Re:filtered #25: The missing layer of media strategy

Welcome to the 25th edition of my monthly newsletter on journalism in a moment of systemic disruption.

Two years since I left my latest (or last?) corporate job to start Gazzetta, I've had the luck to work with amazing technologists and media thinkers on all continents.

In the work I'm lucky to do, I keep seeing the same tangle:

Challenges are almost always rooted in a conflation of three different layers of media strategy: content, product, and service. Keeping them tangled has real costs and leads to dead-ends.

You can't solve a product problem with more content thinking, or a service problem with more product thinking, or a content problem with more service thinking.

My own path followed this sequence. I started as a reporter and editor, obsessed with the story, the scoop, the craft of telling. Then I moved into product, building teams focused on how journalism reached people and how they experienced it.

Recently I've been spending a lot of time reading and learning about service design, the strategic layer underneath: what promise are you making, to whom, and how do they experience it.

The tangle is a legacy of how journalism evolved when it had more capital and status. As both wane, we're only slowly recognizing that service is the layer most media ventures neglect, and the one AI really can't easily replace.

This is just how I think about it. Others may organize these ideas differently. I find it helps to have some way to distinguish what we publish and how people access it from the ongoing commitment we're making.

Content, product, service.

Content is what you publish.

This is where journalism schools focus, where awards get given, where most of our industry's attention goes. It's also where large language models are most capable. Models can summarize, explain, and generate at scale. The perfectly tailored individual content piece is what they're built for.

This doesn't diminish what skilled journalists bring. The craft of finding new information remains powerful, including when augmented by AI, and remains urgently needed even amid growing intermediation. The craft of telling stories well is genuinely beautiful. But at the commodity layer of daily digital content, it's just not where scarcity lies anymore.

Product is the interface people use to access content.

This is all about your site, your app, the newsletter template, the onboarding flow, how membership works. Product optimizes how people consume and navigate.

News product thinking has been central to how the industry responded to the loss of its privileged position over the past decade. There's real craft in building experiences that respect people's time and attention. I'm learning a lot from the people doing this work well, like what my onetime colleague Michaël is building at Trustfund for collaborative growth.

The next frontier is how product adapts as AI interfaces increasingly mediate how people access information, and how to preserve informed decision-making on a product level with proprietary information (see NPA's Data Commons work). I'm looking forward to the just-announced 2026 NPA Summit in Oct. that centers on exactly this question.

Service is the repeated promise you make and keep.

This is the layer most media ventures neglect, and the one AI can't easily replace. In my work, I've found few journalism ventures that can articulate a reliable and relevant commitment that people can return to, with clear expectations and support when something breaks. Some benefit from serendipity, others from legacy status, but neither is a strategy.

Consider what it means to help parents navigate a school district. An LLM can summarize board minutes (useful as a tool within a larger operation) but it cannot maintain a years-long commitment to helping parents in a specific district understand an opaque system, update them when policies change, and be there when the automated answer fails.

The distinction matters because trust works differently at each layer, and journalism, the profession, has conflated them into a generic "news" concept too broad to be addressable.

At the content layer, credibility is demonstrated per piece. Good sourcing, accurate facts, corrections when wrong. This is what journalism ethics courses teach and what awards recognize. It's necessary but insufficient.

Aristotle identified three modes of persuasion: logos (logic), pathos (emotion), and ethos (character). Only ethos requires time: a single story can be logical and can move people emotionally, but ethos is demonstrated through a pattern of repeated behavior.

The sociologist Georg Simmel observed that trust requires a leap beyond what evidence alone can justify. We can verify that yesterday's article was accurate. We cannot verify that tomorrow's will be. Trust is the wager that fills that gap, a suspension of uncertainty that makes ongoing relationships possible.

Neither thinker was describing what happens when someone reads a single story, and couldn't have imagined things like a podcast, a chatbot, or an AI agent. They were describing relationships that require consistency, accountability, and some mechanism for recourse when things go wrong. That's service.

We've neglected this because in the big platform era we got away with narratives of bringing about societal change in tacit complicity with big tech platforms, directed at passive mass audiences whose minds we thought could be shaped by optimized content creation and distribution.

