Way back in 2017, I predicted in a VentureBeat guest post that PR people will be getting superpowers in a single year. Unfortunately, I was off by about seven years. Regardless, it’s here now, and it’s far bigger than I imagined.
The influence of AI on journalism and, in turn, on the profession of public relations is undeniable. It is indeed altering the words and ideas we put on a page.
I recently had the pleasure of a virtual sit-down with Pete Pachal, founder of Media CoPilot, a new media company dedicated to demystifying AI for media professionals. Our conversation delved into the current sentiment of AI among journalists, the impact of AI on content creation, and the innovative approaches being employed by forward-thinking writers.
Below is the abridged Q&A, edited for clarity.
Listen to the uncut interview on the Glitch Podcast on Spotify and Apple.
Aaron Cohen (Glitch PR): How and why did you start Media Copilot?
Pete Pachal (Media Copilot): I’ve done quite a few different things in media and have watched it change quite a bit over the last few years for a lot of reasons. There’s stuff going on with search, social, and audience relationships. The thing that’s clearly changing is the craft at all ends, in terms of the business, the content generation, and how roles are being defined. I could see that happening all throughout 2023.
I was at CoinDesk when ChatGPT came out. Like many other publications, we quickly launched a taskforce internally to examine this, and to figure out if it’s something we could use. I was in charge of that taskforce at CoinDesk. We tried a bunch of different things. We prototyped a few different ways to use AI. The thing that was happening the most around then was kind of infamous face-planting using AI. A lot of publications were experimenting in this way. CoinDesk tried to use AI to essentially scale up content and have AI write articles. Some of the articles had people checking them, but they weren’t the right people. Some of them weren’t checking them at all. Those were more fly-by-night operations. I concluded that this doesn’t work well as an article generator. For AI, that’s just one use case.
There are so many other points in the news production—gathering, and the distribution pipeline where AI can speed up things, create efficiencies, possibly even surprise you. It’s clear that it has been changing, and will change journalism, news, and media consumption for years to come.
I found myself not at CoinDesk because it was part of the uptick in layoffs. The idea of getting ahead of this AI wave, or at least trying to ride it a little better, seemed like the most logical thing. So I founded The Media Copilot.
At the time when I did it, I honestly didn’t know if it was going to be a newsletter that I was just doing in my spare time while I was looking for my next gig or something bigger. It rapidly became something bigger. There was just a tremendous amount of interest. I got a lot of signups. We have thousands of subscribers now. I started offering classes and those went really well, whether I was doing an hour, or three hours, or an hour and a half. I was teaching myself at the same time of just learning all these models, all these services, all these different tools that create content for you with varying degrees of comprehensiveness and ease.
I passed that knowledge on. It just seemed like something was really taking off here.
Now we offer classes on a regular schedule as well as custom classes. We’re consulting with marketing companies, PR agencies, industry associations, journalists and one-on-one. It’s hard to keep up with demand! Now, I’m faced with the issues of building a business and figuring out how to prioritize. “How do I make time for the business side, like squaring the books, while also still putting good content out there?” That’s still the best engine to get more interest.
So that’s my background and how I ended up starting The Media Co-Pilot. It was a bit of a survival mechanism to some extent, but I’m surviving!
Aaron Cohen: Do you have an entrepreneurial background? Are you a family of entrepreneurs and founders? Or are you the first in the lineage to try this, to go out on your own?
Pete Pachal: No! Not at all. No family, entrepreneurs, no old money here. It was just putting myself out there and taking a big risk. I don’t have a lot of passive income. It was trying to leverage the best of my experience and aligning that with the industry I know really, really well. I know a few things well. I know tech, media, and to some extent I know crypto and crypto media. I’m happy to say AI is one of the big things I know a lot about. Although there are days where I feel like I know nothing, because the pace of this industry, even if you restrict yourself to generative AI and content creation, is so fast.
