Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
I think a lot of it is companies and people trying to get too far ahead and
SPEAKER_00focusing too much on value or ROI right away. When a lot of this, like you have to experiment, you have to fail. That's the only way you're really going to get to the true value over time.
SPEAKER_01Hello, and welcome to The Media Co-Pilot. It's a podcast that explores how AI is changing media, journalism, and content creation. I'm Pete Paschel, longtime tech journalist and founder of the Media Co-Pilot, and I'm excited to bring you fascinating conversations with fellow journalists, media executives, and the people building how the business of news will work in the future. This week, I'm excited to talk to John Levitt. John is the co-founder of Elvex, which is an AI enterprise platform, but don't hold that against him. Elvex has gotten a lot of traction in the media industry,
SPEAKER_01actually, and some big names have adopted it, like the Boston Globe and McClatchy. There's a reason for this. LVX isn't some code heavy SaaS platform meant for engineers. It's meant for individuals, people who want to get stuff done faster and better. With LVX, John is helping newsrooms move beyond one off AI experiments towards real operational tools. But what are those tools? And how are managers and leaders at those companies measuring those three letters that everyone's obsessed with ROI?
SPEAKER_01We also talk about the tension between experimentation and application, the things to consider when you want to advance AI in your newsroom, but also might have a staff that's skeptical of the technology, and why the next evolution of AI isn't just about prompts, it's about context. I hope you enjoyed the conversation, and if you do, I'd be grateful if you could leave a review for the podcast. Reviews, especially in Apple and Spotify, really do help the show, so please leave one if you can. Also, a quick note,
SPEAKER_01you may not know this, but a big part of my business at the Media Copilot is teaching AI to editorial teams. I host a monthly virtual AI class for beginners. It's only 60 minutes, and like everything I do, it's focused on practical takeaways for media professionals, people in journalism, PR, or communications. Plus, it's only $49. Each one's a live class, so it's always up to date, and I guarantee you'll walk away with something you can apply to your day-to-day. You can sign up at mediacopilot.ai. Okay, now it's time for my conversation with John Levitt,
co-founder of Elvex.
SPEAKER_01John, welcome to the Media Co-Pilot. Thanks for having me, Pete. Happy to be here. Cool. Well, I'm super thrilled you're here. I think you are, at Alvex, you're particularly well positioned to help me understand something. And that's kind of where we really are at with AI, especially on the ground level. Like, you know, I talk to a lot of people about strategy. I think about big picture things like GEO and sort of what's happening to our information space. But I really would love to probe with you more about what's
going on in the day to day.
SPEAKER_01Like, what are people doing with it? what are the concerns people have with it and like where are people seeing like real like benefits assuming they are um but before that like please just set the table for us a little bit give us a little bit of context on elvex itself you know the platform and then tell me a little bit about your uh your clients and and what their general mood is right now like are they still curious are they excited are they sort of transitioning to a sort of more applicable and real
world, you know, uses of AI.
SPEAKER_01What's the mood out there?
SPEAKER_00Yeah, so I can give you a little background on LVX. So we are building this company because we believe that every company in the future is going to have a horizontal AI platform that empowers their workforce to be the builders of the solutions that they see a need for, for their own team, for their personal use. And so functionally, what we do is we connect to all the different AI models, give them the context of their business systems, their knowledge, their data, and allow them to build agents that can do any number of different things across
the business that are meaningful to the way that work gets done for them so we work with a lot of media and communication companies but we also work with software companies consultancies fintech health tech kind of across the board so we have a a really interesting position to see what are the differences between industries and
SPEAKER_00as you're talking about the mood um you know i think A couple of years ago, there was much more of a perceived threat of how is this going to impact me or my company? I think there's still some of that lingering around. A little bit depends on maybe company to company, person to person. But there's definitely been a trend toward excitement. And some of that, I think, is due to the capabilities of the models themselves have just really, really improved to a point where people can start to have
SPEAKER_00interactions with them that feel magical and they can start to see, oh, this can take away some of the drudgery of my work and allow me to focus more on the things that I get really excited about. But it all also comes down to the tone and the direction of the leadership at these companies. And that's a real responsibility that I think sometimes gets skipped in some of these conversations. Executives have a lot of the power here in terms of how they want this technology to be used across the company and how they want people to kind
of take the reins of
SPEAKER_00it and make it useful.
