This episode of Newsroom Robots is supported by The Lenfest Institute for Journalism.

At the Seattle Times, an AI tool identified a new prospect and helped close additional ad revenue in a single day.

The Seattle Times is one of 11 news organizations participating in the Lenfest AI Collaborative and Fellowship Program, funded by OpenAI and Microsoft. Through this initiative, AI engineering fellow Rajesh Barade worked with the advertising team to address their challenges and developed a prospecting agent. This tool enables sales representatives to enter a category or focus area, identify businesses spending in that market, find decision-makers, access advertising spend data, and generate personalized outreach.

The Seattle Times deliberately began its AI efforts in advertising rather than the newsroom. Early successes helped build momentum across the organization. This led to listening tours across finance, HR, the audience, and eventually the newsroom, where a municipal meeting listening tool is now in development.

This week on Newsroom Robots, I speak with Kati Erwert, Senior VP of Product Marketing and Public Service at The Seattle Times, and Tristan Loper, Head of National Programs at the Lenfest Institute, to explore how the prospecting agent works, why starting in advertising was key to unlocking organization wide adoption, and what it takes to turn early AI successes into lasting capability.

In this episode, we cover:

02:07 — How The Seattle Times built internal momentum for AI adoption

05:51 — What the Lenfest AI Collaborative is and how it works

09:29 — Inside the prospecting agent: how a sales rep closed business in one day

14:11 — Why starting with advertising drove cultural change across the organization

17:07 — The buy vs. build decision in a fast-moving AI landscape

19:03 — How open-source code sharing works across the 11-newsroom cohort

24:44 — Building the municipal meeting listening tool and beyond

37:07 — Navigating the tension of blocking AI bots

45:47 — How smaller newsrooms can get started with AI today

🎧 Listen to the full conversation with Kati Erwert and Tristan Loper on Apple Podcasts, Spotify or your preferred podcast platform.

P.S. I’ll be in Berlin, Germany, and Vienna, Austria, later this month—reply if you’d like to meet up. But first, I’m headed to the International Journalism Festival in Perugia, Italy, next week. A few things I’m looking forward to:

If AI is the new intern, what’s the new on-ramp into journalism? — I’m moderating this panel with Astrid Maier (dpa) and past podcast guests Jeremy Caplan (City University of New York) and Djordje Padejski (Stanford University). Friday April 17 at 10:30am.

dpa AI Soirée – Bring Your Own AI Problems! — Join me and Astrid Maier from dpa, Germany’s leading news agency, for an evening of real-world AI problem-solving. Bring your messiest challenges, and we’ll talk through them together. Spots are limited, so register now to secure yours. Friday, April 17 at 6:30 PM.

Steal this: AI projects newsrooms actually shipped and how they got it over the line — Loved this conversation? See it live. Kati and Tristan will be joined by Emily Brown (Microsoft) and Kevin Hoffmann (The Philadelphia Inquirer). Saturday, April 18 at 9:30 AM.

Share

For years, the aggregator model was simple. Curate the best journalism from thousands of publishers and send audiences their way. Now that contract is being rewritten and Yahoo News is one of the most interesting places to watch it happen.

In this episode of Newsroom Robots, I spoke with Kat Downs Mulder, Senior Vice President and General Manager of Yahoo News, about how the platform is layering AI across every surface of a product that reaches over a 180 million people in the U.S. alone each month. Kat previously spent more than 14 years at The Washington Post as chief product officer and managing editor before taking on the challenge of modernizing one of the internet’s original news destinations.

Yahoo’s purchase of Artifact, the AI news app built by Instagram’s co-founders, gave the platform a new recommendation engine that prioritizes time spent reading over clicks. Its new AI answer engine, Yahoo Scout, synthesizes information with rich citations and visual context. Its AI-powered daily audio digest turns personalized news into a listening habit. Each of these products makes Yahoo more useful to its audience.

Subscribe now

But each of them also changes the relationship between Yahoo and the publishers whose journalism powers the platform. When Scout can give you the answer you need without a click-through, when an audio digest summarizes a story so well you never open the article, when personalization makes Yahoo the destination instead of the pass-through, the old aggregator math stops working for publishers. I pushed on this in the conversation, and Kat was candid. The compensation models haven’t been figured out yet. Yahoo is working with the Microsoft Publisher Content Marketplace to develop new economics, but the industry is still writing those rules in real time.

She makes a strong case for how Yahoo is approaching this differently, from how Scout prominently surfaces publishers, to the rev-share model they operate, to why she believes the quality flywheel they are building actually rewards better journalism.

Kat believes original, distinctive journalism will become more valuable in an AI world because AI agents will seek out what is unique. The question is whether the economics will catch up to that belief before publishers run out of runway.

This conversation raises a bigger question about the future of the aggregator model: what it owes the journalism that powers it, and what comes next.

In this episode, we cover:

02:00 — Why Yahoo acquired Artifact and how it shifted recommendation algorithms

06:20 — The shift from click-based metrics to deeper engagement signals such as session time and retention

08:50 — Inside Yahoo Scout, Yahoo’s new AI answer engine built to support publishers and the open web

14:10 — The changing economics of news as AI platforms begin generating answers instead of sending traffic

17:40 — Yahoo’s personalized AI-generated audio news digest and why multimodal news experiences matter

22:00 — How Yahoo combines human editorial judgment with algorithmic personalization

31:00 — How AI is transforming newsroom product development and prototyping

36:10 — The tension between personalization and journalism’s civic responsibility

41:30 — What smaller newsrooms can learn from the AI product playbook’

🎧 Listen to the full conversation with Kat Downs Mulder on Apple Podcasts, Spotify or your preferred podcast platform.

Subscribe now

AI is settling in as infrastructure within newsrooms, a layer quietly reshaping how journalists discover information, how stories move through production, and how audiences increasingly expect news to reach them.

In this episode of Newsroom Robots, recorded live in New York City at TV News Check’s News Tech Forum, I brought together four industry leaders to examine the tangible ways AI is transforming newsroom operations. The conversation featured Ryan Struyk, Director of AI Initiatives at CNN, Rubina Madan Fillion, Associate Editorial Director of AI Initiatives at The New York Times, Arlyn Gajilan, Global Editor of AI Development and Integration at Reuters and Burt Herman, Co-Founder and Principal of Hacks/Hackers.

Despite representing fundamentally different newsroom models—broadcast, digital subscription, wire service, and local news—they came together to confront common questions about where their industry is headed.

