Last week, a tech entrepreneur named Sahil Lavingia published “DOGE Days,” which reflected on the almost two months that he spent as part of Elon Musk’s “Department of Government Efficiency,” or “DOGE,” within the Trump administration. Lavingia spent his 55 days of DOGE employment embedded with the U.S. Department of Veterans Affairs.
Last week, I ended three and a half years working for a digital services company and federal contractor. I spent the entirety of my 1,279 days in that job on a contract with the U.S. Department of Veterans Affairs, as the accessibility specialist on the VA.gov Platform Governance team.
Lavingia’s blog post is a day-by-day accounting of how he spent his time at the VA. In that post and in subsequent media appearances, he describes his intentions and goals, and tries to distance himself from the most visible parts of the Trump administration’s effort to reshape the federal government. It’s significant as a political document, and fascinating to read as someone who overlapped with his time there.
I never directly interacted with Lavingia, although I did watch some of his activities in Slack and I heard plenty of stories about him. I don’t think he’s a bad person, although I also don’t have a lot of sympathy for him. I definitely perceive his experience at the VA differently than he does. I wish there was some evidence that he’s reflected on the harm that he caused, or took responsibility for the ways that his actions diminished the VA’s ability to accomplish the positive goals that he says he hoped to achieve. Perhaps some day he’ll have enough perspective to realize that his approach to government service was fundamentally misguided.
But that’s my perspective, and his is very different. That’s okay. In his post he gets to be the hero of his own story, and more power to him.
My blog post, the one you’re reading now, isn’t actually about Sahil Lavingia. But his version of his story has loomed large in my mind as I’ve reflected on the chapter of my career that has now closed. Thinking about Lavingia’s tenure in tandem with my own has been really helpful in framing and processing.
Just as he gets to be the hero of his story, I get to be the hero of mine — and that’s what this post will be, some super self-indulgent reflection. I don’t mean to put his experience and my experience in opposition to each other, except maybe for all the ways in which I do.
The biggest stage in the world
I’m not a veteran, and when you’re not a veteran it’s easy to forget just how big the Department of Veterans Affairs is.
There are 1.3 million active duty members of the United States Armed Forces, with hundreds of thousands more serving as reservists. Every year, tens of thousands of service members separate from the military and become classified as veterans. That means there are more than 18 million currently living veterans in the US, a population greater than 14 states combined and roughly equal to the population of the Netherlands.
The VA operates the largest health care provider in the United States, which is a tremendous enterprise on its own. But it also helps veterans finance their education, buy a home, find a job, and more. It’s an enormous organization, and when it’s at its best it offers millions of Americans benefits and services that can transform your life for the better.
The VA tracks service-connected disabilities — meaning, medical conditions classified by the government as disabilities that the government directly attributes to something that happened during someone’s time in the armed forces. That’s a fairly narrow definition of disability, and yet 27 percent of veterans have a service-connected disability. If you focus just on post-9/11 era veterans, that jumps to 43 percent with a service-connected disability. And this doesn’t count the disabilities that you might acquire over the course of your civilian life after serving.
The implications of some of those numbers are something everyone should spend some time sitting with. But for the self-serving purposes of this blog post, working in digital accessibility at the VA was a huge opportunity for me. If you make accessibility your career, the VA is the biggest, most visible stage in the world. More so than in any job before this, I had an opportunity to make a big difference in a lot of people’s lives.
What I did
I’ll warn up front, there’s going to be some gratuitous name-dropping here. I worked with so many extraordinary people and there’s no way I’m going to name all of them, but I’m going to squeeze a few of them in here.
On my last day, Matthew Dingee and Shira Goodman — incidentally, two of the most talented and dedicated people I met working in government — shared with me that in three and a half years I documented more than 500 accessibility barriers in pre-launch reviews. Of those barriers, 128 of them were categorized as launch-blockers. (The standard for creating a launch-blocking issue at VA is relatively high — basically, something that’s very likely to prevent someone from completing a task.) I suspect that I logged more issues in our review process than anyone else in my time there, possibly many times more than anyone else. I caught a lot of problems before they made it to production, and I made the veteran experience better in a real and measurable way.
I also participated in more than 200 design review meetings, providing the teams building digital products for VA.gov advice and course corrections to prevent accessibility barriers from being coded in the first place. I helped to refine design system patterns and components to help designers make the best and most accessible choices as they worked — my very small part in building the most impressive design system I’ve ever worked with. I made suggestions for usability research sessions with assistive technology users, and learned so much more from the results of those sessions.