SPEAKER_01That's interesting. I'd like to click in on that a little bit because I think a common thing you hear about when people really start applying AI or trying to apply AI team-wide is that there's a bit of a disconnect sometimes between the people empowered to sort of make the decisions around it and the people who are going to be using it. And not just sort of just probe that very specifically. It's like there might be a disconnect in terms of the amount of experience or even education those two parties have with it. And then just expectations and also skepticism,
SPEAKER_01because frankly, like at the sort of ground level, there's probably a little more anxiety, right? Because it's like, is there a job replacement? Yeah. you know, what's my relevance here is sort of AI continues to encroach. And then probably concerns about quality and sort of the specific outputs. So like how much of your job is kind of like bridging that and maybe even educating one party or the other? Tell me about that journey when clients and customers sort of come to you, like where are they at and how do you sort of walk them through it to sort of get
SPEAKER_01to something real?
SPEAKER_00Yeah, it's a really good question. And we're a SaaS platform, we're not a consultancy, but we take very seriously the responsibility of how we introduce this technology to people. And we have expertise and experience that again, runs across industries, runs across companies. We've seen things that work. We've seen things that don't work. And we want to, you know, make sure that our customers know, like, here's our advice here. Here's how you can get started and guide them through that. So we find it really important that we're hands on with
SPEAKER_00onboarding, training, strategy all the way through. And some customers listen really well to us and we have successful implementations. Others want to do it their way and sometimes their way works, sometimes it doesn't. But that sort of conversation is really, really important. I think when you talk about the potential for anxiety for staff within a company, a lot of that, again, comes down to the leadership and how things are being introduced. So For any leaders that are listening here, I would really heavily encourage that you have to be really
mindful about how you
SPEAKER_00are talking about why AI is important. Is it important just because you need to say it so that markets can think that you're doing the right things and they'll value your company more? It's not a good reason to do it. Are you doing it because you truly believe that there is value for the company? That's a good start. Are you doing it because you believe that there's actually value for the individuals that work in the company and the customers that you're serving? That's a great reason to do it. And if you can reflect that this is something that is going to be important, not just for the shareholders,
SPEAKER_00not just for the company, but for the individuals. And here's how we help you grow and upskill. That's something that people can get behind a lot more.
SPEAKER_01Yeah, I definitely want to get to the value part of it and sort of see what people are getting and expecting versus what they're sort of seeing in terms of gains. But before that, I actually really want to drill down a bit on specifically your media customers. And I know you have some major newsrooms like the Boston Globe. as customers. And you can feel free to, you know, tell me which one if there's others also like where they sort of skew, right? Like, are they more bigger newsrooms? Or are they more local?
SPEAKER_01Are they more specialized? And but mostly I'm curious about use cases like for for these various types of media customers that you might have. What are the most common ones? What are the most effective ones? Like where do they go? sort of out of the gate? And then what are you seeing that, how their expectation with a platform like Elvex sort of jibes with what they actually end up doing?
SPEAKER_00Yeah, so our experience is mostly with larger organizations like Boston Globe, McClatchy, Gray Media, Graham Digital, some smaller organizations as well. So there is a good mix there. And I would say, you know, the use cases we see are pretty varied and our tool can be used across the entire organization, not just in newsrooms. The initial implementation sometimes is more focused on certain departments and teams. So we might start in a newsroom with one organization, we might start in a sales organization in another company.
