Dave Davies on LLM content SEO shortcuts, attribution loss, and agentic AI

Agentic AI is reshaping search and generative engines – and soon could transform your SEO strategy.

This was just one hot topic discussed in my interview with SMX Advanced keynote speaker Dave Davies, Head of SEO at Weights & Biases. Other topics covered in our conversation included:

• Why shortcuts with LLMs are hurting your content and how to do better.
• Weird new technical SEO issues.
• Why generative engines are the future of SEO.
• A preview of his upcoming keynote at SMX Advanced 2025 in Boston.

Reminder: SMX Advanced returns June 11-13 in Boston. Get your tickets now!

This transcript has been edited for length and clarity.

Danny Goodwin: Hey everybody, I am joined today by none other than Dave Davies who will be keynoting SMX Advanced on [June] 13th. Friday the 13th, that’s a good omen for you right Dave?

Dave Davies: I kind of hope so.

Danny Goodwin: Dave, if you want to introduce yourself. I know you’ve been around the space a long time. A lot of people know you, but just in case somebody may not know you, who are you and what do you do?

Dave Davies: My name is Dave Davies. I am the head of SEO for Weights and Biases an MLOps, AI ops. So we sort of build the tools that AI systems like ChatGPT are trained on and built on. And just got purchased. So my new employer is now Coreweave, a hyperscaler. So the people that OpenAI and stuff get the GPUs from to train their systems. So yeah that’s who I am and I head up the SEO for those companies.

Danny Goodwin: It’s great to have you here. So, we’ll talk a little bit about your SMX Advanced keynote on June 13th in just a little bit, but since I have you and you’ve been following the industry for so long, I wanted to get your thoughts on the state of SEO as it is today. And I figured I’ll just start by asking, from your perspective, what are you seeing us do right as an industry when it comes to adapting to all the changes that are going on? And what are we also doing wrong right now?

Dave Davies: There’s going to be who am I watching when I’m talking about that, because of course you get people going in both directions on everything, We all remember when okay, you and I are old enough to remember when PageRank was a thing and we were all chasing green bars and stuff and there were people doing it right and there were people doing it wrong. I was a little bit of both because I was a bit of a gray hat at the time. So, depended on which version of Dave you were talking to. But I think in this case a lot of people are getting it right – and this isn’t unique to our sector – this is something that I think every sector is guilty of right now is it’s so easy to shortcut it’s just so easy to shortcut that we are all myself included finding ways to shortcut. How can I produce this content faster how can I this into a model into one of the LLMs to sort of have it proof word things a little bit better.

Dave Davies: And it’s not to say we shouldn’t be using these tools. They’re amazing tools. But the shortcuts are where the big problem is. I’ll list sort of a shortcut that I have taken and then an improved version of said shortcut that I now take much more often. One that you could do is grabbing an article because you’re an editor of a major publication like Search Engine Land. I’m not saying you’re guilty of this. I’m just just noting that this is something that applies to all of us who deal with content all day. And just dropping it into ChatGPT and going, check this for errors. How can I word this better, Davies wrote it, so this thing is way too long, make this thing more succinct. And that’s something that I’ve been guilty of, okay, this is a big thing, right?

Dave Davies: And I’m dealing with some stuff where these are machine learning engineers into their doctorates and stuff like that. I don’t even always necessarily understand everything that I’m looking at or some of the math involved in it and stuff like that. So okay, drop it in there and okay, that’s a shortcut and it doesn’t work. Now, another thing you can use it with projects.  I tend to go sort of offline and use local versions or connect to the API. But rather than doing that, I’ve sort of created these buckets for all of my key authors. And it’s like, okay, here’s a bunch of their work. Now, learn off their work and now take a look through where I’ll sort of go I’ve got scripts running and I’m like, okay, the author is this. Here’s the article. Now let’s make sure we’ve got some consistency in there. I’ve given it a few templates that I like to run under it and sort of like outlines, right?

Dave Davies: Just to make sure that I’m still using a tool, but it’s chopped it from it no longer takes me 5 minutes. It now takes me 20 or 30 minutes, but that’s still better than 3 hours, right? But the finished product is much more to what the original author, their style, their writing, what they would want. It’s catching things where it’s okay to be a little casual. That’s how they write, right? And it’s catching those nuances a lot better. That’s the sort of biggest problem that I’m seeing right now. And again, SEOs aren’t the only ones doing this. Everybody’s doing it. We’ve all gotten crappy emails written by I don’t know. I’m throwing ChatGPT under the bus. There’s just every model is getting used, but a lot of us are using the GPT series because they’re easy to use and the APIs are easy. So it’s not just us, but that’s probably the biggest problem.