This is an unreflective and opportunistic continuation of the broadcasting mindset, and where "theory of change" frameworks come from. They persist in media strategy long after the conditions that made them plausible.

In this current, more disruptive moment, in which that symbiosis largely no longer exists or has been hijacked by bad actors, we can either become subservient subjects to the platform overlords or emancipate ourselves from them in service to fellow people.

To be clear: this isn't about rejecting platforms. They remain useful distribution channels. But without a service-level strategy, their distribution parameters become your only relevance signals, and start defining what you do.

A growing number of people in media see this. No wonder service design tools (even if they're not called that) are becoming central to many independent media strategies.

Examples of service promises:

• Functional: "We help you figure out which options actually work for people in your situation."
• Emotional: "We help you feel less lost when no one's explaining what's going on."
• Social: "We connect you with others dealing with the same thing."

Service includes mapping how people discover what you offer (including in the physical space), how they use it, and what happens when something fails. A service is a relevant promise, with operations attached.

Current media strategy often toggles between "publish better" (content) and "ship features" (product) but under-invests in "keep an appreciated promise, reliably" (service). That missing layer is where trust and habit accumulate and translate into social capital.

Why service matters now

Almost everyone agrees that AI is reshaping where journalism has a competitive advantage, but the solutions diverge depending on which layer you're working at (content/product/service).

We documented last year how models fail people in specific, predictable ways. They over-index on narrow sources. They echo political framing back at users. They give advice detached from local reality. These gaps are exactly where journalism has always found its purpose, where institutions fail to serve people, where information that should be accessible isn't. The niches are just getting less obvious.

As our research shows, LLM output is often not systematically aligned with users' interests. Service-oriented journalism can remain resistant to AI displacement precisely because it requires ongoing relationships, local specificity, and accountability to actual outcomes. It also requires something AI structurally cannot provide: the willingness to take sides with people against systems that fail them.

And yet the journalism industry is retreating

The Reuters Institute just released its always insightful Journalism, Media, and Technology Trends and Predictions annual report, based on a survey of 280 senior leaders across editorial, commercial, and product roles from 51 countries. The report is a valuable snapshot of the thinking of many people setting strategy at traditional and digital-born publishers globally.

Their priorities for the year ahead: more original investigations and on-the-ground reporting (+91 percentage point difference between "more" and "less"), more contextual analysis and explanation (+82), more human stories (+72).

What they're scaling back: service journalism (-42), evergreen content (-32), and general news (-38). The reasoning? They expect these categories to become "commoditized by AI chatbots."

The report doesn't define what it means by service journalism (Nic: if you're reading this, correct me if I'm wrong!) and I may be arguing against a narrower definition. The framing suggests a content category: sports scores, event listings, how-to guides, recipes. Generic utility content. And yes, AI can do that.

But the industry has set up a false choice: generic content that AI can replicate versus distinctive content that makes us special.

The actual distinction that matters runs on a different axis: whether the journalism reliably helps specific people with specific problems over time. That's not a content category. It's an operating model that transcends content formats: An investigation can be service. Analysis can be service.

Much of the industry is still working from inherited frameworks, treating content (or now content + product) as the whole game. While leaders debate vertical video versus text and fret about chatbot competition, we're not having a harder but more necessary conversation: what reliable service do we provide, and to whom?

Value and service thinking can break through this. It can generate revenue. Services that reliably solve problems create retention, referral, and willingness to pay that content alone rarely achieves. I don't see a sustainable revenue path without it.

And it's the actual path through AI disruption: not retreating to status work, but doubling down on the promise-keeping that other people's machines can't replicate.

Reframe service as commitment rather than content category, and the strategic logic flips.

What service thinking unlocks

Services create continuity. They turn one-off consumption into an ongoing relationship with expectations. "I know what I'll get here and what happens if I get stuck." Content gets consumed and forgotten. Products get used and abandoned. Services get relied upon.

Services force clarity. "We cover education" is a beat description, not a service. For whom? To help them do what? Until you can answer those questions, you can't know if the coverage actually serves anyone (other than perhaps yourself).