The discussion focused on defining questions for the news industry: Where is AI already delivering real operational impact? How should newsrooms adapt to a world of “liquid content” and AI-mediated distribution? Is human-in-the-loop governance sustainable, or is it already breaking down? As trust in news declines and trust in AI interfaces rises, what becomes journalism’s true competitive advantage?

In this episode, we cover:

03:10 — Where AI is already embedded inside CNN’s newsroom workflows

04:25 — How The New York Times uses AI to power investigative reporting and the “Manosphere Report”

7:30 — How Reuters compressed story production from minutes to seconds and feature development from three months to three weeks

11:44 — Why Hacks/Hackers is urging small newsrooms to think from first principles before adopting AI

15:15 — The rise of liquid content and what it means when audiences reshape journalism into their preferred formats

23:24 — Why local news holds a unique advantage in an AI-mediated information landscape

29:12 — Five years from now: What newsrooms hope they get right

🎧 Listen to the full conversation with Ryan Struyk, Rubina Madan Fillion, Arlyn Gajilan, and Burt Herman on Apple Podcasts, Spotify, or your preferred podcast platform.

Most major newsrooms have now moved beyond early experimentation with AI. The main challenge now is determining how to govern effectively, scale consistently, and strategically position AI across the entire organization, while maintaining public trust as a central priority.

This week on Newsroom Robots, I sit down with Uli Köppen, Chief AI Officer at Bavarian Broadcasting (BR), to talk about what it really looks like to lead AI strategy inside one of Europe’s largest public broadcasting networks.

This marks Uli’s third appearance on the podcast. In our first conversation, we explored how BR’s AI lab was pioneering personalized audio news through its Remix Regional prototype. In our second, we caught up on how those early experiments were beginning to scale.

This time, the scope is much bigger. Uli is now responsible not only for BR’s AI lab, but also for AI implementation across the entire organization.

Uli makes a compelling case for why every newsroom should establish a dedicated AI leadership function, backed by an interdisciplinary governance structure. We also spend time on a question that is defining the next phase of AI strategy for many newsrooms: in a world of AI overviews, zero-click search, and agent-driven information retrieval, how do you maintain your brand as a recognizable, trustworthy source? Uli shares why BR opted out of AI crawling and what they are building instead, including a vision for a verified content data pool that could power new products across multiple media organizations.

In this episode, we cover:

02:09 — What it means to be Chief AI Officer at a public broadcaster

06:30 — Why every newsroom needs an interdisciplinary AI board, not just a single AI leader

09:06 — The skills newsrooms need to build for an AI-driven environment

11:00 — Why reinventing workflows starts before adding any technology

16:28 — Inside the Oktoberfest Chatbot and the collaborative content pool powering it

23:40 — Using AI for smarter community engagement and real-time moderation

26:30 — The personalized audio news briefing that users love and where it’s headed

36:00 — How BR’s AI guidelines evolved from broad guardrails to clear, example-based rules

41:40 — The strategic question: be part of AI platforms, or build recognizable products of your own?

🎧 Listen to the full conversation with Uli Köppen on Apple Podcasts, Spotify or your preferred podcast platform.

This episode of Newsroom Robots is supported by the Lenfest Institute for Journalism.

Chicago Public Media operates two distinct news brands: WBEZ, the public radio station, and the Chicago Sun-Times, the legacy newspaper. With audio and print journalism, both membership and advertising revenue, and decades of archives in multiple formats, they’re a unique case study for AI in local news.

When CEO Melissa Bell joined the organization, there was interest in AI but no dedicated resources for experimentation. Through the Lenfest AI Collaborative, they brought in their first AI engineer. A year later, Spanish translations that used to take days are now published the same day. Forty years of WBEZ audio, previously unsearchable, are being transcribed and made searchable for journalists.

A note on this week’s episode

David Chivers, who you’ll hear in this episode, passed away on January 1, 2026. He was the lead advisor for the Lenfest AI Collaborative and this episode was recorded the previous month.

David was deeply committed to building capability in newsrooms. He was generous with his time, sharp in his insights, and always had one of those big smiles that would light up a room. He will be missed.

In this week’s episode, you’ll hear from Chicago Public Media leaders Melissa Bell (CEO) and Aron Pilhofer (Chief Product and Membership Officer), along with Mark Chonofsky (AI Fellow) and David Chivers (lead AI advisor for the Lenfest AI Collaborative).

We get into the specifics of how they're thinking about AI as part of a larger membership strategy, how they decide what to build versus buy with limited resources and what it looks like to lead through a public AI failure.

In this episode:

02:55 — Where Chicago Public Media started with AI a year ago

08:08 — What AI use looks like inside the newsroom

15:42 — How product development is evolving with AI tools

27:28 — Collaboration with OpenAI and Microsoft

28:26 — How AI fits into Chicago Public Media’s membership strategy

36:05 — Build vs. buy with limited resources

37:44 — The Chicago Sun-Times AI-generated book list incident

42:18 — Advice for leaders navigating AI mistakes publicly

🎧 Listen to the full conversation on Apple Podcasts, Spotify or your preferred podcast platform.

This week on Newsroom Robots, I sit down with Alessandro Alviani, Lead for Generative AI, and Fabian Heckenberger, Managing Editor for AI, at Germany’s Süddeutsche Zeitung to discuss how they’re using AI to build the next generation of news products.

This conversation looks at what happens when AI becomes a permanent layer in a newsroom’s product stack. You start thinking less about “tools” and more about systems for adoption, transparency, quality control, and learning from real usage.

Alessandro and Fabian walk through how they’re designing AI experiences that fit naturally with reader behavior and how they’re developing new distribution and accessibility formats that would have been impossible to sustain manually.

We also go deep on a topic that’s becoming a defining competency which is operational trust. What do you monitor once an AI product is live? How do you categorize failures? And how do you respond quickly when something goes wrong, without panic and without eroding your brand?

In this episode, we cover:

02:52 — How editorial and product roles complement each other in AI strategy

13:13 — Addressing skepticism and fear around AI in the newsroom

25:17 — Inside building the German election chatbot

31:10 — The design framework that signals AI content without eroding trust

35:30 — Real-time risk management and monitoring for live AI tools

48:50 — The two questions every newsroom should ask before greenlighting an AI project

54:55 — Closing reflections and personal AI use

🎧 Listen to the full conversation with Alessandro Alviani and Fabian Heckenberger on Apple Podcasts, Spotify or your preferred podcast platform.

2025 wasn’t just another year of AI experimentation in the media industry. It forced us to confront a bigger question: what happens when AI stops being just a newsroom tool and becomes the layer audiences experience journalism through?