Dave Davies: At the same time, it’s a wonderful advantage for us in data collection, especially if you have access to deep research and stuff like that. Now, always got to fact check – even deep research is wrong a whole bunch. But allowing us to sort of collect together a bunch of information or NotebookLLM from Google, right? Grabbing those massive articles that I have and going, I have no idea what I’m about to Dump it in there. Get a podcast version of it. Listen to that for 20 minutes before I try and prove this thing, so there’s all these nice little things that all of a sudden you can do or just go like I don’t actually need to read that, but this is massive white paper, right? That I’m sort of interested in reading.

Dave Davies: I’m not Bill Slawski. I don’t have that in me to get through all of this. And folks, if you don’t know who Bill Slawski is, look him up. He is awesome and one of our sort of like bygone era folks, Danny, I know you know who I’m talking about here. Folks should become familiar with him. He did that sort of stuff, but I’ll just drop it in there. I’m on one break. Listen for 20 minutes and learn. So, we’ve got those tools and I think they’re great for that. But it’s the shortcuts that I think are problem. That was probably the the longest way I could have said exactly what I just needed to say. I probably could have said it in about 30 seconds.

Danny Goodwin: You can throw that into ChatGBT and say, can you shorten this? There’s been a lot of talk with the rise of generative engine optimization, answer engine optimization, etc., etc. We’re kind of moving away from the era of ranking and a lot of people are saying it’s more toward relevance. How do we provide more value to our clients and our stakeholders in 2025 and beyond if we want to continue to provide value through search and SEO.

Dave Davies: Yeah, this is a nightmare. It’s sort of like the worst. I feel sorry for most people. I’m lucky in this one. And by lucky I mean the people that I work for are responsible for building the systems that I’m now being graded on. They understand this is how LLMs work. They understand all of this. So when I’m like, “Okay, I’m going to go way outside the box here and I’ll talk about what people need to be thinking about.” It’s just way outside the box and going, “I need to go there.” And they’ll be like, “Yes, you do need to go there. We understand.” And by saying that, what I mean is we used to go, you, me, everybody, we’ll talk about that little PageRank bar again, And link building. I’m not saying link building is a bygone thing, but just like we used to focus on link building when we thought of external strategies. So things outside of our website here. We would think naturally you’d be thinking of link building stuff like that.

Dave Davies: Now, all of a sudden, if we think for just a second about how LLMs are trained, they’re looking at this massive corpus called the web and they’re training on all of that data. Do they care about links? We can debate whether they care about links or whether Google’s still looking at that. And I’m certainly not suggesting that link building is dead or that links don’t count. But thinking outside of that, we know that there’s some skewing in this, but in general, they look for the next token. They’re building word after word. I’m going to shorthand and call these words that they’re building after that instead of tokens. But, they’re sort of shorthand going, okay, What’s the logical next thing I should see. I’ve now written that. What’s the logical next thing that I should be typing in there?

Dave Davies: So, if the only experience they have with my website and with my brand is what I’ve put on my website, it’s probably not that likely that if somebody’s looking up, hey, what’s the best MLOps tool out there into ChatGPT? If the only place they’ve seen that is on my site, it’s probably not something that’s going to make their probability of the next available word or token. So we need to start expanding beyond that and going okay there’s this massive corpus of data that they’re trained on some of it they’re going to trust more in my sector it might be GitHub in my sector might be media right but everybody’s sector is a little bit different going it’s not about links again not saying that’s dead talking about a different thing though. It’s not about links I actually need to make sure that when somebody’s going where’s the best place to vacation right that I happen to and the tourism bureau of Hawaii, I need to be that. So you need to make sure that in all of the places that would be considered authoritative, you may not be the only thing in there because it would be logical that in an article talking about the best places it would have, in the United States, for example, it would have the Outer Banks, it would have Florida, it would have Disneyland, it would have all of these things. 