Services change the internal conversation. Instead of "what should we publish next?" and the somewhat arbitrary self-expression of individual storytellers (instincts that matter, but need grounding), you can ask "what promise are we trying to keep to whom, and where is it failing?" That question leads somewhere useful, and rewarding.

Services make impact legible. You can evaluate a service as a system: outcomes, completion, repeat usage, resolution. Not just content performance metrics that tell you nothing about whether anyone was helped.

Finally, we can emancipate ourselves from the "theories of change" that often treated people who consumed our work as passive recipients of benevolent interventions, to be molded by our content.

Consider this difference:

Content framing: "We produce award-winning investigations into school board decisions and regular reporting on education policy."

Product framing: "We built a searchable dashboard where parents can filter board decisions by school and subscribe to alerts."

Service framing: "We help parents in Brooklyn track how budget decisions affect their specific schools, with alerts when relevant policies change and explanations of what those changes mean."

The content version lets you coast on abstractions forever. The product version describes features without saying what promise they fulfill. The service version has accountability built in. You can measure whether you're delivering it. You can improve it based on feedback. You can explain to funders and supporters exactly what their money enables.

A common response: "But we can't really promise outcomes."

True, but we can articulate the outcomes we seek, which is different from dressing up self-expression as public service. Impact becomes a journey from clear intent to consistent service, with follow-through when possible.

In practice

Last year, Madison Karas and I ran what we called the Audience Help Desk with support from the Lenfest Institute for Journalism. We offered dozens of free brainstorming sessions for anyone wrestling with questions of relevance and utility and how to be more than just a content factory of things you care about.

In conversations with newsrooms from around the world, the challenges we faced kept pushing us toward service design tools: jobs-to-be-done, journey mapping, service blueprinting. To help meaningfully, we had to go there. They unlock conversations that traditional audience research can't.

One example: A team came to us frustrated that their investigative work wasn't getting traction. Great stories, minimal response. The instinct was to fix distribution: creator partnerships, SEO, email list clean-ups, different platform-specific storytelling, etc. We didn't start there. We mapped what happened after someone read a piece, subscribed to a newsletter, or donated.

A reader finishes an investigation into landlord violations. Then what? There was no way to check if their building was on the list. No explanation of what the violations actually meant for tenants. No guidance on what options existed, legal, practical, or otherwise. The journalism exposed a problem. It didn't help anyone figure out what to do next.

The storytelling was strong, but the service didn't exist. Readers were informed and just felt more stuck in a depressing hellscape than they did before.

Once they saw that gap, the question shifted from "how do we get more readers" to "what would it mean to actually help someone navigate this?" Not solve their landlord problem. Journalism can't do that. But move from "here's something wrong" to "here's how to understand your situation and what your options are for better decisions you can choose to make."

The work evolved from "answering audience questions" to "designing a repeatable support capability." We started calling it the Service Desk.

This isn't solutions journalism, which is a valuable content approach focused on stories about what's working. Solutions journalism answers "what kind of stories should we tell?" Service thinking operates upstream: "what promise are we keeping to whom?" before any content decisions get made.

A service-oriented newsroom might use solutions journalism as one of its content approaches. They're complementary, not competing. The investigation into landlord violations might be exactly the right content. The question is what surrounds it.

The invitation

After months of reading up on research I wish I had read in J-school (Huge thanks to Lou), I'm getting certified in service design in the coming months to learn more about what works in other industries. I'll share what translates to newsrooms (and what doesn't) in upcoming newsletters and workshops.

In the meantime, the Service Desk is open for mostly free brainstorming sessions and resources. No consulting contracts, no setup required, just bring an open mind. Claim any free slots in our calendar, or if they're taken, email servicedesk@gazzetta.xyz.

Slots are free for U.S.-based media ventures. For international ventures, we fund free sessions through paid slots for journalism-support organizations and technologists, inspired by the caffè sospeso model (pay-it-forward).

I know this framing is easier to apply when you're not facing layoffs or working within institutional constraints or hyperbolic funder expectations. The Service Desk exists partly for that reason, to help people find what's possible within their actual situation.