And it's actually really interesting maybe to dig into some of those dynamics for,
SPEAKER_00you know, why do we get started in one place versus another? And I think, again, that comes down to like the leadership initiatives and just where the focus is and not trying to do everything all at once, but trying to build and scale responsibly when you're introducing new technology and solutions. As for practical use cases, there's a lot within newsrooms and content delivery in particular that is about repurposing content, taking it from one format, putting it into another, chopping it up for social, editing, fact checking,
SPEAKER_00which I think honestly maybe doesn't get talked about enough. We can talk about it here. We can talk about it here. Yeah. Research also really important. You know, how can we comb through interviews? How can we, you know, a use case I really like a practical one is if you have a draft written, create an assistant or an agent that is a persona that matches your audience. And have the agent ask you, you know, what questions might the audience have after reading this? What gaps do I need to fill?
SPEAKER_00And, you know, in that way, I think a lot of the anxiety that we see in newsroom sometimes is about, well, you know, there's the, what about the potential for factual errors? And, you know, AI isn't going to write as well as I do. You don't have to use it for that. Use it for the things that it can be useful for, not for the things that, you know, other people say you should be using it for.
SPEAKER_01Yeah, well, you did mention fact checking, though. So I don't actually wouldn't mind double clicking. I've written about this before, because I do think it's like, for someone who might not be immersed in this, it sounds a little silly, right? Like, it's kind of like, oh, AI is known for hallucinations, and humans have to check it. how could you possibly use an AI to then check the work? And it feels like, and again, this is, I want your thoughts on this and how you sort of implement this or how people have implemented it with your tool rather. Like it feels, people sort of say,
it feels like you're just amplifying the errors or something.
SPEAKER_01What's the misconception there? And then how are you seeing people or rather your customers use it for this effectively?
SPEAKER_00Yeah, so it all depends on the design of the system, right? So like I can say, even for our own use cases, we have some agentic workflows that we've built that can help us turn, you know, transcripts from customer conversations into proposals. And those proposals need to identify like, you know, here's the use cases that were discussed. Here's the tools that were mentioned that need to be integrated with so on and so forth. And so we can pull that from the transcript, create the proposal,
but then we have a step to loop back into that transcript and do a search and confirm that the details that are referenced in the proposal are actually there. So same sort of thing you could imagine in a journalistic workflow, right? Like we're creating content from an interview that we've recorded or from a city
SPEAKER_00hall meeting or from research that we've done through science journals or whatever it is, as long as we have that documentation available, we can ask the agent to loop back through it. Does that necessarily replace all of the responsibility of the human to go through and review and fact check? Of course not. But I think just being realistic about the state of media companies right now, news organizations, We have fewer fact checkers and editors than we've probably ever had for the volume of content that we're producing. And so we need to find solutions that can improve upon the
quality of work that we can put out.
SPEAKER_00I don't have the expectation, whether it's through AI, other technology, or humans, that we're going to be 100% accurate with all of the work that we're doing. The point is, can we improve on what we're doing today? And not for the purpose of cutting headcount or reducing costs, just for the purpose of taking the resources that we have and doing more with them.
SPEAKER_01Nice. And yeah, the way I sort of think about fact checking for the AI, it's less about sort of that amplification thing. And it's more about like adding another filter, you know, like you have one layer of AI that might have contributed on some level. And then another one that is specifically designed to catch things. So it's not perfect. It doesn't guarantee things to your point, and humans still need to be in the loop. But having an extra filter probably helps more than it hurts most of the time, again, if you're implementing it the right way.
SPEAKER_00Yeah. One of the ways that we would recommend doing that too, again, depending on a little bit on the use cases and the tools that you have available, but you might have a fact checker agent that is reviewing the work of the editorial assistant, right? And because you are separating these into two separately designed agents that are programmed to do a certain thing in a certain way, you might have more confidence that the fact checker is doing things without getting biased by the conversation with the prior assistant, right? So sometimes removing context,
we always talk about like adding more context. Sometimes removing context actually can be helpful with some of these tasks.