Dave Davies: So making sure that you’re included in these massive corpuses and these massive articles to making sure you’re placed in there, link or not, it’s going to then understand okay the next available token is probably going to include Hawaii right like that this is just a logical thing in the token sequence and that in the vector embedding diagrams that we’re seeing that okay Hawaii is often included related to best vacation United States right think things like that. So I think that’s one of the big things that to me is fundamentally changing about the way we need to look at things. It’s not just I need to train large language models now and I need to provide them enough collateral to understand that depending on the question I’m the next logical token right and that the way I want that framed is the next logical thing right what now it can understand sentiment a lot better now it’s actually understanding that just because I’m there I might be bad. Understanding that I need to be in all places related to the corpus that it’s being trained on. And to that I mean just a little finding just this is just like a freebie takeaway and you can pull this if you want. I’ll be including it in the keynote though. But a thing that just hit me it was about two three weeks ago we use prerendering. Our site a big chunk of it it’s in JavaScript so we need to use a rendering engine.

Dave Davies: A lot of people control their bots in different ways. Something I had never thought is we use prerender.io. And I had never thought, hey, go in there because we define each of the bots and we want to let this through. And I’d allowed Google-Extended, I’d allowed in all of these additional Google bots that needed to come in there. And I never went Common Crawl. CCbot. That’s basically Common Crawl is used by all the major language models to train on. And I had never gone, I need to make sure that Common Crawl is let into our prerendered area. I couldn’t find that anywhere else on the web either, I couldn’t find other people talk about this. So that’s just an additional sort of gotcha – 40% plus last I sort of looked now. it’s been months since I looked, but either way a massive amount of the training data used is from Common Crawl.

Dave Davies: You can actually go to index.com crawl and see which of your pages are actually indexed by Common Crawl. And I noticed that’s where I’m like, “Okay, a lot of our pages are, but not the pages inside this area that are prerendered. Dang it.” I’m like, “Okay, I need to go in and change our pre-rendering scripts.” And ll I haven’t had enough time to see how that plays out. If you haven’t seen the new Common Crawl data sets. But it’s just a weird new era with some weird new technical SEO sort of like gotchas built into it.

Danny Goodwin: Absolutely. That’s a good tip. And it was interesting just listening to you, so I’ll tell you a little story of something I just found out the other day about the fun of SEO in this new era. I did a search on I think it was ChatGBT 4o and search just asking who’s the editorial team at Search Engine Land.

Danny Goodwin: It turns out based on not our site but some random site that I had never heard of Matt McGee is the chief editor, whatever they call him, and it’s also working there are… pretty much the rest of us. So it was pretty interesting. Do you think this is going to be a problem? it is odd because it’s like for so long we’ve had to build this authority on our own site, but now it’s almost like we’re flipping that where authority comes from all the other sites and maybe even just some random not I don’t want to call it a spam site, but low-quality sites because they’re just slurping up all this text and god knows how they’re coming to their conclusions in the end.

Dave Davies: I mean, we can get into the sort of ethical gee, what side of history are we all on when it comes to exactly what you’re talking about, right? And now if I start building scripts that start now creating more content based on ChatGPT’s wrong content… Now all of a sudden, we’re proliferating this very problem. and everybody else, I like to think, hey, I’m a kind of intelligent guy. But as we’re starting to get AI Overviews in more places, AI Mode, clearly they’re going to be baking that right into everything. So, we’ll all just be able to have this very chatty sort of experience. My attention span for the hell, I was talking about I used to read white papers. Now, I just go summarize this, make me listen to this in a 20-minute podcast, just get me to the nuts and bolts.

Dave Davies: I like to think it’s because I’m more busy than I used to be, but somehow I managed to stay up till 10 o’clock reading white papers and now I just find something else to play around with. But I think this is part of the problem that we’re going to be hitting just sort of as a species. Is just this okay, what do we do now? As a marketer, you can’t like black hat, white hat, gray hat, as a marketer, you can’t help but see the opportunity built into this.

Dave Davies: You can be concerned about it. There’s certain types of misinformation that we don’t want there. But anytime we’ve moved look at PageRank right little green bar appeared and suddenly Dave went I can charge 600 bucks for that link on this page right that’s like as soon as new technologies come up immediately there is opportunity to go with it I’m not saying to run off and sell links. As a note, I had a site banned and I lost ample revenue from that site because I got greedy and lost so just take my lesson. Don’t follow my lead on doing black hat stuff. It’s not a good idea unless you’re all in on that and you understand the risks of what you’re doing. But with new technologies, the people that adapt the quickest will be the ones to win. So you could look at something like that and go, “Oh, okay.” maybe I should start a company.