There are so many new things happening every week. Just this past week, there was Sora from OpenAI, and a new iteration of Google Gemini, at least as a product, because now they’ve done away with the Bard name. Every week there’s all these new tools and new products, new businesses being created with AI. It is just breakneck. I get whiplash every time I look at my inbox and see all the news I’ve signed up for from all these newsletters, because everyone’s covering it too.
Aaron Cohen: What is the elevator pitch for Media Copilot? Did you have an “aha moment” to do this? If so, what was that like?
Pete Pachal: The mission to demystify and remove the fear people still have with AI.
It’s funny, I’m starting to get in a bit of an AI bubble when I talk to people and the services I deal with. You look at these online communities and you get the sense that everyone’s excited. Then I come out of that bubble. I talk to journalists and others who are still very skeptical of AI. That skepticism is born out of fear to a large extent. I don’t think it’s quite as quid pro quo as “AI is going to take my job.” A lot of the people that I know are fairly senior and understand what their own value is, and the companies they work for understand their value as seasoned people.
What I think they’re looking at is a bit of a real fear to some extent with AI, that it’s going to atrophy good things about journalism and good writing. There’s some merit to that. A lot of the fear is unfounded with regard to the journalism profession. Because even though there are a lot of places that we’ve seen AI for “nefarious” ends, whether that’s for marketing or just cheap content, the practice of journalism still requires humans and human insight—human ingenuity. That hasn’t changed. I don’t think that will change for a long, long time. Sure, everyone talks about AI systems getting better by the day. While that’s true, the curve of AI is that the writing today gets us to about 90 to 95% of what a human can do. If you accept that as a general rule, the first 95% was a lot easier to do than the last 5%. The last 5% is where journalists can make a difference. The last 5% is not going to be conquered in the near term. As good as these systems are getting, they’re going to get incrementally better going forward. As much as everyone likes to hype up AGI (or whatever the hell that is), I don’t think we’re going to get a human equivalent intelligence for a long, long time still.
Aaron Cohen: So we’re not going to have an AI [agent] call up a company for comment? They won’t be talking to journalists anytime soon?
Pete Pachal: I think AIs will start calling for comments, but to understand that comment and repeat back a question that is actually useful and probing, where they need to probe with that human intuition, that’s going to be very difficult to simulate. That’s simply not going to happen for a long, long time.
If I could just give one pragmatic example of that. One of the things that happened to journalists and reporters in the 2010s was that we all had to become content marketers for a while. You all had to figure out how SEO worked, what the social networks were and how to speak the language of those networks just to get your content out there, competing with everyone else’s content. One of the promises of AI is there’s going to be a bit of a retraction of that, partly because social just isn’t worth what it used to be. Same with search to some extent, but also because AI can handle it.
I strongly believe that in most Content Management Systems (CMSs) within the next six months to a year, we’ll have “generate for me” buttons that essentially does what we would consider the Google Chrome around content. You write an article, then the Tweet, then the Facebook post, then all those meta headlines, then the SEO headline and come on, no one wants to do that. No one ever wanted to do that. I mean, there’s a few, there’s a few people who are great at it. God bless them. I got pretty good at it, but no one looks forward to that. No reporter anyway. Now we’re going to see pretty quickly that being done by AI. It’s essentially a dropdown menu of auto-generated responses that you can tweak. That’s something journalists could get behind.
Listen to to the full episode on Spotify here (go here to listen on Apple):
Aaron Cohen: Are journalists pro AI?
Pete Pachal: In general, there’s still a bit of a stigma. So not pro, though more and more every day seeing the value that it can bring to tedious tasks. AI will get used more in content in distribution. Journalists need to adapt or die, frankly. As a profession.
Now, all that said, the concerns that some of them have around these tools are valid but exaggerated. For example, most journalists that aren’t just starting out, they’re probably split between writing two kinds of articles—the kind of news to crank quickly versus something that is bespoke, like enterprise reporting or a feature story. Those are what you really want to spend time on. Features, or enterprise reporting, is going to remain the same. That’s going to stay human because you want it to be human. That’s really what’s adding that sort of magic. I love those! Everyone loves those articles. Come on, you’re sitting back with a nice long piece from New York Times Magazine on a Sunday afternoon—if you’re into news and into reading, that’s awesome!