SPEAKER_01Interesting. I'm going to dive into that in a minute. But I'm also curious just on the use cases. Well, just not less on the use cases, but like you said, obviously, it's not just editorial, like, you know, the whole company can benefit from LVX. Do you see any like at media companies in particular, do you see any sort of enthusiasm, more enthusiasm from any particular sort of department like product, data, growth? Who's who are the most voracious users of AI?
SPEAKER_00We do see a lot on the revenue side. And when you think about what's required there, again, a news media company, a lot of the work that they are doing to service advertising clients has to do with campaign analysis. And that's a lot of data that's relying on other teams to pull the data and to put it into structures that teams can then work with and spend lots of time putting into sales decks and presentations and You know, coming from a sales background, like that's the type of work that salespeople and account
managers don't really
SPEAKER_00love doing. They want to talk to more customers. They want to help more people and worry less about just copy and pasting metrics into a different format. So there's a lot of value that can happen there. Proposal building is another use case that we see on the revenue side a lot. We see product organizations, you know, like you have user surveys and feedback that just might overwhelm you with the amount of data that you come back. So you either let it sit and go to waste or you'd find a way to make it useful. And luckily, I can help us extract important insights and,
SPEAKER_00you know, immediately take action on things that are important to take action on. So, you know, those are all, you know, really consistent things that we see our customers doing. I'll say like odd. One other observation that I have from working with our customers and, you know, I think everyone is very excited to jump to the highest ROI, you know, use cases, right? Like that's the thing that you want to do. I mean, you know, let's, let's take something.
Just to show the example, something really, really ambitious for a news company.
SPEAKER_00Let's take our entire 200 year archive, digitize it, create a chatbot on the site that can interact with extremely high accuracy with any question that's thrown at it and provide a totally radically different experience to news consumers on our website. If that's the first starting point for an AI project at your company, I would say, let's dial that back significantly and start to create value, even if it's something, you know, relatively simple,
like everyone has an HR team in their company that has to answer questions all the
SPEAKER_00time. And HR assistant is a pretty simple agent that you can build, especially if you have LVX. It's like under five minutes to build that, integrate it with Slack, and then you're good to go. Why is this important? Is this going to be the highest ROI thing you can build for your company? No. Is it a good way to make people see what's possible with this technology and get inspired for other things they can do that might help them and their teams? Absolutely. So I think those those small steps toward the bigger ambition is something I always advise.
SPEAKER_00Like, let's let's make sure we're crawl, walk, run before just trying to sprint in the Olympics right away.
SPEAKER_01I'm glad you went to the ROI thing pretty quickly, because I definitely wanted to ask you about that. I know we spoke at ONA a little bit ago, and that was a big topic that you were talking about there. And so, you know, I'm curious, specifically, like, obviously, everyone sort of wants to know, wants to move the needle on something with respect to AI, implementing it even at smaller organizations, you know, there's some resource that you need to dedicate to it.
SPEAKER_01How do you sort of walk them through that in terms of like, whether it's metrics to look at, expectations, you know, starting with an HR thing, that sounds great. And then I would think with that, well, at the very least, not only the inspiration part of it, but it's like your HR team isn't answering a bunch of dumb questions anymore. So maybe they can actually do some stuff, which might be hard to measure. But how do you guide people in that sense?
SPEAKER_00Yeah, so very important to know what the starting point is for the company and the people within the company. And we do work with some companies who literally like no one has ever generated an account with open AI or an API key or anything like that. And so, you know, what is the ROI for that organization versus someone who's already got, you know, a bunch of engineered AI solutions across their business? The target might just be, hey, six months from now, 12 months from now, we have people who understand this technology and have started to build some things that we could see becoming valuable in the future.