Dave Davies: I’m sure Jason Barnard is on all of this stuff and going like okay my job is to actually find where are they slurping up their data from and going this is wrong, right? So okay how do I craft this right start running tests does schema work does it not just all of a sudden start really solidifying yourself as when there is the wrong information when you do have a reputation management problem. With search it was just like okay we’ll just flood the top 10 with everything good. It’s fine. Your reputation management problem is gone. But all of a sudden, now you’ve got these big answers appearing when people are looking up your brand. It’s probably going to be a little harder to sweep that under the table, So, the people that start to figure out, okay, this is how we make sure these are the sources that are more trusted for this type of data. But they’re not perfect. And as you found in that case, it can be like a B or C list site that’s all of a sudden coming up out of nowhere.

Dave Davies: And I’m like, seriously, that’s beating me? We’ve all seen that. We’ve all looked in that and gone, I have this brilliantly optimized piece. It’s on a highprofile site. It’s like, why? What is going on that I’m okay being beaten, but by that. We’ve all had that happens where it’s like, I’m okay as long as I know my competitor is good. But you’ve really got some crap in there or just it’s not crap, but it’s just really not from a source that’s trustable, where it might actually be decent. And maybe that’s something that they’re able to extrapolate a little better than maybe they could in the past in this case a clear example of them not being able to do it very well.

Dave Davies: But I think there are opportunities for people like Jason Barnard who does reputation management and brand management to sort of step in and go, “Okay, we’re figuring out how all of this side works and we’ll fix that for you, Danny. It’s just going to cost you X dollars, right?

Danny Goodwin: Exactly. With change comes opportunity. So I’m curious, for you – what are you most excited about in SEO world right now?

Dave Davies: I’m excited by a lot of stuff. Just naturally. It is just fundamentally different than it used to be. The thing that keeps me up at night. So, I’ll sort of go here’s the thing that keeps me up at night. And it always sounds like a copout when you have to say it. But me and every SEO on Earth are facing the same copout answer right now, which is – how the f do I get attribution anymore? How? Where is my attribution coming from? People are using GPT on their phones. Are using it okay you’re clicking or it’s like coming in from some other site but it’s actually GPT but it’s actually like they’re running their own GPTs in this little app so it’s coming from nowhere but the optimization we’re like how do we make that work?

Dave Davies: Click through rates are dropping AI Overviews are expanding right and I’m in a codebased sector you look at on average I think we’re looking at 20% drops through of AI Overview presence which then it amplified but we’re looking at those sort of ratios but if you look at Stack Overflow when ChatGPT came out they were dropping 50% in their traffic when ChatGPT came out because coders are just like hey this is an easier way to do it. That’s my site. My site isn’t Stack Overflow but my users are those same users right? These are the same I referred to it and I talked to my team about it I’m running up a sand dune here right as fast as I can run. It’s slipping down faster than I can run, and that’s fine. It’s just moving over here and we can’t show it as well. Now, John Mueller, I saw it yesterday. I was on vacation. Just got back yesterday. So, it might have been a couple days before that, but John Mueller was talking about they are finally going to be putting the impressions and the clicks into Search Console. That’ll be a saving grace.

Dave Davies: I don’t know how they’re going to surface it, but at least having that confirmation is I’ll see how often we showed up. Click-through rate is still going to be abysmal. I just know it. But at least we’ll be able to surface that to our employers and go here it is. But then how do you match that sort of attribution down funnel, which is really really important for us to. Be able to go we saw the brand, but if they’re not clicking, how do I go? In a multi-step funnel, it actually led to this sort of action. And again, for me, if it’s generating the code and then the person’s using a Jupyter notebook or a Google Collab or something and then going, “okay. I’m going to sign up for the thing here” because it generated all that code for them. personally, I’m getting no attribution for all my hard work to optimize for ChatGPT or Gemini and whatnot. I’m getting no attribution because it’s actually going to show up as director local host, right?

Dave Davies: When they actually engage with our product and click through to our site to sign up, it’s actually going to be from a completely different source. So that’s I think the biggest problem. At the same time, it is nice to know that all of us are dealing. It’s one of those okay but it sucks for me but at least I know that all of my friend when I’m standing in Boston chatting with Danny and we’re all talking to a bunch of other people, all of us will be going I don’t know what are you doing about this? Does your boss understand what you’re talking about when you’re like, “Yeah, I know we’re down 10%.” But it’s okay, we were supposed to be down 30%. This is actually 20% growth. Getting those sort of estimates together going we’re actually doing better. There’s just less available traffic. Yeah, I know the impressions are staying the same. It’s just a bit of a disaster. Again, I have the benefit of working for a company that has the people that train these systems, right? So, they do understand what I’m talking about in that.

