We’d like to use cookies to see how visitors use this site and to measure our advertising. Nothing is tracked unless you allow it.

Fix government data sharing before AI transformation

alt=""

Key takeaways

1. AI transformation depends on better government data sharing

AI can speed up government processes, but without secure access to reliable information across agencies, it risks accelerating the same fragmented workflows instead of improving them.

2. Governance is often harder than connecting systems

APIs and integrations can move information between systems. The harder questions are who owns the data, who can access it, under what conditions, and who is accountable when something goes wrong.

3. Secure data sharing doesn’t require centralized data

Estonia’s X-Road demonstrates how independent organizations can retain control of their data while exchanging information through a shared framework for secure, trusted communication.

4. Building trust is the real work behind interoperability

Flexion’s X-Road prototype showed that establishing trust between organizations required more effort than exchanging data. Certificates, security, and membership approval were essential to enabling cross-agency data sharing.

5. Start with a real problem, then choose the technology

U.S. agencies can borrow X-Road’s principles without replicating its implementation. Keep data ownership local, make access auditable, establish reusable trust mechanisms, and begin with one cross-agency problem that delivers measurable value.

A few years ago, I changed my last name through a court order.

Simple enough, right?

Except that I then had to update my information with 16 different organizations. Banks, government agencies, businesses, and everyone else who needed to know. Most of those updates eventually happened. But to this day, my property deed still has my old name because finding time to go to the county office has never quite made it to the top of my list.

It’s a small inconvenience in the grand scheme of things. But it illustrates something that has bothered me for years about how we design government services.

Why are we still asking people to provide the same information over and over when it already exists elsewhere?

The answer is more complicated than outdated systems or missing integrations. And as agencies explore AI in government services, this question is becoming even more important.

The government data sharing problem isn’t always the technology

When we talk about modernizing government services, the conversation often turns to technology. Better systems, modern APIs, cloud infrastructure, automation, and now AI.

All of those things matter. But connecting systems is only one part of the problem.

Even when two agencies have the technical ability to exchange information, that doesn’t necessarily mean they can or should. Five questions decide whether they can:

  1. Ownership: Who owns the information?
  2. Authorization: Who is allowed to access it?
  3. Conditions: Under what circumstances can it be shared?
  4. Accountability: Who is responsible if something goes wrong?
  5. Quality: How do we know the information is accurate and current?

Those questions aren’t really about technology. They’re about governance, trust, and accountability.

A few years back, we decided to explore this problem more deeply at Flexion; we looked at how other countries approach government data sharing, particularly Estonia, whose X-Road-based approach has become a widely recognized example of digital government interoperability.

And rather than stopping at research, we built a working prototype of X-Road, the open-source technology behind much of Estonia’s data exchange infrastructure.

What we learned changed the way I think about interoperability.

What is X-Road?

X-Road is free, open-source data exchange software that lets independent organizations share data securely without pooling it in one central database. Estonia began its national deployment on December 17, 2001. Today it is developed by the Nordic Institute for Interoperability Solutions (NIIS), a joint organization founded by Estonia and Finland in 2017.

Now let’s understand what Estonia’s X-Road got right.

What I find most interesting about Estonia’s approach is that it doesn’t require every organization to put its information into one giant database. Instead, organizations keep control of their own data and exchange information through a shared framework.

X-Road provides the technical infrastructure for secure communication. A broader governance structure sets the rules for joining and participating in the network. No organization has to abandon its existing systems or consolidate its information in one place. That distinction matters.

We often approach modernization as though the answer is to replace or centralize everything. But in government, where agencies have different responsibilities, legal authorities, funding structures, and technology environments, that isn’t always realistic or desirable.

A more useful question is: How do we let organizations work together without asking them to give up control of their systems and information?

That was one of the ideas we wanted to test with our prototype.

What we learned when we actually built an X-Road prototype

I’ll be the first to admit that understanding an architecture on paper and actually implementing it are two very different experiences. Our prototype was intentionally small. We set up:

  • the central infrastructure that manages network membership,
  • security servers for the organizations producing and consuming data,
  • certificate management, and
  • the supporting services required to establish trusted communication.

The actual data exchange we wanted to demonstrate was relatively simple. Getting everything around it working was not.

We spent a significant amount of time configuring certificates, managing security requirements, troubleshooting the environment, and figuring out how organizations were registered and approved to participate. At times, it felt like we were spending far more effort establishing trust between systems than exchanging any actual information.

And we were.

Initially, that seemed like a lot of overhead for something that should have been straightforward. But it led us to an important realization:

If you’re connecting two internal systems that already operate under the same authority, much of that overhead may be unnecessary. But if you’re connecting dozens of independent organizations, each with different responsibilities, security requirements, and legal obligations, the situation changes. You need a way to establish trust, control access, and demonstrate accountability without negotiating a different approach for every connection.

That’s where the value of a shared governance framework becomes clear.

That doesn’t mean every challenge we hit was justified. Some of the tooling and administrative processes were more cumbersome than they needed to be, and better automation and operational practices could reduce that burden.

But the experience reinforced something I strongly believe about technology decisions: we have to understand what problem we’re solving before deciding whether a solution is too complicated or too expensive.

Why data sharing matters even more for AI in government

This is where the conversation gets particularly interesting today.