That’s the question I keep coming back to as we head into 2026, because the stakes are shifting fast. AI is reshaping distribution, workflows, trust and the economics of publishing all at once.

On this episode of Newsroom Robots, I’m joined by Francesco Marconi and Scott Austin for a candid, high-signal conversation about what 2025 really changed for AI and news, the trends that flew under the radar, and our predictions heading into 2026.

Francesco is the co-founder and CEO of AppliedXL. Before that, he led R&D at The Wall Street Journal. Earlier in his career, he built some of the earliest AI and newsroom automation systems at The Associated Press. If you’ve been listening for a while, you might remember him from his previous appearance on the Newsroom Robots podcast back in 2023. You can revisit that episode here.

Scott leads business development at Symbolic.ai, an AI-assisted publishing tool. He is also a journalist and digital media veteran who spent years at The Wall Street Journal as a reporter and award-winning editor, leading teams that covered some of the biggest tech stories of the decade. He later led content partnerships at Dow Jones, working across major platforms and partners.

The three of us have also been collaborating on our upcoming book, The Science of First, coming out in 2026. Working on it has kept us immersed in the biggest questions about how news organizations survive and compete in an AI-mediated information ecosystem, even as the industry continues to shift in real time.

In this episode, we cover:

03:10 – Why 2025 was journalism’s operational reckoning year

08:55 – The shift from search to answers, and why it breaks old business models

14:40 – Proactive AI, and what ChatGPT Pulse reveals about the next distribution layer

20:30 – Journalism’s hidden work: why persistence, source building, and human judgment still matter

23:30 – Why news orgs must move upstream, from content to structured knowledge

36:10 – AI agents: what they actually are, what they are not, and why transparency matters

41:20 – The overlooked shift: Model Context Protocol (MCP), and why it is a major newsroom disruption

51:05 – Our predictions for 2026

🎧 Listen to the full conversation with Francesco Marconi and Scott Austin on Apple Podcasts, Spotify or your preferred podcast platform.

This episode of Newsroom Robots is supported by the Lenfest Institute for Journalism.

The Philadelphia Inquirer never had an AI engineer on staff until the Lenfest AI Collaborative & Fellowship program changed that.

The collaborative is a $5 million partnership between the Lenfest Institute, OpenAI, and Microsoft that placed 10 AI fellows in American newsrooms for two years. These engineers work within the organizations, building tools that solve real newsroom problems.

This week on Newsroom Robots, I sit down with Jim Friedlich, CEO and Executive Director of the Lenfest Institute; David Chivers, lead advisor to the Lenfest AI Collaborative and Matt Boggie, CTO of The Philadelphia Inquirer, to walk through how the program works and what the Inquirer has built as a result.

The Inquirer came to the collaborative with an idea to build a full-archive search tool that would let reporters query decades of journalism. They expected it to take 24 months. Within two weeks of a Microsoft hackathon, they had working code. The tool, now called Dewey, searches everything the Inquirer has published since 1978.

Subscribe now

In this episode, we cover:

03:02 — How the Lenfest AI Collaborative got started

05:34 — Can newsrooms trust big tech partners?

08:33 — How the fellowship works day to day

14:52– Inside the Microsoft hackathon that built Dewey in two weeks

21:37 — Training journalists to understand LLM limitations

24:07 — How AI literacy has changed newsroom culture

29:45 – How small newsrooms can get started with AI

35:14 — AI answers, search decline, and the future of audience traffic

38:15 — Rethinking journalism’s role in an AI-mediated world

41:23 — Closing reflections and personal AI use

🎧 Listen to the full conversation with Jim Friedlich, David Chivers and Matt Boggie on Apple Podcasts, Spotify or your preferred podcast platform.

P.S. I wrote a piece in Nieman Lab’s Predictions for Journalism 2026 titled “AI Will Rewrite the Architecture of the Newsroom.” It looks at how AI is reshaping news discovery and why 2026 is the inflection point for rebuilding newsroom infrastructure.

Subscribe now

What happens when a newsroom turns an internal AI solution into a global business?

That is exactly what happened at Zetland, a Danish audio-first newsroom whose internal transcription fix evolved into GoodTape, a global AI business.

This week on Newsroom Robots, I sit down with Tav Klitgaard, CEO of Zetland, to unpack the origin story of GoodTape, an AI transcription tool that began as an internal newsroom solution and grew into a profitable product used far beyond journalism.

GoodTape was not born from an AI strategy or a product roadmap. It emerged from a familiar newsroom pain point. Journalists were spending hours transcribing interviews, and existing tools fell short, especially in non English languages like Danish.

In our conversation, Tav breaks down how GoodTape went from an internal experiment to a standalone, subscription based product that quickly became profitable, generated millions in revenue, and was eventually divested. We also get into what building GoodTape taught Zetland about AI adoption, organizational learning, and where newsrooms should and should not use generative AI.

In this episode, we cover:

05:50 – How a prototype using OpenAI’s Whisper sparked GoodTape

08:36 – The moment Zetland realized GoodTape could be a real product

12:34 – How journalism’s trust and privacy standards became a product advantage

13:59 – What actually improves transcription quality beyond the model itself

15:27 – How GoodTape became profitable and contributed to Zetland’s revenue

17:36 – What building an AI product taught Zetland about newsroom AI adoption

28:14 – A real-world example of AI use that forced Zetland to rethink its own guidelines

30:34 – Why principles matter more than rigid AI rules in newsrooms

🎧 Listen to the full conversation with Tav Klitgaard on Apple Podcasts, Spotify or your favorite podcast platform.

How do you redesign a newsroom’s entire workflow when AI is no longer a single tool, but a collection of agents, voice interfaces, and ambient intelligence changing how journalism gets produced?

This week on Newsroom Robots, I sit down with Markus Franz, Chief Technology Officer at Ippen Digital, one of Germany’s largest digital media networks with more than 80 online news and media portals. We recorded this episode live at the Digital Growth Summit in Stuttgart, where Markus and I explored how his team is building some of the most forward-looking AI experiments in European media.

Markus leads Ippen Digital’s Incubator Lab, an innovation unit focused on reimagining how publishing and AI-driven experiences will evolve. With 16 years inside the company, Markus has been central to Ippen’s digital transformation and now leads efforts around multi-agent architectures and building adaptive workflows for the newsroom.

In our conversation, Markus breaks down how his lab is experimenting with multi-agent “virtual teams,” voice-first newsroom interfaces, multimodal content production and an ambient AI-powered newsroom where intelligent systems support journalists in real time. He shares what his team has learned from early prototypes, why the biggest challenges are cultural rather than technical, and how news organizations should think about guardrails, platform dependency, and the rise of self-evolving models.

