Modernization Has to Start With the Problem

Roshni doesn’t see federal agencies struggling because they don’t understand the need to modernize. In many cases, leaders know what needs to improve. The difficulty is that almost everything they are trying to change is interconnected.

Modernizing a system affects processes. Processes affect people. Data determines what can be measured. Governance determines what can move forward. And workforce changes can eliminate institutional knowledge just when organizations need it most.

“I think they have the strategy and they have the will, they have the mission,” Roshni says. “But it’s all those different components. And they’re siloed to some extent.”

That is why one principle sits at the center of her approach:

“Start with the problem, not the technology.”

A veteran may be struggling to access care. A clinician may be navigating an overloaded schedule. An employee may be trying to understand benefits. An agency may struggle to determine which information is accurate.

Those are the problems. Technology may be one part of the solution.

Too often, Roshni says, organizations reverse that sequence. They begin with the technology they want to implement and then try to determine where it fits. Meanwhile, the less visible work required to make that technology useful, such as workflow redesign, change management, policy, governance, and adoption, gets squeezed out.

Especially in a constrained budget environment, she believes agencies should also look harder at investments they have already made. Enterprise platforms initially purchased for one purpose may now have capabilities that can address additional needs.

“What are you trying to solve? What do you already have? What part of that technology do you really need? And then make your investments.”

The point isn’t to resist new technology. It is to make sure modernization produces a better outcome, not just a newer system.

AI Makes Human Knowledge More Valuable

AI intensifies the need to get the fundamentals right.

Data has to be usable. Ownership has to be clear. Privacy and governance have to be established. Leaders need to understand what information is entering a system and how its output will be evaluated.

“You can’t just say, ‘I put AI in,’ and that’s your measure,” Roshni says. “What really got accomplished?”

She also wants leaders to move beyond hypothetical use cases. It is one thing to demonstrate what AI could do. It is another to show that it has worked against a similar problem, with a similar population and under similar operating constraints.

But perhaps her strongest concern is the assumption that AI can simply replace the people who understand how these organizations work.

“AI is only going to go as far as the people that are directing it.”

Someone still has to interpret what AI produces, validate it, and recognize when a recommendation that appears logical does not fit the realities of a particular agency. That judgment often comes from years of institutional and operational experience.

Roshni sees a related risk at the beginning of the talent pipeline. AI can undoubtedly accelerate some work traditionally performed by junior employees. But that work has also been part of how people learn.

“If you replace all of that work, at some point you’re not going to have good middle management,” she says. “They weren’t trained.”

The opportunity is to develop people differently. Instead of spending enormous amounts of time gathering information or producing routine deliverables, employees can use AI to get to information faster, and then learn how to question it, interpret it, and solve more complex problems.

In that sense, AI doesn’t eliminate the need for human capability. It changes where that capability creates value.

Move Faster by Testing Smarter

Roshni also pushes back on the idea that federal agencies should simply operate at the speed of the private sector.

“I don’t think the government should be working at the same pace as the private sector,” she says. “The government needs to be deliberate and very aligned to what its mission is.”

Government operates with public money, statutory requirements, and responsibilities that require checks and balances. In areas such as healthcare, defense and public safety, moving faster cannot mean abandoning safeguards that exist for a reason.

One way to reconcile innovation with accountability, Roshni believes, is to use pilots much better.

A pilot should realistically test how a solution will operate in the environment where it eventually needs to work.

“It’s the full picture,” she says. “It’s basically a microcosm of what you’re going to do, but at a lower cost and not at the larger scale.”

That means testing more than the technology. Does it connect with existing platforms? Is the data usable? Does it fit the workflow? What governance or policy questions emerge? Most importantly, does it improve the outcome the agency set out to address?

It might mean combining expertise in technology, data and analytics, digital transformation, healthcare and finance around a specific mission problem, and then testing the solution before asking the agency to make a much larger commitment.

If it works, scale it. If it doesn’t, leaders have learned that before spending significantly more time and money.

The Bottom Line

Public sector modernization will be defined by whether that technology makes the mission work better.

That requires leaders to understand the problem before choosing the solution, preserve the human knowledge AI depends on and create practical ways to test innovation without sacrificing the accountability government requires.

As Roshni puts it: “We can’t just think that technology is going to solve everything.”