I formalized the accessibility review process for products launching to VA.gov — a piece of documentation I’m incredibly proud of, and something that I hope will benefit whoever is in this role after me. With my teammates, I helped re-imagine and re-write the standards that define a digital experiences at the VA. (Sara Sterkenberg and Erin White, two of my all-time favorite people, get the credit for the vision on those standards. But I’m proud of my part in it.) With my colleague Ryan Smith, I co-wrote an assistive technology testing procedure for design system components that’s more thorough than anything else I’ve seen anywhere in the world.
I also collaborated with incredibly talented people like Josh Kim, Eli Mellen, Jeana Clark, David Kennedy, Jasmine Friedrich, Rachael Penfil, and a dozen or so other brilliant accessibility practitioners on so many problems and projects. And I had the incredible pleasure of working with people like Allison Christman, Keegan Kitagawa-Bosch, Rebecca Ferguson, Jim Ryan, Cheryl Evans, and others — none of whom had “accessibility” in their job titles but nonetheless have great instincts and often flagged accessibility barriers before I ever found them.
There’s a lot that I could have done better, and so much more that I wish I did. But I still accomplished a lot, and I’m proud of the collaborations I had with people across uncountably many teams. I genuinely believe what I did made people’s lives better, and the experience was so very rewarding.
What I learned
I’ve left that job and moved on to a new one. It’s the right move for me for several reasons — but those reasons aren’t what this blog post is about. Knowing that it’s time to move on doesn’t diminish the value of an experience. I learned a lot over the course of so many audits and so many design reviews.
The things I learned are largely about accessibility, but also I’d suggest that maybe they’re applicable to a lot more than just accessibility. I don’t know how insightful any of these lessons are, but they feel important to me.
1. Large organizations can do good
A large organization can be a tremendous vehicle for positive change in the world. Working with a government agency or at a big corporation or a major university can be an opportunity to build a meaningful capacity to do good, and to improve a lot of people’s lives.
I realize that “good” isn’t really well-defined here, and I’m not going to try to define it either. But let’s just assume for now that my version of good matches yours, especially as it pertains to accessibility.
It’s easy to have a negative view of bureaucracy. When an organization grows past a certain point, it becomes depersonalized to outsiders (and sometimes even to insiders). It’s no longer perceived as a collection of real humans trying to do their best work. It becomes a giant soulless blob.
But it is, in fact, still a collection of real humans trying to do their best work. And when those humans work together toward a positive goal (like building accessible government services), they can do some genuinely impressive things.
Leadership support and values are an important part of building that capacity to do good, although I would argue it’s not the most important piece. Dedicated and knowledgeable accessibility practitioners are an important part of it too, but I also think that’s giving people like me more credit than we deserve. Expertise is necessary, but it’s not sufficient.
Instead, I would point to a sustained network of people who embraced accessibility as a norm — an organizational culture that says “of course we’re not going to launch that feature if it doesn’t work for screen reader users” and “we care about getting the semantics of buttons and links right” and “oh wow I’d never thought about how this would work for voice users, where can I learn more?” That culture is definitely not present across the whole of the VA, which is a massive organization and can’t be defined by a single shared ethos. But it’s a culture that’s embraced by a big segment of the organization, big enough that while I worked with the VA we built an incredibly sophisticated and successful accessibility program.
Building culture — and building the capacity to be a force for good — is, again, not a task that can be solved with code. It comes from people working together, from a shared sense of purpose and a shared understanding of the problem space. These are human tasks that can’t be automated.
2. Most accessibility issues are really governance issues
On my last day, David Kennedy asked me if I had any final hot takes to share. This was the first thing that popped into my head: Most accessibility issues are really governance issues.
As soon as I said it I was absolutely sure it was true (the feeling you should always get for the hottest of your hot takes), so much so that I repeated it a couple of times throughout the day in other contexts. And now I’m saying it again here.
I’ll acknowledge that there’s a certain tautology to this. In a space largely defined by legal compliance, failure to meet compliance requirements is necessarily a governance failure. In this sense it’s not really new information.
But beyond that, I would suggest that gaps in governance processes are ultimately what generates the vast majority of accessibility issues. A developer doesn’t intentionally create a keyboard trap. A designer doesn’t intentionally create a confusing, high cognitive load user flow. But these things happen because some governance process failed.