In this episode, we cover:

02:22 – Why Ippen Digital built an Incubator Lab and how it’s structured as a future-focused R&D unit

04:49 – What multi-agent systems look like inside a newsroom

9:42 – The case for voice as the next major interface for both journalists and audiences

14:41 – The shift from human-in-the-loop to human-on-the-loop workflows

17:40 – Guardrails for agent systems: grounding, bounding, editorial policies

19:33 – The vision for an ambient newsroom powered by AI companions and real-time intelligence

27:31 – Why vendor lock-in and self-evolving LLMs pose new strategic risks

30:08 – Multimodal personalization and rethinking how news is experienced

34:27 – Why most AI pilots fail and what experimentation looks like in practice

49:19 – Markus’s personal AI stack and how he uses these tools day-to-day

🎧 Listen to the full conversation with Markus Franz, recorded live in Germany, on on Apple Podcasts, Spotify or your favorite podcast platform.

How do you bring AI into a newsroom as big and globally distributed as the BBC, an editorial network that stretches across 42 languages and more than 5,000 journalists?

This week on Newsroom Robots, I sit down with Olle Zachrison, Head of News AI at BBC News, where he leads the BBC’s efforts to advance AI use and strengthen its journalism and audience experiences. Previously the Head of AI at Swedish Radio, Olle has spent the past few years implementing practical newsroom AI workflows while upholding public-service values.

In our conversation, Olle breaks down BBC’s four-part AI strategy, covering large-scale translation and transcription, content reformatting, investigative tools, and early experiments with synthetic audio and conversational news. He shares what’s working inside one of the world’s largest news organizations, what routinely stalls AI projects, and why the most challenging part of AI transformation isn’t the technology but the collaboration required across editorial, product, and engineering. Olle also reflects on what it means to innovate as a public broadcaster in an AI-driven ecosystem, and why archives, credibility, and direct audience relationships will determine which journalism remains indispensable in the years ahead.

In this episode, we cover:

03:39 – The BBC’s four-part AI strategy: Boosting productivity, reformatting content, augmenting journalism, and innovating user experience as the core themes

05:10 – Using AI for large-scale transcription, tagging, live pages, alt text, newsletter production, and translation to save time and make content more searchable.

06:59 – Why AI-assisted translation is one of the BBC’s most advanced use case, and how human sub-editors remain essential to uphold editorial responsibility.

08:17 – Reformatting content across platforms and formats

20:59 – Innovating user experiences with synthetic audio and conversational formats

31:59 – How the BBC uses strategic themes, clear metrics, and fast pilots to decide what’s worth building and scaling

46:59 – Inside the BBC’s fine-tuned LLM and Style Assist

52:01 – What it means to be a public broadcaster in an AI-driven ecosystem

01:02:58 – Olle’s personal AI stack

🎧 Listen to the full conversation with Olle Zacharison on Apple Podcasts, Spotify or your favorite podcast platform.

Can AI write, analyze, and create? Absolutely. But empathy, imagination, and care remain firmly human.

This week on Newsroom Robots, I sit down with Vilas Dhar, President of the Patrick J. McGovern Foundation, one of the world’s foremost philanthropies advancing AI for public good. Vilas leads a $1.5 billion endowment that has committed over $500 million to projects spanning climate action, public health, education, and democratic governance. He has served on the UN Secretary-General’s High-Level Advisory Body on AI, is the U.S. government’s nominated expert to the Global Partnership on AI, and was named a World Economic Forum Young Global Leader in 2022.

Across philanthropy, policy, and technology, Vilas carries one central conviction: technology may accelerate, but the future of journalism and society must remain human-centered.

In our conversation, Vilas discusses his three-part framework for ethical AI deployment (responsible data, clear boundaries, and transparency) and explains how to translate abstract principles into concrete newsroom decisions. He unpacks his LISA framework (Listen, Involve, Share, Assess) for audience-centered AI design, and tackles the hardest questions facing newsroom leaders: Should we buy or build AI tools? How do we balance innovation with environmental sustainability? What happens to human creativity when machines can create?

But perhaps most powerfully, Vilas challenges a deeply held belief in journalism: that media organizations can remain ‘just’ media companies in an AI-driven world. There is no way to be a media organization today without also being a technology organization, he argues, and that shift requires not just new tools, but a fundamental reckoning with organizational identity and purpose.

In this episode, we cover:

00:31 – Introducing Vilas Dhar and his human-centered AI vision
Why technology should serve dignity, equity, and democracy—not just profit

02:17 – The three-part framework for ethical AI
Responsible data, clear boundaries, and transparency as actionable principles

07:08 – Questions leaders must ask before deploying AI
Who’s involved? Who’s accountable? Who has editorial control over AI use?

10:16 – The LISA framework: Listen, Involve, Share, Assess
Turning AI experimentation into behind-the-scenes reporting that builds public trust

13:30 – Navigating ethical dilemmas around AI-generated content
Voice, attribution, and distinguishing between human and machine work

13:51 – The three phases of newsroom AI adoption
From individual experimentation to strategic alignment to building custom tools

18:54 – Why “we’re not a tech company” no longer works
The buy vs. build debate and restructuring organizational identity

23:12 – Organizational reckoning in an 18-month transformation cycle
Redefining purpose when change that took decades now happens in months

25:23 – Reconciling AI’s environmental costs
Why smaller, targeted models and collective action matter more than massive systems

29:14 – Fighting misinformation with AI
Provenance standards, verification tools like Gigafact, and rebuilding trust

34:13 – What journalism is missing compared to other industries
The courage to tackle operational efficiency and low-hanging fruit

37:01 – The evolving role of human creativity and agency
Why AI can find patterns but never understand vulnerability, empathy, or love

39:33 – The McGovern Foundation’s North Star
Moving from AI done “to us” to AI done “for us” to AI done “by us”

44:23 – How Vilas uses AI personally
Family storytelling and flipping AI into a Socratic questioner to sharpen his own thinking

🎧 Listen to the full conversation with Vilas Dhar on Apple Podcasts, Spotify or your favorite podcast platform.

How does a 182-year-old global magazine stay ahead in the age of generative AI?

