT for such bad systems? Aren’t we still failing our users? Isn’t the unemployment insurance system still incapable of scaling to meet demand?, the answer is Well, we have the digital service, and the conversation moves onto more legible issues (and issues that political leadership feel more comfortable with). Ironically, an effective digital service can provide just enough success to excuse the lack of true reform.

What’s clear now is that Colorado intends to actually do the reform. The announcement this week essentially says that OIT, which is orders of magnitude bigger than CDS, is now going to operate at scale the way CDS has been operating in pockets. (We are not saying Colorado is the only one – a few other states are arguably doing something like this more quietly.) How they got there is important for every other state, city, and the federal government to understand. Over the past six years, CDS worked alongside agency teams on the projects that make government function: child welfare case management, unemployment insurance modernization, health data interoperability, and the digital identity work behind myColorado. When COVID hit, the team scrubbed in on covid19.colorado.gov, contact tracing, and exposure notifications in weeks rather than years. But the projects were never really the product. The capacity was. Along the way, CDS helped evolve procurement by bringing engineers onto evaluation panels, paying for sprints rather than fixed requirements lists, and writing contracts that assumed iteration. They built a talent pipeline of senior technologists who would never otherwise have considered state service. And they accumulated the quiet, specific evidence that a different operating model produces different outcomes: faster delivery, lower cost, better experiences, and agency teams that wanted to work the new way.

One of the secrets of CDS’s success, unsurprisingly, is their HR strategy. And we don’t just mean that CDS hired good people, though they did. We’re talking about how they worked with their agency partners to help them hire the right people for the right jobs. That meant rewriting job descriptions in agencies to align with the product model. A project manager working in a legacy IT framework simply does different work, and is measured by different success metrics, than a modern product manager. The team around them has different skills as well (user researchers, engineers). It’s not about doing what IT used to do but better. These are literally different jobs.² A digital service team that works with an agency that continues to hire and manage for the same jobs they did last year, and structures the work in the same way it did last year, is just a tourist. Reassign the team to the next agency partner and everything snaps back to how it was. But change the job descriptions, the hiring assessments, and the success metrics of agency staff, and you’ve rewired the machine.

That accumulated capacity is what made the transformation thinkable. The reset Colorado is announcing, the move from a project-based operating model governed by time, scope, and budget to a product-based one governed by outcomes and customer value, did not appear because someone wrote a compelling strategy memo. It became possible because six years of CDS work had already proven, under real conditions, that the new model worked, and had built the standing with agencies, the legislature, and the Governor’s office to defend it. The state had the talent. It had the procurement muscle. It had agencies that had felt the difference. By the time the window opened, the leap had already been walked, one demonstration at a time.

Clearly, it also had the political leadership willing to back this. We don’t know who exactly in the power structure of the Colorado governor’s office or elsewhere deserves a medal, but we’d be delighted to pin one on them. But we’ll order in bulk, because more are going to be needed. This moment is not the end of the story. At the risk of being grandiose, it’s not even the beginning of the end, but perhaps the end of the beginning. A former White House official used to say privately, “If there wasn’t an assassination attempt on USDS this week, it’s only because we didn’t hear about it.” Those were internal threats, and though USDS faced different dynamics, there are always going to be forces that resist change in government. But there are plenty of outside forces who may be displeased with the changes Colorado is making. There are plenty of vendors who welcome healthy, much-needed disruption, but if this works, some vendors stand to lose contracts. Those vendors have lobbyists, and those lobbyists have an enormous surface area for influence, not just with the current administration, but with a governor-elect in November, and with the legislature. (We are not by any means implying that a new administration is particularly vulnerable to this kind of influence, but we think it’s safe to assume there are actors who will try, and are well resourced.) It would be smart to pair celebration with preparation. The new OIT is likely to come under attack in the coming years, and while both of the front runners for governor have signaled support for the new operating model generally, they’ll need to understand their role in protecting this change, and be prepared to pay the costs of it. Change is rarely free.

One cost of change is the pain of layoffs. 173 OIT employees lost their jobs in this transition.³ We don’t know the details, but our general understanding is that these are people whose jobs were designed around the legacy project model, whose training and expertise is no longer fit for the jobs that will now need doing. As we described above, moving to the new model is not about doing the old thing better – it’s about doing a fundamentally different thing. Tracking progress on a set of requirements is a valid skill set that many public servants demonstrate well. But being good or even excellent at that does not automatically make you good at understanding user needs and making informed judgement calls about what features will meet those needs and how they should be designed. Over the past six years, many project managers, for example, have had the opportunity to learn product management. If we understand correctly, many of those people have been retained. Those 173 affected by the layoffs can apply for some of the 95 new jobs that are now being posted, for roles designed to work in the product model. We imagine this was a difficult choice for leadership to make, but it does nobody any favors to have people in the wrong jobs. This layoff, while hard on those affected, appears necessary to align the workforce around the new model. Without it, the stated goals and the day to day activities will diverge, causing confusion and incoherence. It’s a mark of courage that Colorado’s leaders were willing to go through with layoffs, and presumably a sign of sensible flexibility in Colorado law (not always present elsewhere) that the layoffs were possible.

What other lessons should we learn from Colorado’s bold move here? For other states, cities, federal agencies, and even central federal government strategists, the first point is that the digital service itself was never the point. (Another headline for the Colorado story could have been “Colorado Digital Service Quietly Shuttered.” We’d still be celebrating.) The point (and apologies for the broken record) is that the operating model our government runs on is not fit for the purpose it needs to serve, and needs a comprehensive update. Changing how we build and buy technology in government is a wedge into that larger problem. You start a digital service in order to pry that wedge ever wider, starting with the elements that need to align to adopt product model development: the funding, the staffing, the decision-making authority, the success metrics that must be shared by not just the team, the agency, the governor’s office, but also the oversight bodies that might still hold you accountable to the old model. Soon you’re way out of the IT lane, changing job descriptions, hiring and management protocols, and funding structures, right-sizing and redesigning process and procedure, educating a wide swath of stakeholders about what you’re doing and why. It’s a team sport — leadership cannot expect the digital service team to do this without enormous, ongoing, dedicated support from above and around.

If you’re a leader in another state thinking about starting a digital service, here’s our advice: if all you want is to announce a new initiative, don’t bother. You’ll perform reform without doing it, and it’ll make true reform harder when the requisite political will is present. On the other hand, if you and the power players in your administration have the appetite for the kind of journey that Colorado has been on, and you’re willing to see what you’ve started through, even if it upsets some apple carts, then study these documents Sarah Tuneberg and her team have generously shared with the world (we’re serious – everyone should read them!), and craft a strategy that backs into a future state much like the one she describes. You will still start small, just as CDS did. You will still have to build trust and proof points. You will still have to explain yourselves over and over and over again to an ever-widening set of stakeholders confused or upset by the changes you are making. And even getting as far as Colorado just has won’t mean your entire government is operating in the new model — the digital piece is just a good starting place. But you now have a roadmap that extends past your first few steps, and where it can take you is a place very much worth going.

1. The term “digital service” can be confusing, because it refers both to the actual service people experience, like enrolling in a benefit, for instance, and to institutions within government who build digital services. When you’re talking about the institutional entity, it’s service in the singular (not services), just like the Postal Service, the Forest Service, or the Food and Nutrition Service. The UK started the trend when Mike Bracken and others founded the Government Digital Service in 2010. The US Digital Service is named after the GDS.

2. See Chapter 10 of Recoding America for specific illustrations of this.

3. That’s about 15% of its 1,150-person workforce