But I think you also look at stats on the failure rate of proofs of concept and
SPEAKER_00things like that. I think a lot of it is companies and people trying to get too far ahead and focusing too much on value or ROI right away. When a lot of this, like you have to experiment, you have to fail. That's the only way you're really going to get to the true value over time. you know finance leaders don't like to hear that right they're like i'm approving a budget uh i have to know why i'm doing this but it is sometimes that soft roi that's the starting point and then when you get a little bit more mature you might
say hey yeah let's take like a newsroom use case well if we can produce x more uh content on you know a subject area that we don't cover really well right now but we
SPEAKER_00know we have lots of interest in What does that mean for advertisers that we can bring into the business, an audience that we can grow? If it's on the sales side, if we can respond to five times more RFPs than we can handle right now and we win even a couple more deals, that could be a significant boon to the business. So you got to try it out, right? Like there's no guarantees here and people actually have to use the technology and build with it. But I think that's like the right way to think about these things.
SPEAKER_01And would you say that's like one of the main selling points you have? It's kind of like, obviously to be able to experiment with things, there's always concerns about data privacy. There's always concerns about like, yes, experiment, but also do it in a safe way that a platform like yours is basically like, okay, you can easily create sort of sandbox. And it inherently is kind of a private sandbox environment and you can do that. But essentially, how do you help people, help companies,
SPEAKER_01I guess, like figure that out of like, okay, here's the stuff we're actually doing and here's the safe space to experiment in? Because this is not talked about all the time, right? You need to, people need to experiment, people need to experiment. Well, how do you do that safely with, you know, your data.
SPEAKER_00It's one of the key pillars of our platform. Any of the work that you're doing in our platform is not used to train models or to do anything other than just provide the service back. And when you look at a lot of the fear and anxiety that persisted across businesses, especially when chat GPT first became popular, but even today, it's just the fear of, oh no, if someone puts in private data into one of these models, what's going to happen with it? How at risk are we as a business? And, uh, simultaneously you have those you know ceos in front of
investors saying ai is going to transform our business we're going to use it everywhere um but then they're saying but you can't use it with any company data or personal data or anything like that well how am i going to make it useful if i can't bring it into the context of how we're working it's a requirement right so that's one of the um
SPEAKER_00uh like baseline challenges that companies need to solve The other inhibitor I would say there is, again, going back to that fictional CEO saying it's important for everyone in the business to start using AI. But then you have the problem of how do we get people access to AI? And most commonly, software companies charge per seat. And that gets really expensive to scale out across a company of 500,000, 10,000 people, especially when the expectation for current day adoption is like 10%,
SPEAKER_0020%. That's 80% of the expense that is going to waste. So why would I do that? So, you know, we have adopted a bit of a different model. We've said all of our contracts are unlimited users. So we have to create the path to invite everyone to experiment, knowing that they're doing it in a safe, governed, sanctioned space. And then, you know, you kind of pay appropriately for what you use from there. So I think also like there will be a trend in software, especially with the emergence of AI,
SPEAKER_00maybe away from seed based pricing and more toward usage based pricing. And that may change the way that the ROI conversation also gets evaluated by some of these companies.
SPEAKER_01Interesting. Okay. So it's essentially metered is how you do it as opposed to the seeds. So it's like the more you use it, the more you'll pay, but presumably you'll also get more return from whatever you're doing.
SPEAKER_00It seems like a smart way to do it. Yeah. And presumably you're starting at a lower cost from day one too. So less risk, more opportunity. Yes, you may pay more over time, but If it's truly an investment, then you're happy to pay for value, right? So that's how we're thinking about it. And that's how we hope our customers see it too.
SPEAKER_01Well, I also think that model allows or encourages anyway and incentivizes the business from minimizing that compute, right? And so like to your point of like being able to choose any model, you can get the most efficient model. for whatever your use case is. And I also find this might be, I don't know if it is, actually I'd love to get your thoughts on this, on sort of the skepticism of AI and concerns about it.