This week on Newsroom Robots, I sit down with Ludwig Siegele, Senior Editor for AI Initiatives at The Economist. After more than 25 years reporting from San Francisco, Berlin, and London, Ludwig is now leading the publication’s AI strategy. He explains how The Economist created an AI Lab, a startup-style group operating inside the organization but with the freedom to test bold ideas and move quickly. The lab is tasked with looking years ahead, preparing for a future where much of journalism’s supply chain may be automated, and figuring out how to protect The Economist’s identity in an AI-driven media ecosystem.

From practical newsroom wins like AI-powered translation and research pipelines to bolder experiments such as TikTok video dubbing and the SCOTUS bot, Ludwig shares how The Economist is testing, iterating, and learning in real time. He also reflects on what hasn’t worked, the barriers to adoption, and why the next phase of journalism may require rethinking the role of the journalist itself.

Subscribe now

In this episode, we cover:

00:00 – Introducing Ludwig Siegele & The Economist’s AI journey

01:31 – How AI experimentation began at The Economist

04:53 – Building AI infrastructure and upskilling staff

07:10 – The tools and vendor partnerships powering experiments

08:29 – Why adoption is harder than building tools

12:10 – Translation, research, and NotebookLM as newsroom game changers

16:06 – How automation could reshape the journalist’s role

18:41 – Launching The Economist AI Lab

24:11 – Audience-facing AI experiments (TikTok dubbing, Espresso app, SCOTUS bot)

26:05 – Partnering with Google NotebookLM while protecting the brand

30:02 – Scraping, monetization, and the future of publisher revenue

33:41 – Measuring ROI on AI initiatives

37:40 – The biggest barriers to newsroom AI adoption

39:14 – How Ludwig uses AI personally in art and culture

40:40 – Closing reflections

🎧 Listen to the full conversation with Ludwig Siegele on Apple Podcasts, Spotify or your favorite podcast platform.

Subscribe now

In Estonia, a country of 1.3 million people, Delfi Meedia has built one of the strongest foundations for AI in journalism. With one of the highest digital subscription rates in the world, Delfi has moved beyond the buzz around AI to put it into everyday practice, supporting both its journalism and business. Under Chief AI Officer Ivar Krustok, the company has treated AI as infrastructure, spending years on data foundations before rolling out tools that accelerate reporting, connect teams across borders, and give managers real-time answers.

When I spoke with Ivar for this week’s episode of Newsroom Robots, what struck me was the pragmatism of Delfi’s approach. Rather than chasing hype, they zeroed in on the fundamental questions every newsroom faces: How do you help reporters cover more stories? How do you serve readers across multiple languages? How do you ensure staff trust the tools being rolled out?

The results are impressive. Delfi now runs a live translation feed that lets reporters in Estonia instantly read stories from colleagues in Latvia and Lithuania. A searchable archive spanning 25 years of reporting helps editors pull quotes or trace narratives in seconds. Journalists no longer sift through endless court filings as bots scan new cases and deliver quick summaries to editors, already helping find stories that might otherwise have been missed. Even the advertising team benefits, with bots that answer inventory questions in industry-specific language.

Delfi’s AI strategy is organized into three components. The first is everyday newsroom tools such as transcription, translation, and summarization. The second is long-term experimental projects, such as their in-house ChatGPT-style toolkit trained on archives and government data. The third is company-wide literacy, ensuring that everyone, from reporters to managers, can confidently use AI. This blend of immediate utility, forward-looking experiments, and staff training has laid the foundations for Delfi to be an AI-ready newsroom.

Here are three takeaways from my conversation with Ivar Krustok:

1️⃣ Build the foundations for your data infrastructure before AI

Delfi spent years cleaning subscriber records, standardizing article tags, and unifying advertising and editorial data. That foundation is why their AI tools actually work. Skip this step, and you’re building on quicksand.

2️⃣ Build what is core, buy what is not

Delfi develops its own systems for sensitive or proprietary work and relies on vendors for routine tasks like summarization or polls. This approach keeps costs down while safeguarding what makes them unique. If something drives your competitive edge, keep it in-house.

3️⃣ Move fast with prototypes

With tools like Manus AI, Google Stitch, and v0.dev, Ivar can spin up working prototypes in under three hours. Instead of debating specs in meetings, editorial and product teams test real tools immediately, provide feedback, and shape what gets built. This speed has created both momentum and trust across the newsroom.

Delfi’s experience shows that the hard work of AI lies not in the models but in the management. Clean data, clear ownership of core systems, and rapid prototyping are what turn AI into real results. Before adding another tool to your stack, ask whether your data is ready, your newsroom knows how to use it, and your team can test it quickly. That is the difference between experimenting with AI and building with it.

🎧 Listen to the full conversation with Ivar Krustok on Apple Podcasts, Spotify or other major podcast platforms.

As a new academic year begins, journalism schools face a defining challenge: how to prepare students for a profession being reshaped by AI.

At Stanford, Djordje Padejski continues a journey he began years earlier. Now associate director of the JSK Journalism Fellowships, he has been wrestling with this question longer than most. A veteran investigative journalist and educator, he launched one of the earliest AI-focused journalism courses at Arizona State University before bringing it to Stanford last year. His classroom has become a lab where students not only use AI tools but also examine them.

In this week’s episode, Djordje shares how he structures his course and what journalism schools must do to prepare the next generation of journalists.

Here are three takeaways from our conversation:

1️⃣ Don’t Ban AI, Make It a Learning Tool

When AI first appeared in classrooms, the knee-jerk reaction was to ban it. Djordje took the opposite approach: he required students to disclose exactly how they used AI and reflect on what it got right or wrong. His students analyzed the technology’s limits instead of avoiding it.

Thanks to Djordje’s support, we’re sharing his syllabus with you here as a resource.

2️⃣ Focus Less on the Product, More on the Process

Instead of grading polished essays that AI could easily generate, Djordje emphasized in-class discussions, live experiments, and peer critiques. His students also pursued semester-long projects exploring AI’s role in areas such as surveillance, healthcare, and self-driving cars.

3️⃣ Every Beat Is Now an AI Beat

Whether covering healthcare, politics, or business, reporters are already seeing AI reshape their subjects. For Djordje, that means AI education isn’t an elective but a core requirement. You don’t need to be a coder, but you do need to understand how AI works.

As students return to classrooms this fall, it’s a moment of opportunity. AI can churn out headlines and summaries, but it can’t exercise judgment, understand an audience, or hold power to account. Djordje’s approach is a reminder that if we teach journalism students to examine AI, not fear it or blindly embrace it, we’ll shape a generation fluent in tomorrow’s tools and rooted in journalism’s values.

🎧 Listen to the full conversation with Djordje Padejski on Apple Podcasts, Spotify or other major podcast platforms.