When we first explored X-Road, the focus was secure data exchange between organizations. Now we’re talking about AI assistants, intelligent automation, and AI agents that can work across multiple systems and perform tasks on behalf of users. Even Estonia now frames its next step as an “agentic state” built on top of its data exchange foundation.

I keep coming back to the same question: If we haven’t figured out how organizations should securely share and govern information, what happens when we put AI in the middle of it?

Imagine an AI assistant helping someone apply for government benefits. Instead of asking that person to repeatedly provide employment information, income records, or other documentation, the assistant could help retrieve authorized information from the right systems. From the user’s perspective, that could make the experience much easier.

But behind the scenes, the agency has to be able to answer six questions:

  1. Does the assistant have permission to access that information?
  2. Is it using the authoritative source?
  3. Can we explain where the information came from?
  4. Can we audit what was accessed?
  5. What happens if two systems have conflicting records?
  6. Who is accountable for the decisions and actions that follow?

AI doesn’t make those questions disappear. On the contrary, it makes answering them more urgent. Federal policy points the same way: OMB’s April 2025 memo on federal AI use, M-25-21, directs agencies to speed up AI adoption while keeping strong safeguards for privacy and public trust.

We can build an impressive AI interface, but if the underlying information is fragmented, unreliable, or inaccessible under appropriate governance, we’ll still struggle to deliver meaningful value.

That’s why I distinguish between using AI to make an existing process faster and using technology to rethink the process itself.

If we simply add an AI assistant to a process that still requires someone to submit the same information to five different agencies, have we really solved the underlying problem? Maybe we’ve made it a little more convenient. But we haven’t made it better.

What does this mean for the U.S. government?

It’s tempting to look at Estonia and ask why we can’t just implement something similar here.

But the United States is a very different environment. Federal, state, and local governments have different authorities, funding mechanisms, privacy requirements, and existing technology investments. Even within the federal government, agencies operate under different missions, regulations, and organizational structures.

Technical interoperability standards and data-sharing initiatives already exist, but implementing a shared approach across organizational boundaries remains hard. In many cases, the hardest questions aren’t about whether systems can communicate. They’re about:

  • who has the authority to set the rules,
  • who operates shared infrastructure,
  • how participation is governed, and
  • how those arrangements survive changes in leadership, priorities, and funding.

That’s why I don’t think the answer is to replicate Estonia’s implementation. We should learn from the principles behind it.

Four principles U.S. agencies can borrow from X-Road

  1. Organizations keep ownership of their information. No giant central database.
  2. Access is controlled and auditable. Every request is authorized and logged.
  3. Trust isn’t reinvented for every integration. One shared framework replaces one-off agreements.
  4. People stop acting as the go-between. Residents shouldn’t have to carry information between agencies that are already authorized to exchange it. Estonia calls this the once-only principle.

There are opportunities to apply these principles incrementally, especially where multiple agencies already coordinate around a common service or outcome. Benefits eligibility, identity verification, and employment information are good places to start.

The key is to start with a real problem and the organizations responsible for solving it, rather than starting with a technology platform and looking for somewhere to use it.

When does X-Road make sense, and when doesn’t it?

Our research didn’t conclude that X-Road is the answer to every interoperability problem. Quite the opposite.

SituationBetter fit
A few systems inside one organization, under one authorityA simpler point-to-point or API integration. Cheaper and easier to maintain.
Many independent organizations, different legal responsibilities, long-term data-sharing relationships, strong accountability needsA shared trust framework like X-Road. The extra complexity pays for itself.

The important thing is understanding where the additional complexity creates value.

During our research, we also explored what a practical business case might look like. Rather than connecting an entire ecosystem at once, we considered a staged approach:

  1. Visibility: Give participating organizations better visibility into information they already exchange.
  2. Reuse: Introduce a service that reduces how often people must submit the same information.
  3. Expansion: Grow into broader cross-organizational capabilities where governance and the business case support it.

Each step should deliver something useful on its own, like vertical slices. That’s an approach I believe in regardless of the technology.

We don’t need to build everything before we can demonstrate value. Just identify the right problem, define what success looks like, and deliver improvements incrementally while learning along the way.

The question we should be asking

Going back to my name change, the challenge went beyond outdated systems. I had become the link between organizations, responsible for carrying the same information from one to another. It’s a familiar example of how disconnected government systems create inefficiencies for the people who depend on them.

I had to figure out who needed the information, provide it repeatedly, and make sure everyone eventually updated their records. And that is still how far too many government services work.

We have better technology today. We have more mature approaches to interoperability. And we have AI capabilities that create real opportunities to rethink how people interact with government.

But none of those advances automatically solve the underlying challenges of governance, trust, and accountability. The more capable our technology becomes, the more important those foundations are.

For me, the biggest lesson from our X-Road research comes down to three questions we should ask before choosing a technology:

  1. What problem are we solving?
  2. Who needs to work together?
  3. What has to change for the solution to deliver value?

Sometimes the answer will involve a new platform. Sometimes it will involve better integration. And sometimes the most important work will have very little to do with technology at all.

That’s the kind of thinking we bring to our work at Flexion: looking beyond the immediate technical fix, understanding the broader system, and helping organizations make practical decisions that lead to meaningful, sustainable improvements. Because ultimately, the goal is bigger than connecting more systems.