So like something that I feel like is bubbling up is because of all the news about
SPEAKER_01AI infrastructure that's always constantly happening, there's more rumbling and even objections to using it because of its immense energy use. You know, and so and I've seen this on sort of a ground level at ONA. In fact, I saw you might have seen this, too, while you were there, like a couple of people come up to microphones and say, I don't want to use AI because it's killing the planet, which, again, I'm not trying to minimize that concern. It's obviously a big macro concern around all this stuff. But from business leader standpoint, you want people to be experimenting and doing stuff.
SPEAKER_01while addressing that concern. So yeah, it just seems to me that idea of like, let's experiment and then optimize around it is a good way to at least show you're doing it in sort of a thoughtful way as opposed to just, oh, we're just going to take some blanket cost and go ahead and use ChatGPT 5 Pro like all the time or whatever. Yeah.
SPEAKER_00Yeah. I mean, technology is expensive in a number of ways, regardless of what type of technology we're looking at. AI is, you know, in the early days in its current form. And you're already seeing over the last couple of years that we're finding ways for it to become more efficient. And I think that continues to happen. Now, simultaneously, as the usage volume goes up, the obviously the net effect there may continue to increase for some time. But I think AI probably will be impactful in figuring out how
to make itself more efficient.
SPEAKER_00Right. So I think there's still more to go, more investment. And, you know, this all has to be balanced with other benefits and costs as well. Right. So, yeah, you know, there's going to be some impacts in some ways, but, you know, not this isn't something that LVEX is working on. Right. But like. AI broadly is helping scientists and doctors cure illnesses and find solutions to other problems that exist. So with everything, there's always some trade off.
SPEAKER_00And I think we have to, as an industry, be very considerate about finding the right balance and not act irresponsibly in the way that we're building this technology and making it available to folks. But I do believe that as we progress, we will find more ways to make it more efficient and more beneficial and like overall industry-wide ROI. Nice. I feel the need to switch gears here because we drifted into sort of macro stuff. And again, I want to get into sort of the nitty gritty of it with you a bit.
SPEAKER_01So, you know, given how... people like have, I wouldn't say traditionally used AI, but like how the usage of AI has evolved over the last couple of years. I feel like, you know, a couple of years ago or when AI was still relatively new, and this is still true to some extent, prompt engineering was a big focus. There was a lot of frameworks and talking to the AI the right way. As probably predicted, the AIs have gotten better at sort of understanding what we
want.
SPEAKER_01There's thinking models now and all this. So, you know, the buzzword I'm hearing more and more is context engineering, which sounds right to me. You know what I mean? We're getting away from this kind of like you need to sort of talk AI and now it's just... You can talk like you, you just have to bring it the right information. Are you seeing that too? And then if you are, how are you adapting Elvex and its approach to sort of match that moment?
SPEAKER_00Yeah. So interestingly, maybe a year and a half ago, I had a friend who was kind of considering like what next steps they wanted to take in their career. And they asked, like, I think I'm going to go back to school and study prompt engineering. And I said, I think that's probably a really bad idea. I don't know that there are going to be a lot of prompt engineer jobs in the future, five years from now, however long of a timeline we want to put on it. And the reason for that is,
SPEAKER_00yes, the models are already getting much, much better about following instructions without having to engineer the prompt that you're giving to it. Companies like Elbex are also considering how do we take natural language instructions and turn them into the right prompt for the models. And that could be not just a generic prompt for all models, but... each model provider and each model within those family of available models might require a little bit of a different instruction and so like
that's built into our platform already we will help make those decisions for the user we can route certain prompts and requests to the right model so that it unburdens the user from
SPEAKER_00that And I think people that have understanding of how to correctly engineer prompts without needing the assistance of a tool like LVX to do it will have more of an advantage. But I don't think that skill in and of itself is something that is going to be a job or at least a heavily available job. additional step of, you know, context awareness and thinking about like memory tools that some of these providers have implemented, all super useful.
I think really important to recognize when those features are being used and not used because it can change the way that you think about some of these things.