How your journalism is packaged, credited, and consumed is shifting — and Particle News offers a glimpse of where aggregation may be headed.

Launched in November 2024, Particle News gives readers a fast way to see how a story is covered from multiple perspectives. It pulls in reporting from across the political spectrum, generates concise summaries, verifies each claim against original sources, and adds a bias meter to show how the story is framed. For users, it’s an incredibly polished news experience. For publishers, it’s a preview of how aggregation technology could reshape the first point of contact with audiences and what that might mean for engagement with their journalism.

This week’s guest is Sara Beykpour, co-founder and CEO of Particle. In this conversation, she takes us inside Particle’s approach to building trust and transparency, and explains why personalization doesn’t have to create filter bubbles.

Here are some of my takeaways from our conversation:

1️⃣ A Methodology Built to Minimize Hallucinations

According to Sara, Particle’s summarization process is designed to keep accuracy front and center. A story isn’t touched until at least three articles from two different publishers are available, ensuring a diversity of viewpoints before any AI is involved. Once a draft summary is produced, it runs through a second “Reality Check” — an additional AI pass that double-verifies each claim against the original reporting, generating citations and a transparent audit log. By combining multi-source input and a verification layer, Particle has built a summarization pipeline that sharply reduces hallucinations while maintaining nuance.

2️⃣ Making Bias Visible Without Creating Filter Bubbles

Rather than trying to eliminate bias, Particle makes it visible. Its bias meter shows how coverage leans left, right, or center, using nonpartisan human-sourced data that’s updated automatically as stories develop. Personalization is based on topics, not sources, so a reader interested in politics will see perspectives from across the spectrum. For users, this approach offers a more complete understanding of the news environment. For publishers, it means your framing of a story becomes one voice among many in a blended, multi-perspective view.

3️⃣ The “Facts-as-API” Future

We discuss a strategic shift that is happening where the “facts layer” of news may eventually function like an API that is licensed, freely exchanged, and remixed while analysis and opinion become the premium offerings.

For an aggregator, that’s a logical and potentially lucrative evolution. Particle already has licensing deals with multiple publishers, including one with Time via Tollbit, ensuring compensation for original reporting.

But here’s the critical point for publishers: when the aggregator becomes the primary interface, your news is no longer the complete narrative package under your masthead. It’s distilled into verified facts, quotes, and bullet points — reassembled alongside competitors’ coverage — in an experience the reader may trust more than your own brand.

In a world where the first touchpoint with your work might be through an aggregator’s summary rather than your own homepage, what happens to your editorial identity? Today’s aggregators often act as traffic drivers, but in the next generation of news aggregators, the summary itself becomes the primary reading experience.

For some publishers, this could be a brand amplifier putting their work in front of audiences they might never otherwise reach. For others, it could feel like a brand filter, where their distinct voice is distilled into a few sentences alongside competitors. The challenge will be to decide how much of their journalism they want to live inside someone else’s container.

🎧 Listen to the full conversation with Sara Beykpour about how she’s building the next AI powered news aggregator on Apple Podcasts, Spotify or other major podcast platforms.

What if the future of journalism isn’t locked behind the paywalls of big tech companies, but freely available to every newsroom willing to embrace it?

Too often, the conversation around AI in newsrooms centers on big tech, like OpenAI’s ChatGPT or Google’s Gemini. These are powerful tools, no doubt but they come with caveats: mainly cost, limited transparency, and little to no control over where your data ends up.

But there’s another world of AI rapidly evolving in parallel and it might be journalism’s best path forward: open-source AI.

In this episode of Newsroom Robots, I welcomed back Florent Daudens for his second appearance on the podcast. He first joined us last year on the podcast while leading AI innovation at Radio-Canada. Since then, he’s taken on a new role as Press Lead at Hugging Face, the open-source AI platform. Described as “the GitHub of AI,” Hugging Face hosts over 1.5 million models used by 7 million people daily.

Here are three key insights I took away from our conversation:

1️⃣ Open-Source AI Lets Newsrooms Reclaim Control from Big Tech

Here’s the fundamental difference: When you use a tool like ChatGPT, your data is sent to OpenAI’s servers, processed by their systems, and governed by their privacy policies with all the potential risks of third-party data exposure. With open-source AI, you download the model (think of it like downloading any other file) and run it entirely on your own computer or servers. That means greater control, better privacy, and fewer surprises.

"Each time you're running an open source model on your device, it is free to download, it's free to run, you don't pay any cost," Florent emphasized. For newsrooms handling sensitive investigations, confidential sources, or unpublished content, this local processing eliminates major security risks entirely.

But the advantages go far beyond privacy. Open-source models offer full transparency: you can see how they were trained, what data was used, and what biases might be built in. Proprietary models like ChatGPT are black boxes — you have no idea what’s happening under the hood. With open-source models, you can inspect every component and modify them to fit your needs.

Perhaps most importantly, you’re never locked into a vendor’s decisions. When OpenAI changes ChatGPT’s capabilities or pricing, you have no choice but to adapt. With open-source alternatives, you control the entire pipeline. “You can switch models in one line of code,” Florent said — meaning newsrooms can experiment freely, compare results transparently, and never worry about outside companies changing the rules.

2️⃣ Smaller Models, Greener Footprints

I’ve noticed that one of the most significant barriers to AI adoption in newsrooms isn’t just technical, it’s ethical. Many journalists are wary of AI because of its massive environmental footprint, and they’re right to be concerned. Florent pointed to research from Hugging Face showing that every time you generate an image with ChatGPT, it consumes about as much energy as fully charging an iPhone. Now scale that across millions of users making thousands of requests every day and the carbon impact becomes staggering.

Florent offered some practical advice on how newsrooms can reduce their AI footprint. First, he emphasized the importance of asking a simple but critical question: “Is AI actually needed here?” He explained, “The news industry has focused on productivity as a first step with AI and that’s okay. But sometimes you have to ask: do we really need AI for this specific task?” In other words, just because you can use AI doesn’t always mean you should.

Florent’s second recommendation focused on where newsrooms can gain the most efficiency: swapping out large foundational models for smaller, task-specific ones. “What I’m seeing,” he said, “is that more often than not, the news industry is using big foundational models for tasks that could easily be handled by smaller, specialized models.” This is where the small-model revolution becomes crucial — delivering faster results, lower costs, and a far lighter environmental footprint.