And you want that human connection. But for those quick turnaround stories, particularly ones that are pretty rote, whether it’s rewriting a press release or just a quick piece of news, a quick hit, as we call it in the trade, a lot of that will either get AI-written or AI-assisted.
To be clear, none of that should be going out unfiltered without humans in the loop. Humans are necessary for checking it because any AI can hallucinate. The writing tends to be fairly soulless. There are things you need to do to help string thoughts together for these systems. But again, a fine-tuned system, particularly if you’re untrained on your publication style of writing, can do a pretty decent first draft. You just need to put it together. That’s what Samafor is doing with Semafor Signals. That’s a brand new product they just announced the other week where they get an AI to both find and summarize news from around the world on a particular beat. Then a journalist comes in and strings it together and adds their own knowledge. That to me is the right way to use AI. Particularly, I would even say this is probably a necessity to have the human be someone with subject matter expertise.
There was a tech publication that was doing AI-generated articles. They were being checked by humans. The problem was they were junior level editors without subject matter expertise. So even though they could correct for grammar, spelling, and style, and even to some extent take some logical leaps, they didn’t have that granular knowledge when the AI would just make something up that was false.
Aaron Cohen: What do the publishers, the lawyers and all of these people behind the scenes, the editors and managers think about the journalists using tools that actually write the first draft?
Pete Pachal: The most forward-thinking places have cautious optimism where they have learned from the missteps of those publications that went too far too quickly and are looking at a process that really is more like a copilot situation where most of the articles are human written. You can lean on an AI for SEO, to fill out parts of an article so that the page is better equipped to rank on Google. There’s a bunch of tools that do this better than Chat GPT or Claude. But that said, there are guardrails. Subject matter expertise is the main one. Ultimately, by the time it gets to a publication, you might not even see a note about AI having contributed because it was so deeply embedded in that process. You could probably still detect it if you ran the article through an AI detector, but it would be only a chunk of the article somewhere, probably lower down. Again, those are the folks that are doing it ethically.
Again, I’m not talking about content operations that are essentially farms that are just cranking out stuff. That’s happening too and it’s affecting the whole ecosystem. But the sentiment for forward-thinking folks is cautious optimism. Editors are implementing it. Bit by bit, they’re figuring out the ways and places that AI can help. The business managers love it in this sense that they were the ones certainly pushing to cut down on overhead and crank up output. It was a mirage. You can’t really do that because then you’re Sports Illustrated. I’ll name them just because it was such an infamous case and essentially the publication is now shuttered, not necessarily because of that reason directly, but certainly contributed. The business managers are all, “Let’s do more AI where we can.”
There are some roles that will likely, if not fall by the wayside, change or become integrated into other roles. I mentioned social media already. That role will become much more administrative and less content based. Obviously, you still need someone to manage it, but in terms of writing the tweets and all the things that go out on various networks, that’s mostly going to be taken over by AI. Another one is copy-editing. The copy editor role and copy desks have been going away for years, and certainly with AI now getting really good at not being an editor—the better services can show you a paper trail. You can go back and see not just what changed, but why. For almost all publications, that’s going to be handled by software, which is too bad because a real-life copy editor does have that extra [knowledge] that they can give by being a subject matter expert. If you’re at a traditional place like The Times, you’re copy editing on a specific desk. You will know which questions to ask and how to fill in stories. That will be at this point still better than Chat GPT, but the services are getting so close to that.
Aaron Cohen: One of the roles of journalism is seeking out truth and keeping the foundation of our democracy in check, holding folks accountable. There’s a lot of media companies going out of business right now. There’s a lot of layoffs, right? I don’t think it has much to do necessarily with AI at all. Can you comment on that though? Are you thinking that this is a bright future for journalism right now? Are we in a rough patch right now?