There's real joy in this work when it clicks and when the abstraction becomes gratitude from someone whose day you made easier or whose perspective you shifted.

Looking back

Our research at Gazzetta this month shows this gap in practice. We published findings from our ongoing work in Iran.

In the days after the internet blackout earlier this month, we asked thousands of Iranians what was on their minds, using experimental outreach methods we developed with support from the Open Technology Fund.

We expected them to talk about politics. Instead, hundreds of respondents wanted to talk about money, how to pay for food, how the shutdown itself had cost them income, how to earn something online. Only 7% mentioned political themes; 32% mentioned livelihood.

This is the service gap in action. While the content layer focused on political analysis, the service need was about livelihood. That gap is the opportunity. Media ventures and media funders often assume "the news" is self-evident and just needs better distribution.

The responses show that even (or especially) during a major crisis, people's actual information needs may still be about getting through the week. That gap between what outside observers think or project matters and what people on the ground are asking about is exactly where service-oriented journalism can start.

We also just published the preliminary results of a pilot audit with our partners at ASL19 examining how five large language models source Iran-related information in Persian.

Why bother? Because so much fact-checking has gone stale, a kind of social platform-funded rent-seeking that stops being useful when platforms stop caring. But as AI intermediates more information, understanding how models source and weight that information becomes critical. This kind of audit can differentiate one LLM from another. We hope to show how similar work could matter again.

We found that just asking models and logging answers doesn't lead to useful measurement. It's too simple. We wanted to know not just what they say or what they cite, but how quickly political framing in the prompt changes both.

Three findings stood out:

• The most meaningful split wasn't "good vs bad" but whether models held their position or echoed the framing of politically charged prompts. Small shifts in prompt vocabulary flipped outputs from human-rights documentation to state-aligned framing. (This is the 2026 equivalent of Facebook suggesting additional Neo-Nazi groups to people who have joined one already.)
• We found that Persian outputs are still shaped heavily by English-language and Western-indexed retrieval infrastructure. That advantage is contingent and temporary. It could shift as more inference providers build out their own pipelines, with Chinese open-source language models especially becoming something to watch and test carefully.
• "Credibility" collapses into legacy-media signals. The models favor institutional markers over actual evidence. That's a problem because it can be so easily gamed.

This builds on our earlier DeepSeek testing in Chinese, where we found both Western and Chinese models disappointing workers seeking practical labor guidance. DeepSeek contains useful, contextualized knowledge about Chinese workers' daily constraints but wraps it behind moderation barriers aligned with the Communist Party's hyper-capitalistic authoritarianism. Western models face no such restrictions, yet offer impractical advice detached from local reality. Neither serves the people using them well.

The pattern holds across contexts: AI closes some information gaps while creating others. We're expanding the testing to other models and regions, and building a workshop to help teams think through how models shape their information space.

If you want the full report, subscribe to Gazzetta's Field Notes, and we'll send it to you in a few weeks.

If you're curious about who we work with in Iran, ASL19 and other partners got coverage in the New York Times this month for their great work helping Iranians bypass internet censorship. Worth a read.

Looking ahead

Reminder: new Service Desk slots are open. Claim one in our calendar, or email servicedesk@gazzetta.xyz if they're taken. They went quickly when we emailed the waitlist a few days ago.

We're preparing an in-person service design workshop for media in Philadelphia in March. More details soon. If you're interested, let me know and I'll keep you posted.

And a loose idea: a friend from the Russian exile media space is visiting New York in February. I thought it might be nice to get together with like-minded people working on similar challenges. No agenda or plans yet. If that sounds appealing, let me know.

Thank you for reading. If you work in strategy, or with someone who does, I hope you see the opportunity here. I'd be grateful if you passed this along to anyone making these decisions.

And if any of this sparked a thought, I'd love to hear it. Signal at patrickb.01 or reply to this email.

Until next month.

This newsletter was not funded by the International Foundation for Media Buzzword Adoption.

Impact metrics: Three consultants are already repackaging this for a grant application. Zero dashboards created. No theories of change harmed. One promise kept: this arrived in your inbox. And you read this far: that's a great metric! Thank you.