A big foundational model is something like GPT-4o or o3 — the kinds you access through ChatGPT. But thanks to the small-model revolution, compressed models are now achieving remarkable capabilities while running entirely on local devices. Florent spoke about a 250-million-parameter vision model — smaller than most smartphone apps — that can shot-list an entire video in minutes. And when you run a model like that locally, you're using a computer that's already powered on for other work — no need for extra servers or transmitting data across the internet.

Compression advances are accelerating fast. Models that once needed massive server farms are now being distilled into lightweight versions that run smoothly on everyday laptops and all while retaining most of their core capabilities.

3️⃣ Preparing for the Post-Website Era

The most existential threat to journalism’s business model may be the rise of AI agents that eliminate the need for audiences to visit news websites altogether. These autonomous systems can browse the internet, read articles, synthesize information, and deliver personalized news briefings, all without users ever seeing a headline, ad, or paywall.

“You can ask your agents to retrieve information for you. They’ll browse the website, extract the content and you’ll never see the UX, the UI, or the ads,” Florent explained, describing a reality that’s already taking shape with tools like ChatGPT’s search feature and Perplexity’s AI-powered search.

This shift threatens to obliterate traditional revenue models built on page views, time-on-site, and ad impressions. When readers can ask their AI assistant for the latest news and get a full briefing without ever visiting a news site, what happens to digital advertising? And if agents can instantly synthesize information from multiple sources, why would anyone click through to a single article?

But this disruption also opens up new opportunities for newsrooms willing to rethink their relationship with audiences. Instead of competing with AI agents, the most forward-thinking outlets may need to start building for them: creating content and experiences that work whether consumed by humans or by machines acting on their behalf. In this new landscape, the winners won’t just own the content, they’ll own the experience.

🎧 Listen to the full conversation with Florent Daudens on Apple Podcasts, Spotify or other major podcast platforms.

We in journalism often find solace in the phrase “human-in-the-loop,” believing it is our ultimate safeguard against the unpredictable tides of artificial intelligence. But what if that comfort has become our greatest strategic vulnerability? What if our unwavering insistence on human oversight is actually preventing us from unlocking AI’s truly transformative potential?

I had the privilege of moderating a provocative live panel at the Nordic AI in Media Summit. My guests were three newsroom leaders who are at the forefront of integrating AI in journalism: Gard Steiro, Editor-in-Chief and CEO of Verdens Gang(VG) in Norway; Fabian Heckenberger, Managing Editor and Senior Editor for AI at Süddeutsche Zeitung (SZ) in Germany; and Naja Nielsen, Media Director at Sveriges Television (SVT) in Sweden and former Digital Director at BBC News. We discussed the big questions facing media organizations today: How do we align AI with our core editorial values? What does it take to earn trust in AI-powered news products? And how should we structure our teams and strategies to stay agile amid relentless, dizzying change?

Here are three takeaways from our conversation:

1️⃣ Embrace Uncertainty and Foster Internal Debate for Agility

In a landscape where regulations are fluid, and the future remains hazy, focusing on what you can control internally is paramount. Gard Steiro of VG shared his perspective: “My most important task as a leader is not to point the direction, but to foster a discussion within the organization … People disagree and we have that discussion at VG and we work with scenarios.” The goal isn’t to possess all the answers but to cultivate an environment where uncertainty is openly acknowledged—enabling constant experimentation and the emergence of diverse solutions. Naja Nielsen reinforced the point, arguing that top managers must be “brave enough to make the damned-if-you-do, damned-if-you-don’t decisions” and adopt Silicon Valley’s iterative, test-and-learn mindset. Our biggest legacy constraint, she says is the lingering belief that “the editor-in-chief must be the most clever man in the room and therefore he or she must make all decisions of importance.”

2️⃣ Redefine “Human-in-the-Loop” and Prioritize Real User Needs

The traditional notion of a "human in the loop" for every AI output might actually be holding us back. While transparency and admitting errors are crucial for trust, as Gard pointed out, "If you still insist on that [human in the loop], we won't take the advantages of AI." Naja questioned our inherent trust in humans over machines: “If I had to choose between a doctor doing the surgery and a machine right now doing it on my beating heart, I would prefer the machine.” She challenges us to think about what humans truly excel at – "storytelling, the on the ground reporting, everything that only humans can do" – and let machines handle tasks they can do better, even if it means rethinking our established workflows. Fabian Heckenberger of Süddeutsche Zeitung (SZ) echoes this, urging newsrooms to move beyond “playing with AI for the sake of AI” and concentrate on how it boosts subscriber retention and efficiency.

3️⃣ The Article May Die, But Storytelling and Unique Value Will Thrive

In the discussion about whether the article—treated as a commodity—has run its course, Gard responded, “The article will die, should die, but storytelling will not” Naja added a different perspective, suggesting that articles will survive for a niche audience, yet the panel’s shared message was unmistakable: format is secondary to relevance. The real challenge is to pinpoint journalism that as Gard said “can’t be copied, that gives value to the readers,” and rapidly redirect resources there. Fabian also highlighted that in a flood of easily produced, unverified content, there is a rare opportunity for “truly high-quality journalism to double down on what we’ve been doing for the last five to ten years.” The future of news is not about preserving outdated formats but about inventing new, compelling ways to deliver essential and trustworthy information.

This conversation was recorded live at the Nordic AI in Media Summit in Copenhagen. 🎧 Listen to the full episode on Apple Podcasts, Spotify or other major podcast platforms.

What if AI represents journalism’s opportunity to serve everyone rather than only those audiences large enough to attract advertisers?

That question has stuck with me since my recent conversation with Gina Chua, Executive Editor at Semafor, recorded live at New York University in collaboration with the AI networking group Humans in the Loop.

With decades in leadership roles across newsrooms as the former Executive Editor at Reuters and Editor-In-Chief at South China Morning Post, Gina is one of journalism’s most leading voices on AI.

During our conversation, she revealed something striking about her approach to AI experimentation. "I have a very deep and personal relationship with Claude now, where I'm just sort of sitting and we just start talking for hours," she said, describing her "vibe coding" sessions—rapid prototyping that turns ideas into functional tools. These late-night collaborations with AI have led to her experimenting with multilingual search engines, style guide bots, and content analysis systems.

This experimentation also led to her prototyping an AI editorial safety tool—a chatbot to democratize security expertise for journalists. "There's lots of news organizations that can't afford security advisors, don't have nine people full time doing this like we had at Reuters. So why not try to bring that information to the world?"

That prototype has since evolved into JESS (Journalist Expert Safety Support), a product that Newsroom Robots Lab has been developing in collaboration with the ACOS Alliance and CUNY's Journalism Protection Initiative to make expert security knowledge available to newsrooms worldwide.

