Episode 3: Stop Planning, Start Doing — with John Hrastar
Most federal organizations have AI working groups, pilots, and task forces. What they don’t have is movement. In Episode 3, host Jack Moore sits down with John Hrastar — Evans’ longtime CEO advisor — to tackle the gap between studying AI and actually using it. John’s take: stop waiting for perfect guidance. Sort your work into three buckets, clarify what your organization is trying to accomplish, and give your people permission to act.
Host
Jack Moore hosts Progress Over Perfection, Evans’ podcast for federal leaders navigating modernization, workforce change, and the realities of moving programs forward without perfect information.
Guest
John Hrastar is a CEO advisor, former business owner, and interim executive who now works primarily with company owners on building transferable enterprise value. He is a founding member of the Exit Planning Exchange, DC chapter, and a founding partner of the Greater Washington Family Business Alliance. John has advised Evans’ leadership for over a decade and is the kind of advisor who tells you the thing you’re not quite ready to hear.
Listen on
YouTube · Apple Podcasts · Spotify
About this episode
Federal agencies are running AI pilots. They’re standing up working groups. They’re attending briefings and building centers of excellence. And most of them are still in the same place they were 18 months ago. The problem isn’t access to tools. It’s the gap between activity and action — and that gap is a leadership problem, not a technology problem.
Jack Moore brings in John Hrastar, who has spent his career advising organizations through exactly this kind of transition, to make the conversation concrete. They work through a practical framework for sorting work into what AI should own, what stays human, and what disappears entirely — then translate it into three things any federal leader can do starting Monday to get their team moving.
What Jack and John cover
- How do you figure out which work AI should take on and which has to stay human? John’s three-bucket framework sorts every job into: what AI can do on its own, what still requires meaningful human input, and what disappears entirely once AI is involved. The third bucket — work that just evaporates, like status updates and coordination loops — is often the biggest surprise for leaders.
- What happens when you optimize for the wrong thing? John walks through the Klarna example: the company replaced human call center agents with AI, saved tens of millions of dollars, and then watched other key metrics crater because they’d optimized for call volume, not customer relationship. The lesson for federal leaders — just because AI can do something doesn’t mean it should.
- How do you get people to actually use AI instead of waiting for permission? Jack shares an internal Evans story: a team was given a list of 31 processes to consider automating. Instead of waiting for a ranked priority list, one employee built the tool over a weekend — because the direction was clear and the permission to act was implicit. John’s point: you don’t find people like that. You create the environment where they emerge.
- What are three concrete things a federal leader can do starting Monday? First, run a job dissection exercise — have each person go through a structured AI interview to assess how their work splits across the three buckets. Second, dust off the mission statement and make it clear enough that no one can misunderstand it. Third, create the environment: tools, training, direction, and explicit permission to experiment. John offers to share his job dissection prompt directly — find him on LinkedIn.
- What separates leaders who actually transform their organizations from those who just keep talking about it? John’s answer is direct: bias for action. Don’t wait for the perfect policy, the perfect training sequence, or the mountaintop guidance. Set the goal, give people guardrails, and start. The organizations that transform are the ones where someone, somewhere, does something concrete by a specific date.
Why it matters now
The episode doesn’t cite external statistics — and that’s intentional. Jack and John’s argument is that most federal leaders already know the data. What they’re missing is the decision to move.
One signal from the conversation: at one Evans federal client, more than half of the program management structure took early retirement or a DRP in a single cycle. The leaders left behind aren’t waiting for headcount to recover. They’re learning to do more with AI — or they’re falling further behind.
Related resources
- Between Prototype and Procurement: What It Takes to Scale Federal AI — Jesse Lambert
- Abundance in Action: Applying AI Maturity to Public-Sector Modernization — Jesse Lambert
- Mission AI: Where Modernization Gets Real — Jesse Lambert