SPEAKER_00So like if I am using a chat, like the chat GPT app on my phone, which I will use for personal use, And it does have memory of certain things about my life and my interactions with it that can be useful. It can know that I'm like going on a trip in a month. And so it might use that information to guide some of its response. However, if I'm creating an agent that is being used across my company, I don't want that personal context and that personal memory to
influence how others
SPEAKER_00are interacting with it. Or if others are trying to change the instruction for the agent for a specific conversation that they're having or task that they're trying, I don't want that to persist, right? Because I've developed this agent to work in a really, really structured way with the context I've given it. And I don't want that to change unless someone needs to change it just for their, you know, context or, or, uh, sorry, uh, conversation. So when designing these systems, like the intended use case and user, and like, is this a business or is this a consumer?
SPEAKER_00Like all that stuff really, really matters to design the right systems for the, uh, for the intended use.
SPEAKER_01Yeah, yeah, it's interesting you point that out. I do think that's kind of a, it's an interesting differentiator that they've introduced, particularly with chat GPT, the sort of ongoing memory. I mean, they've had memory for a while, but now it's like kind of more persistent and they're remembering more stuff. Which is, to your point, it's really good for the individual. And but, you know, people, when you're starting to use this across teams, it's like you want different, honestly, you just want different context. I mean,
and also like me as a professional, like I, you know, I'm kind of mixed in all the time.
SPEAKER_01That's kind of my own company. But it is like, as an employee, I want different contacts, certainly, than what I'm doing for my company, than what I'm doing at home. So I do feel like there's a perspective of an enterprise platform like Alvex would be very difficult to get right. So beyond... you know, the obvious differentiator for a platform like yours is right off the bat. You can use any model, right? And then on top of that, I think it's sort of this organizational context that not that
you can't get with a
SPEAKER_01chat GPT enterprise or whatever, but I don't know, might be more more you'd be able to be more thoughtfully route that um if that makes sense i don't know you tell me if that's a correct yeah well you know i think when you think about the organizational context and knowledge and information that is required from a business perspective for these tools to really really be valuable um it's a massive amount of information right um and the restrictions that you have
in play by uh the model providers directly like if you're creating a custom gpt or
SPEAKER_00um, you know, something similar with a different platform, there's only so much data you can provide it probably, you know, 10, 20 files like that. Um, which is not organizational context. That's just like ask context or task knowledge. Um, and there's real opportunities through, you know, like providers like LVX where you can say like connects to a whole bunch of systems and platforms and use that as you need to, to complete the tasks. So, um, That is like going back to part of the initial question about
where is the industry
SPEAKER_00and like what education is necessary. These are like intricacies and details that I would say the average person is very far from really understanding at this point, just because it's like not something that's accessible to them in a consumer way. And so how would they kind of know what's available in a like an enterprise business sense? Right.
SPEAKER_01We've tossed around the word agent a lot, and I know it obviously gets tossed around all the time. How do you guys define it? Because sometimes people, I don't know if they're being nitpicky or not, they're like, that's not an agent, that's an automation, that's not an agent, that's an assistant. Where do you see the lines between agent, assistant, automation, and whatever other A words might be applicable here?
SPEAKER_00Love and hate this conversation because the word agent in particular, and even AI is so overused. I've been joking recently, like imagine if a company introduced itself today as an internet startup, like what? That's 25 years out of date, right? I think the same is going to happen with AI. AI has become persistent in almost every software company and beyond. So I think there's a short timeline on saying like, oh, we're an AI startup.
SPEAKER_00It's just going to be like a common part of the technology stack. Agent also like very overused, very misused. You see like a simple chat bot on a website being called an agent. You see things that are truly agentic also being called agents and like these are very different things. For me, I mean, If we're talking about something that can make decisions and kind of like operate beyond the scope that has been pre-programmed into how the piece of technology is working,
SPEAKER_00I would say that has agentic qualities. But honestly, for the sake of conversation and marketing, we may use agent in ways that may not match that just because it might simplify and help the other person be on the same page with us. So a little bit of love and hate for that conversation, to be honest.