Her vibe coding sessions showcase how AI has lowered the barriers to innovation. What once required teams of developers can now be prototyped by a single person with domain expertise and a simple conversation with an AI chatbot like Claude.

Here are three key insights I gained from our conversation:

1️⃣ Personalization Can Scale Inclusion If We Do It Right

For the first time, AI enables journalism to do what was once difficult: offer multiple versions of the same story, tailored to different audiences by geography, language, knowledge level or cultural context.

"You can't possibly write 25 versions of a story for 25 different people but a machine can," Gina said.

That unlocks audiences we’ve historically failed to reach. Newsrooms can now serve groups once considered too small, too remote, or too linguistically complex to justify commercial investment.

However, this power comes with a critical tension: the more we personalize content, the more we risk fragmenting our shared civic experience. Can journalism serve individuals and hold space for the collective? That's the paradox AI throws at us, and the answer will shape the next generation of news products.

2️⃣ Journalism's Value Must Move Upstream in the AI Era

Gina repeated a provocative take of hers she often shares: "We're not in the business of saving journalists. We're not necessarily even in the business of saving journalism. We're in the business of saving information in the public interest."

As AI grows adept at summarizing, rewriting, and even generating articles, much of what we traditionally consider “journalism” like storytelling, packaging, and writing can be automated. Yet the upstream work of reporting, cultivating sources, verifying facts, and making ethical judgments remains deeply human.

What becomes scarce and thus valuable is not the content but editorial judgment: asking the right questions, uncovering hard-to-access information, interpreting complexity, building trust with communities, and holding power to account.

According to Gina, journalism must move upstream toward investigation, verification, and meaning-making to remain essential. The rest may end up being handled by the machines.

3️⃣ AI Adoption Isn't a Tech Challenge, It's a Workflow One

Gina's experience embedding tech into newsroom routines made one thing clear: the hardest part isn't the tech. It's workflow.

AI is often viewed as an app that can simply be dropped into existing systems. Yet true integration demands rethinking how decisions are made—who touches what, and when. That means rewiring habits, redefining job expectations, and sometimes even shifting power structures.

AI adoption can’t rely on isolated pilot projects because without redesigned workflows, the technology will quietly fade away.

🎧 Listen to the full conversation with Gina Chua on Apple Podcasts, Spotify or other major podcast platforms.

When The New York Times appointed Zach Seward as its first Editorial Director of AI Initiatives in late 2023, he stepped into the role just as the paper was making headlines—for suing OpenAI and Microsoft.

“I started the job right before Christmas,” he told me during our live Newsroom Robots podcast taping in Detroit. “It was the week between Christmas and New Year’s that The Times sued OpenAI and Microsoft and I learned about that the way everybody else did, which was the push notification on my phone from The Times app.”

That beginning set the tone for what’s proven to be a high-stakes, high-impact role. Rather than rushing into flashy deployments or quick integrations, Zach and his team have opted for a more intentional approach: slow, strategic transformation built on trust, education, and careful experimentation.

I sat down with Zach for a live taping of the Newsroom Robots podcast during the AI Leadership Summit in Detroit. The summit—co-hosted by Newsroom Robots and the Online News Association—brought together media leaders, technologists, and newsroom practitioners to grapple with urgent questions about the future of journalism.

Subscribe now

Zach, who now leads AI initiatives at The Times, previously co-founded the business news site Quartz in 2012, where he served as CEO, Chief Product Officer, and Editor-in-Chief. His career has spanned editorial, product, and leadership roles—all now converging in his efforts to responsibly guide one of the world’s most influential newsrooms through the AI age.

Here are three key takeaways I had from our conversation:

1️⃣ Build AI Literacy Before AI Tools

Zach’s team has prioritized talking to people across the newsroom—what they call their “AI roadshow.” They’ve visited nearly half of the newsroom’s 2,000 journalists across 40 desks to walk through what AI is (and isn’t), where it can be useful, and how it might fit into their work.

“I’ve been surprised how much of the job has turned out to be talking,” he said.

This approach isn’t about evangelizing a new technology; it’s about expanding imagination and addressing real concerns. One recurring theme that kept coming up in these conversations? Summarization. Whether it’s writing homepage blurbs, SEO descriptions, or internal notes, summarizing content is a constant and time-consuming need. That insight laid the groundwork for what came next.

2️⃣ When AI Starts With the Problem, Not the Tech

One of the most telling examples of The Times’ approach to AI is a tool called Echo—an internal summarization assistant built by Zach’s team. Echo allows journalists to input links to Times articles and receive summaries tailored to specific needs—whether for homepage blurbs, internal notes, metadata tags, or newsletters. What makes it compelling isn’t the technology powering it (it runs on large language models), but how precisely it addresses one of the most common, everyday challenges journalists face: the need to repeatedly distill and repackage information.

Zach shared that summarization came up again and again in conversations across desks—an often-overlooked friction point that was slowing down workflows.

Rather than starting with the capabilities of AI and looking for places to apply them, the team is listening first—identifying pain points and building tools that solve for them in a way that feels native to how journalists already work.

3️⃣ Putting Journalism at the Center of AI Innovation

The team’s work is guided by three pillars, all identified from early on.

“First is using AI for research and investigations,” Zach said. “Number two is internal workflows and processes those like knotty or just robotic parts of everybody’s job… And then the last area of focus is reader experiences.”

A powerful example of the first bucket came from an investigative project involving the Election Integrity Network. The Times received a leak of 500 hours of internal Zoom meetings—roughly 5 million words of audio. Using LLMs and targeted prompts, the team was able to process, analyze, and extract key leads from that data—on deadline. That same pattern, Zach said, has been useful in many areas of coverage.

They’re also exploring how AI can transform reader interaction. The current search function, he admits, isn’t great. But experiments with embedding models are already showing promise. And when it comes to voice? He’s leaning into what’s next: “We should be prototyping what talking to The Times would be like.”

All of these insights reinforce something I’ve believed since launching Newsroom Robots: AI transformation in newsrooms isn’t just about technology. It’s about people, workflows, and culture. It’s about giving journalists the language and tools to experiment, ask better questions, and reimagine what’s possible without losing sight of editorial values.

🎧 Listen to the full conversation on Apple Podcasts, Spotify or other major podcast platforms.

📍And if you’re in New York, join us on June 3 at New York University for our next Newsroom Robots live podcast taping with our guest Gina Chua, Executive Editor at Semafor.

Only a few spots remain—sign up here.

Subscribe now