Pete Pachal: Well, we’re definitely in a rough patch right now. There’s no question about that.
Aaron Cohen: Is there a bright future though?
Pete Pachal: That’s a really, really, really good question. There’s reason to be optimistic. Philosophically, in the 2010s, when everyone was chasing scale, social media was giving tons and tons of traffic, and Google was giving almost as much. It’s going to be looked back on to some extent as an anomaly. That was a weird time in media with everyone chasing scale. A lot of it just wasn’t real, just traffic spikes and whatever the whims of the Facebook algorithm were. That’s not having an audience.
As painful as this time in media is, and as much as I personally wish it wasn’t the case—because I want opportunities for my friends and myself to be out there like where they were in the 2010s—but this will ultimately be healthy for media because it’s going to force every publication to wonder what value they’re bringing, who their audience is, and to optimize for that.
I’ll get back to sort of the AI question, because this is a big wild card that’s been thrown into everything I just said. Search is still the number one way most people find their news. Search is rapidly getting AI-ified. You could argue the zero-click searches Google was doing even prior to the current wave was part of that.
Google was already, for some queries, just giving you the answer. Now with generative search, if you opt into that, that’s what it defaults to. At some point, fairly soon, that’s going to be the default for everybody. Google is obviously the biggest of these, but we’re seeing that with Bing, and Amazon’s building this technology into Alexa, which will finally make it an actual assistant, presumably, because that’s what was always missing. You could say whatever you wanted to, Alexa would understand you. If it didn’t have that answer hard-coded to give back to you, it just would respond, “Here’s something I found on the web.” But now it can start doing that. When Apple [AI] comes out, and it will, guaranteed, in a big way, probably this year—but you never want to predict too much with Apple because anything could change—once that happens, once iPhone has that ability to be your real assistant and we’re not hunting for apps anymore, look out!
So what does that mean? It means everything behind media has to change to some extent. We’ve got to figure this out, and quick. Is it as simple a matter as media companies striking deals with AI companies, which will include these gateway companies like Apple and Google? Google is already an AI company, but like these people who are the traditional middlemen, that seems to be what’s happening. Otherwise, we’re going to need an answer to the problem of the content being completely ingested by these systems and then not getting anything in return with regard to the media companies.
That’s been a de facto agreement between media and platforms forever. It’s like, hey, we’ll give you our content, index it for free, but you give us traffic. Now, the second part of that no longer applies. So what’s the deal then? It’s probably some money exchange. I think the media companies wanted a B with billions in there. They’re going to come out somewhere in between.
But the problem with the tug of war is that tech always has the upper hand here. Because as we’ve seen with social, with Facebook, Twitter, and others pulling back from sharing news, the media needs them more than they need the media. That’s an unfortunate thing for folks out there in the media to face. But again, it’s just sort of a harsh reality. That said, there’s still value to it. There will be many deals made.
But if that court case with the New York Times versus Open AI gets resolved, if they actually answer this, and there’s some way that the law stands on what that information is worth and how media companies should be compensated, that could change things. But I wouldn’t count on it.
Aaron Cohen: The book industry had a lawsuit against Google. Do you remember the outcome of that?
Pete Pachal: Google won. This is the basis of a lot of what the companies that supply the data to Open AI and others are relying on, which is that the act of ingesting a copyrighted work isn’t by itself a violation for these systems.
I’m not a lawyer, just to be clear. There’s four factors that matter—how much of the work, how much it affects the market, is it transformative, and I’m forgetting the fourth one.
But the fact that you couldn’t just get the whole book back generally, and it was partial, was a factor. These services like Common Crawl crawled the entire internet and have these massive data sets that all the LLMs, all the large language models use, to train their models. Now the telltale part of knowing where your work has been ingested is that it can be regurgitated. This is in the New York Times lawsuit. If you say the right thing, if you cast the right spell with ChatGPT, it will give you entirely, almost word for word, an article back.