They could be gaps in how standards are defined, or gaps in onboarding and training, or gaps in QA requirements. They could be inconsistent mechanisms for prioritizing work, pushing teams and individuals to cut corners. They could be inadequate communication channels that fail to share information to the right people at the right times. They could be broken escalation paths through which competing visions for a product can be reconciled.
Accessibility barriers flow from these kinds of issues, and closing this kind of gap goes a long way toward preventing future issues from being created.
There’s probably an argument that strong governance is incompatible with agile methodology, that well-defined processes always result in reduced velocity. I don’t think that’s really true — I think if your governance processes are slowing down your product teams that’s probably a sign that your governance is under-resourced.
I’ll also note that I don’t think this is only true for large organizations. Small teams in small organizations would benefit from well-defined governance perhaps most of all, but are most lacking in the capacity to build those processes.
Regardless, it’s not the sexy and exciting work of coding, moving fast, and breaking things. It’s not the kind of work that someone who believes that complex problems have software solutions would want to do. It’s the boring and slow work of understanding problem spaces, building relationships, and clearly articulating and enforcing requirements.
3. How you communicate about accessibility helps build culture
I know I just said I didn’t want to give people like me more credit than we deserve, but also this is my retrospective blog post. I get to be the hero of my own story, after all.
In my role working with the VA, I was hyper aware that I was the bad guy to a lot of my colleagues. Software engineers and UX designers would work incredibly hard, and then I would come along and point out a dozen problems with what they had done. Sometimes I would tell them they needed to completely rethink their approach to their product. I had the power to ruin a lot of people’s day.
This is something that would keep me up at night sometimes. The work that I did was valuable, and I know that there are veterans whose experiences with the VA are better because of it. But there’s an emotional toll to being a bad guy.
With that awareness, I tried as hard as I could to communicate that I was a colleague and not an adversary, that I was a potential collaborator to help solve the problem. Borrowing a phrase from my friend Crystal Tenan, I frequently told people I wanted to be their best friend — a trusted resource that they could come to for any accessibility questions.
And one part of that communication style that I found incredibly important and effective was deliberately expressing uncertainty.
When I provide feedback, I try to show my work by explaining how I reached my conclusions and what the potential impact is for users. I don’t think it’s ever enough to just say “this violates WCAG 3.3.4” or whatever. Instead, you explain what you’re seeing, how you think some group of users will interpret it, and what the impact will be on their experience. Every bug is a learning opportunity.
But if I’m not certain about something, showing my work also means highlighting that uncertainty. If something is in an accessibility gray area, I make sure it’s clear that I’m providing an interpretation of the situation and that other interpretations are worth considering. I think it’s important to demonstrate that although I have some pretty deep knowledge of accessibility, I don’t have a monopoly on expertise. Sometimes the honest answer is “I don’t know for sure, but here’s what I’m thinking.”
I’ve found that designers and developers respond really well to transparent feedback that engages them in a discussion of the problem, and we’ve come up with some really interesting solutions from those conversations. In my experience, innovation comes from a place of sincere interest in the perspectives of others, a place of valuing their professionalism, expertise, and problem-solving abilities.
The culture of accessibility at the VA existed before I arrived — I can’t take credit for it. But I am hopeful that I helped to grow and sustain that culture. The “shift left” accessibility ethos depends on everyone having a stake in solving the problem, and how we communicate the problem can play a big role.
What’s next
As I wrote this blog post, several other ideas bubbled up — thoughts on the uncertainty at the heart of accessibility, thoughts on the limits of artificial intelligence, thoughts on how big organizations like government agencies are structured. I almost wrote a lengthy digression on disability and identity, featuring an excerpt from an episode of the podcast Gender Reveal (transcript, .docx file) and some Billy Joel lyrics. Hopefully some day I’ll get a chance to organize these thoughts into a few more posts. What I have here is probably enough for now.
This morning Pro Publica published an extraordinary article examining the AI prompts that Sahil Lavingia used to identify contracts to cancel. It’s worth reading, and may give you some thoughts on uncertainty, AI, and the structure of government agencies too. But again, that’s for another time.
What’s next, for me, is a new job. I already miss the amazing humans that I’ve worked with for the last three and a half years. I miss the familiarity of working in an organization I’ve gotten to know well. I’m feeling the discomfort that comes with new people, new processes, and new routines.
But all the same I’m very much excited about the new job. It’s a different kind of stage than the VA was, but it’s still an opportunity to make a big positive difference in a lot of people’s lives. I don’t know how I was so lucky to end up with this as my career, and I’m so grateful.