Back to the Analog: AI and the Future of Public Diplomacy
What follows is a lightly edited version of my conversation with Robert Ogburn on Public Diplomacy Examined, a podcast from the Institute for Public Diplomacy and Global Communication at GW’s School of Media and Public Affairs, produced by Kassidy Corey. We recorded this conversation back in February 2026. You can listen to the full episode here.
Robert Ogburn: You’re listening to Public Diplomacy Examined, where we explore how nations communicate, connect, and compete for hearts and minds in the digital age. Today we’re diving into one of the biggest challenges and opportunities facing global diplomacy: artificial intelligence. Is AI the end of human-centered diplomacy, or could it actually make public diplomacy more effective, more empathetic, and better informed?
Joining me is Paul Kruchoski — former Senior Executive Service member, senior diplomat, and Director of the Office of Policy, Planning and Resources at the U.S. Department of State, effectively the chief operating officer for public diplomacy operations, where he oversaw nearly 5,000 staff, more than 170 embassies, and a budget of around $1.5 billion. Paul is now a director at Guidehouse and a leading voice on AI’s role in foreign affairs.
Now let’s start with the big picture. When you look at where PD is right now, how is AI already beginning to change the way diplomats listen, learn, and engage with foreign publics? And where do you think we’re still stuck in an analog mindset?
Paul Kruchoski: Let me flip the last part of that on its head, because I think the heart of AI in diplomacy is actually an opportunity to focus harder on the analog things — the parts that are central to what diplomacy is, which is people connecting with other people. AI isn’t going to change that. It may change how we prepare for those conversations, how we run our embassies. But diplomacy is still going to be about connecting with people.
We’re in a fascinating moment. State and non-state actors are pumping misinformation and AI-generated content into the internet, and it’s alienating people quickly. People aren’t reacting well to AI personas on Instagram or synthetic content across the other platforms, and it’s driving them off. 2025 saw some of the sharpest drops in social media use, most dramatic among 18-to-24-year-olds. Where are they going instead? To closed channels, to spaces with people they know and trust. That tells me something about the years ahead: the human relationships are going to become more central than ever.
What diplomats are already doing with AI
Robert: You’ve been interviewing practitioners about how they actually use AI today. What surprised you?
Paul: One desk officer in a regional bureau described how their entire media workflow had changed, soup to nuts. They were using projects in StateChat, commercial ChatGPT, and Gemini to pull in large volumes of media — from something like North Star, State’s media analytics platform — analyze it, strip it down to the gist, and use that for media planning and a first draft of talking points, which they then feed back into the system.
Anyone who’s done media work knows how time-consuming each of those steps is. At an embassy you might have two or three people spending most of their day on each one. Condense those down and you free people up to do the two things long-time diplomats have always struggled to protect time for. First, in-person contact work — there’s no way to sit with a group of journalists or sensitive contacts from behind a desk in Washington. Second, the strategic planning and the monitoring and evaluation we’ve talked about for over a decade but never staffed. Given that flight to trusted spaces, we’re going to need diplomats doing that human work more than ever.
Individualized AI, not one-size-fits-all
Robert: Capturing what happened at an event or engagement has always been a labor problem — logging that a hundred people came to a lecture, what they heard, who was in the room. Can AI help with that?
Paul: This is where agentic and individualized AI agents get powerful. One of the ways people are using AI out in the private sector now is not an enterprise AI application that’s the same for everyone — it’s a personal setup people have with connections to your data systems, your credentials for other systems, where it can go in and do tasks on your behalf. Diplomats work the way people work — we each have our own way of taking notes and sharing them. A one-size-fits-all enterprise tool won’t get the same adoption as something more idiosyncratic that still plugs into the enterprise systems of record.
Some people call this “identity AI.” I’m not a fan of coining new terms for their own sake, but individualized, personalized agents are going to be a huge part of what works for diplomacy — because they draw on the natural patterns the work already has. Picture an agent that pulls your meeting notes, lets you photograph something and distills it into text, and uploads it into a CRM to share with others. Now the enterprise software works for you, instead of you working for the software. You certainly need systems of record to put all that into, but the model is entirely broken where we’re asking officers and local staff around the world to put a bunch of data in, at a cost of tens of thousands of dollars, just to get basic insights out.
What the private sector already knows
Robert: You mentioned CRM. Are you gaining insights from what the private sector is doing and trying to encourage PD practitioners to use those?
Paul: Over the last twenty-five years, sales, marketing, and business process have quietly converged with diplomatic practice. Marketers spend their time on experience — how do you build events that change how people feel about an organization? That sounds a lot like public diplomacy. The difference is that they operate inside a multi-trillion-dollar industry with a clear profit motive, so they’re innovating constantly, and we can borrow from it.
Think about the discrete signals: what someone told us in a meeting, how that correlates with traffic on our website, with the emails they’ve opened, with the events they attend. We don’t have the capacity to analyze all of that at State. AI systems start to give you the ability to look at the aggregate data available to an enterprise and recommend what you should do next to move that relationship or contact forward. A human still has to do it — an AI can’t — but you start to get some of those aids in building better diplomatic relationships and aiding the work you do in negotiations and everything else.
A hopeful 2030
Robert: Look ahead five years. What does public diplomacy in Washington and at an embassy feel like near 2030?
Paul: I’ll give you the hopeful version, because nothing here is fixed. Most of the administrative work is gone, or can be — travel vouchers, document processing, the ServiceNow forms you send into the ether after a dozen slow clicks. Our overseas officers cost us about $400 an hour. That’s real money and real time, and it isn’t fungible. I think we can offload 80-plus percent of the department’s administrative work to agents and automation.