Now, to be clear, that’s the weakest part of The Times lawsuit, even though it’s true that this could happen. Generally, it’s easy to patch. No one’s doing that, to be clear. Like five people are doing that, trying to get, like bypass the paywall by asking for articles. And it gets away from the real issue. The real issue is that you’re taking this content and doing something transformative with it, acknowledging that it’s transformative, but doing it at scale, which is really what changes things.
Scale can be a factor. If you look back 25 years earlier, Napster was taking music and scaling the experience of going to your friends and saying, hey, anyone got this track? It’s fine on a small scale, but once you scale that up to a certain level, then it becomes violative. Now, again, I’m not a lawyer because I think technically even the person-to-person stuff was probably “illegal” back in the day, just copying music from someone.
And in this case, it’s probably not illegal to copy something and then transform it and serve it up in a different way. But changing the market and how something affects the market is one of the factors you can look at in a fair use claim for copyright law. I don’t know how the law will rule on this, but things could change quite a bit if this ever actually gets in front of a judge. That’s what The New York Times is waiting for, seeing how the winds are flowing when the case gets closer to actually being heard, while also waiting to see if OpenAI is willing to come and change that million dollars to a billion dollars the next time it comes calling.
Aaron Cohen: It seems like the AI is going to help folks who are on their own, folks like yourself, create another movement, another era of journalism, creating their own audiences and creating their own businesses.
Pete Pachal: This is a whole movement born more out of what’s happening with media than with AI. Again, as socials and search pull back, we’re seeing a correction, and journalists that have a platform can make a go out of it on their own. Again, I think having some experience here, you really have to think through the business model as you do that and how big your platform is.I don’t think there are that many journalists, unless they’re really into doing what they’re doing for the long haul, that can do that quickly.
That said, there are some. I could name Casey Newton, who does a really good job on Platformer, and Jesse Singal, who just walked, and Katie Herzog. Again, they have platforms. They took them to these sort of independent places, or they built them up over time. Casey was their lead guy on social media in Silicon Valley at The Verge before he went out on his own. If you don’t have that platform already, you have to ask, what else can you offer besides just content? A newsletter probably isn’t going to give you anything you can live on, at least not in the short term.
Again, if you’re serving up great content that people want, you’ll find that platform on your own. Building it on your own is super rewarding, because obviously it’s yours and you’ll own it. You won’t have to lose that platform if you ever depart your employer. But again, there’s so many different things you can do in media now, besides just serving up content. I think content is essential because it’s how you get yourself out there day after day. But there’s events, there’s sponsorships, obviously. There’s what I’m doing, which is teaching courses and consulting. If you have at least some kind of platform and network, it’s worth looking into and trying out. My advice for anyone thinking about this now is that it’s a good time to do it.
Aaron Cohen: Why is it a good time right now, do you think?
Pete Pachal: Because it is a good time to figure out who your audience is. This is healthy for media, and healthy for journalists. You probably do have an audience.
You’ve been working at a place for a few years and you’ve been covering a beat. There’s probably people that follow you. You have a network, certainly, and you can lean on that. You can really just start putting things out there. The big advice I would say is don’t overthink it. Don’t waste too much time trying to figure out your business model and what you’re going to serve up, that’s going to change. You just have to start doing it. That’s what I did. Once I decided I wanted to do this, I spent an inordinately small amount of time prepping the launch of the Media Co-Pilot.
Aaron Cohen: That’s so inspiring!
Pete Pachal: I threw together a logo in Canva. I picked a name because I thought, hey, Media CoPilot. Again, this was before Microsoft decided to slap CoPilot on every single thing. So I can’t say I was first. The CoPilot idea has been out there since AI has been around, but don’t confuse me with their products. But I just thought, okay, well, I’ll start doing some newsletters and see what happens. And a lot happened. I’m grateful I jumped in when I did.