SPEAKER_01Yeah. All right. I'll let you get to use your get out of jail free card for that one. So just let's get as we sort of wrap up here, let's look ahead a little bit. You know, you've obviously guided various media companies on how to leverage AI across their staff. You've seen a lot of use cases. You've seen all the questions about ROI and whatever else. So what do you think is the biggest shift that they're going to have to get ahead of in the coming year? We already talked about context engineering,
SPEAKER_01but like as the technology evolves, as it gets more implemented and it just sort of keeps changing the both our ecosystem and how MediaWorks done, like what is the thing that they need to embrace, get ahead of and use it to thrive over the next year? I mean, we could have a whole series just on this topic, but I'll try to get into this.
SPEAKER_00Yeah. So one is just mind shift culture. You know, we've talked about it a bit already, but like how leadership is actually enabling and expecting people to make use of this technology when they're using it on like a day-to-day basis. That's, you know, that's really important and core to the work that we're doing at Allback. So that's very top of mind for me. I think bigger picture for media companies, you know, there's a whole shift that is coming and already here in terms of how people are engaging with news content,
SPEAKER_00media content. And, you know, there was a conference I was just at yesterday where there was a Silicon Valley VC talking about get ready for the shift from, you know, be basically like business to human interactions and It'll be more business to agent interactions. And then eventually it's going to be agent to agent interactions. And he wasn't specifically talking about news and media, just kind of like general marketplaces.
SPEAKER_00And I think that's probably a reality. Like we're already seeing some of that, right? Again, what is the timeline horizon to when we get to a point where it's, you know, we're creating experiences for agents to interact with each other. um i think the silicon valley dc would try to represent that that's right around the corner i don't think it's right around the corner but i think it's maybe closer than your average person might expect so what does that mean for news and media organizations what does that mean for how content is delivered
and consumed and is everything now hyper personalized uh you know how do we monetize this stuff there's
SPEAKER_00lots of questions that are going to have to be answered fairly soon If I had the answers to all those questions, trust me, well, I might not tell you because that'd be a great business opportunity and I'll tell you next time when we figure it out. But yeah, there's a lot that I think the industry collectively needs to figure out.
SPEAKER_01Nice. I'll go halfsies with you on one of those ideas that we have to. John, I think we'll leave it there. Thanks so much for stopping by.
SPEAKER_00Yeah, absolutely. This was a pleasure. Thanks, Pete. All right, take care.
SPEAKER_01Thanks for listening to my conversation with John Levitt of Elvex. If you enjoyed this episode, I'd appreciate it if you just take a minute to leave a rating or review for the Media Co-Pilot in your podcast app. Also, don't forget to subscribe to the Media Co-Pilot newsletter and check out MediaCopilot.ai for our classes on AI for journalists, PR pros, and communicators. Be seeing you in the future.
For years, Yahoo News has occupied a peculiar position in the media ecosystem. It is not a social network, not a traditional publisher, and not a search engine in the modern sense. Yet millions of people still use it as...
This special episode of The Media Copilot is a look back at the conversations that defined the past year… the questions, tensions, and turning points shaping how media, journalism, and technology intersect right now....
AI has made verification a newsroom survival skill. In this episode of The Media Copilot, Pete Pachal talks with James Law, Editor-in-Chief of Storyful, the newsroom partner known for verifying the videos and social...
AI in journalism can feel abstract until you talk to the people actually shipping products inside newsrooms. In this episode of The Media Copilot podcast, host Pete Pachal talks with Darla Cameron, Chief Product Officer...
AI is rewriting the rules of digital media and few people have had a closer view of the shift than Adam Singolda. In this episode of The Media Copilot, host Pete Pachal has a candid conversation with the Taboola founder...