For the data-heavy work, everyone has an individualized agent that plugs into a broader set of agents and data sources. On your way to a meeting: “pull together the facts and figures I need, and the recent press statements my counterpart has made on this issue.” It’s distilled into a single briefing, drawing from intel and everything else. We’re not far from that. And in a negotiation, you can query those sources for recommendations and for the second-order consequences of a move. That personal AI layer, banded with a data layer, lets people focus on the diplomacy and the relationships they haven’t been able to protect.
Accountability comes first
Robert: So how do we use AI to engage with audiences without losing the human element?
Paul: Two principles guide me. First, machines can’t be held accountable, so they shouldn’t be allowed to make decisions — that line comes from an old IBM training handbook and it holds up. The second: if you’re going to be accountable for the decisions a machine is recommending, you are making those decisions — you have to understand, rely on, and trust the information it’s producing for you. Given that they produce false results, you have to know enough about what you asked it to do to go back and audit it yourself. If you can’t do that, you probably shouldn’t be using AI for that task.
That gives you guardrails. When you get into the media question — should I allow AI to go direct to social, putting out posts? No, because it can’t be responsible for the consequences of what happens in the real world. And if you’re asking it to generate content on a topic you don’t know anything about, who’s going to be accountable when that goes out? It’s you. At the end of the day, people are still accountable for communication, at least in our US and Western system. Not everyone feels that way. If you take a look at how Russia and China are already using AI on the internet, they’re using it to pump trash out into the information ecosystem. That’s a reality, and we have to think really hard about how we push back — identify tactics that make that less and less attractive over time. But the public has spoken. They recognize the synthetic content, and they’re leaving.
A council of advisors
Paul: Most of us treat AI as a machine for generating variations on an idea we already have. One of my favorite tools I’ve built for myself is a variation on Edward de Bono’s Six Thinking Hats idea. Edward de Bono is a facilitator and strategist who came up with these six thinking hats, each representing a different way of looking at a problem — looking at something from the position of feelings, from critical logic and risk, from creative alternatives. What I have running in the background almost all the time is a little variation of that. When I send it an idea I have, long or short form, it takes it and runs five or six little separate agents that each look at the idea from a different perspective, and then come back to me like a group of counselors.
Let me tell you, they are mean to me on a regular basis. They’re critical in consistent ways, they talk to each other, they gang up on me — which is perfect. Instead of an echo chamber of me and a bot telling me my half-baked idea is wonderful, I get an epistemic check that tightens my thinking. When you think about diplomats and negotiations, the ability to look at the same set of facts or talking points from multiple different perspectives opens up a whole new universe of how you can use these tools to get better and smarter, rather than just produce more.
Robert: In business school I read Reframing Organizations, which took one case and presented it through several different management lenses. This seems like a natural fit for that.
Paul: It is. The two basics of agentic AI are that it can take multiple steps toward an objective, and that it can spawn sub-agents — each with its own context, its own perspective to represent. A council of advisors with radically different views, condensed into a short window. Running something like that takes me maybe 20 minutes to read, another 10 or 15 minutes to digest and think about. Compared to what we would have done with a country team, with a front office in a bureau, it’s light-years ahead of where we were. It has the ability to accelerate the pace of work and, again, get everyone to the highest-value version of what they could possibly do as quickly as they can.
The money question
Robert: I notice we haven’t hit the brick wall PD people always hit first: budget. Is this affordable for government?
Paul: Even the commercial versions are. I remain an enduring fan of what Anthropic is doing with Claude; the top-tier subscription is $200 a month per user. Set that against $400 an hour for an American officer overseas. Local staff cost less hour to hour, but even at $50 or $60 all-in, it takes only a couple of hours a month to break even. The workflows I described — the full media cycle, drafting and synthesizing grants material, keeping a grants file current — those run to tens of hours a month. The money is there.
Robert: For listeners: the grants process isn’t glamorous, but it’s essential, and everyone who’s touched one has wished for a team to help write it well.
Paul: You can hear PD staff around the world groaning at the word “grants.” It’s a paper file two or three inches thick before the grant even starts. You can reuse language from past iterations, but someone still has to verify all of it. Anything that gets people back to the strategic use of their time feels like a gift. We’re in a moment where diplomats get to decide what kind of future they want — one where AI slowly replaces them, or one where they use it to move to higher-value work and prove what human connection can do.
What AI can’t do — and shouldn’t
Robert: So the future isn’t a humanless office.
Paul: No. I hope it’s an office where people are freed to think big thoughts and guide the AI while keeping the ultimate strategy in view. And I don’t think genuinely novel strategy can be written by an AI, for two reasons that should give hope to anyone who wants to be a diplomat.
First, large language models converge toward an average — of their training data, their weights, your prompt, your documents. A human has to decide what goes in and how to push the trajectory somewhere other than average. Human creativity hasn’t died in the age of AI; it matters more than at almost any point I can name. Second, there are signals AI reads poorly: the nuance of a person-to-person conversation, body language, the small shifts in a room, what a large group does together. We’re years from any system doing that well. In an international negotiation, in the tradecraft of building relationships, the machine can’t read the fine signals or sense the trajectory. It can do all sorts of other things, so let’s let it do that. Let’s not let it take away, or try to do, those things it’s not well designed for — and that frankly are the most fun parts of our job anyway.
Robert: So not two agentic bots from two countries negotiating a treaty on their own?
Paul: Let the humans set the goals and the parameters. Then, sure, let the machines help us work the terms toward them. Let the humans decide where we’re going; let the robots help us get there.
Listen to the full conversation on Public Diplomacy Examined (SoundCloud).
Paul Kruchoski is a director at Guidehouse and the former Director of the Office of Policy, Planning and Resources at the U.S. Department of State. This conversation aired on Public Diplomacy Examined, produced by the Institute for Public Diplomacy and Global Communication at the George Washington University School of Media and Public Affairs.