Beth Diaz is used to complexity. As Politico’s chief information officer, she’s tasked with modernizing a legacy media organization that operates all over the world, under the intense scrutiny that comes with covering politics in real time. But these days, she’s also helping shape the company’s adoption of AI.
In this conversation, Diaz explains why she sees AI as a way to free up time, amplify human judgment, and deliver better insights to readers without losing sight of the critical role of the press in society, which Politico’s founders, editors, and reporters consider paramount. For founders building in high-trust sectors, her perspective offers a nuanced view of how to navigate ambition, ethics, and AI adoption in one of the most under-the-microscope industries out there.
Q. You’ve said that AI should “automate tasks, not automate jobs.” How does that philosophy shape your team’s approach at Politico, especially in a newsroom where trust and human judgment are foundational?
Beth Diaz: Every industry is being impacted by AI, but as I like to say, media is both a supplier and a user of AI: We're powering the models, and we're also using AI to further our business. So it's interesting to see how different media companies and organizations are approaching this.
At Politico and on the team level, we’re focused on what we can do to leverage AI to further the business while still respecting the core ethos of journalism. Maybe we'll get there someday, but right now, I 100% trust the editorial judgment of our founder and executive editor, John Harris, to make decisions about what's most important for people to know, over some sort of recommendation engine or algorithm. That being said, there's a lot that AI can do to take on some of the hard work that’s too much for individuals to do.
We know that our reporters, our journalists, our editors, and our product managers just have better insights than AI. And this is what people pay us for both in their time as well as their money. So let's clearly demarcate the areas where there are opportunities, like taking 1,000-page government documents and summarizing them, so that our reporters can interpret it.
Q. You evaluate new tools and help define the organization’s AI strategy. What advice do you have for founders pitching enterprise AI products to media companies like Politico?
I'm very appreciative when I'm talking to vendors and software service providers that are actually asking me questions and not making assumptions. There’s a difference between asking informed questions and saying, “Here's this 48-page form to fill out so I can tell you why you need to buy from us.” And also… I’m going to be blunt: Treat me with some respect!
I tend to come from the position where I identify a need, and then I start to look to see who's out there. I can imagine it's very challenging for startups working with a relatively conservative industry like media. But I think that figuring out what I need first, and then asking how a vendor can meet that need with a particular eye to security and privacy, is the right way to go about it.
Q. Politico’s audience includes policy professionals who need fast, accurate insights. Where have you found the biggest wins using AI to serve that mission without compromising journalistic responsibility?
There is just such a massive amount of publicly available data, regulatory data, legislative data. The ability to take all of this publicly available information and synthesize it in a way that’s useful for people is a key opportunity. I just want to make it very clear that this is different from reporting on things.
One of our other challenges is to summarize information and see what's resonating with our audiences, and it's constantly evolving. We take all of the reporting that we've done over the past month, or quarter, or six months, and ingest it, both in terms of the details of the story as well as its performance. And then we do a vector analysis and look at summarized results so we can see what people are reading and engaging with. I've spent years doing this — more time than I care to think about! — looking at thousands of individual articles and hand-coding topic areas, which now we can do in just a matter of minutes.
Q. What are some of the boundaries you’ve drawn — places where AI shouldn’t be used — and how do you communicate those internally in a way that builds trust rather than fear?
At Politico, we’ve got a bunch of incredibly smart people who are literally in the business of finding out what's going on. And so taking the approach of “avoid talking about AI because people might not notice it” is definitely not going to work. And so the approach at Politico has really been, “Let's get in front of this.”
We work to identify both what the opportunities are, how we can go about automating tasks, improving productivity, and giving greater insights to our readers and subscribers, while at the same time identifying areas that just aren’t something we're going to be doing.
Q. We’ve covered how AI can help a news organization, but what about the opposite? Do you think that there’s any weight to the anxiety that AI will replace real journalism?
The years of experience and the capabilities of our journalists cannot be substituted by something that can take mass volumes of information and summarize it. You need somebody who’s on the ground talking to the people that are actually making these regulations and who can identify what's coming down the pipeline that you haven't seen yet. What’s the analysis on the potential implications and potential next steps?
People who are coming to Politico know that they're coming to us and paying for the services we provide because of that.