Dave Davies: But it’s still a bitter pill to swallow when you’re I know my traffic’s, stable and that’s actually good, just holding steady is actually growth. It’s a tough pill to swallow. But at the same time, seeing all of these new opportunities, like academically speaking, while we’ll all be battling those things until Google comes up with a better way of framing it and gets us the data in Search Console where we can go, we’re showing up at least and we’re showing up in good positions. We’re not getting clicked, but look at this and I can look at the click-through rates and go, that’s what I’m talking about. We’ll actually have the metrics. Once you have that, that’ll be great.

Dave Davies: But I think academically for a lot of us here especially I tended to lean technical even when I’m doing content SEO it’s like okay what’s the math what are the formats what are the structures of things that tend to work better than others. Even when I’m sort of looking at that stuff, right now mean there’s the Allen Institute put out a great paper on this that I wrote on a while back on sort of like what’s working now they were using GPT3 when they were doing it because it was just when the paper was being done and what they were using. A lot of the techniques they talk about still work. And it’s just like right the formats tend not to … keyword density and keyword stuffing don’t work for example but putting things in a table on a page don’t copy and paste that image put it in a table on a page because Google slurps that stuff up. Cite your sources at the bottom of the page because Google looks at that and goes okay and it’s not like external links count it’s just like we understand now what’s going on here connecting the dots there. Gemini loves quotes in the page, they just love that and statistics, and it makes sense when we think about that and go, “Right, LLMs, as smart as they are and impressive as they are are dumb.” Like 4r-year-old kid dumb. They don’t understand.

Dave Davies: That’s why we have the wrong information showing up for who’s on the editorial team at Search Engine Land because it doesn’t understand what it’s looking at. It’s just reciting back what it’s seen before, right? So for us now we have this great opportunity to go okay I can be on the forefront of sort of trading these things my way and that for me is exciting in what formats work. I have found the formats that work for featured snippets for the same reason they featured snippets work for LLMs too. Because it’s short or lists or depending on the type of feature snippet you’re getting. It’s like this is easily digested.

Dave Davies: It was easily digested and understood when we were targeting featured snippets so they could pull it out and understand they need to put it on a page. It’s easy now that they’re trying to train tokens to go I understand that this heading is this question and then it’s followed by a paragraph that’s 315 words right or characters or whatever and that it fully answers the question. Fantastic. OK, off we go, right? And then there’s one of the tricks that I do, right? That’s like put the heading, it’ll imply a question.  Follow that up with a list or a paragraph with specific character counts. For those folks wondering, I’m playing with them right now. Find about 310 to 320 characters is pretty good for that opening one. But for both featured snippets and training LLMs to understand your answer is the primary answer. So yeah, those are sort of what I’m excited about right now. And that’s just like constant testing.

Dave Davies: Something that I think is a shortcoming I’m looking forward to either developing myself finding is there aren’t really tools built for the kind of testing we’re doing now. Right now I just keep notion docs running where I’m like here’s kind of what I’m doing but there aren’t really very good tools for tracking the type of experiments that we’re trying to run right now. And I get why and I don’t know even how you’d build them but hopefully somebody will figure that out and get a good tool together. wise, I’m just stuck with a notion doc with a big list of experiments that I’m running in an inline database.

Danny Goodwin: That’s more good information. So, I’m curious just SEO as you’re seeing it right now. Do you think as an industry SEO is heading toward growth, decline, or reinvention?

Dave Davies: I would have to lean toward reinvention. I get asked periodically by my team. This just happened about a week right before I was going on vacation. I was like, “Wait, when I get back, I’ll talk to you all about this, but this isn’t a conversation to start when I’m going on vacation.” And I had been sent here’s how we’re ranking, and it was an organic ranking. And I’m like, “Okay, I’ve got three main areas that as an SEO for our company I need to be looking at right now. I won’t get into all of them, but two of them were traditional SEO and generative SEO. And I’m like, some teams are talking to me about how we rank generatively. Some teams are talking to me about how we rank organically. I’m a person. So, I need to know here’s my three things. What are my priorities? Right?

Dave Davies: And that I’m sort of like laying out right now, but I know what I’ll be presenting and I’ve already noted this. I’m like what I’m going to be saying the primary thing we need to focus on is generative engines. Then after that we have our other two priorities, and that I’ll have a big conversation about okay but first can be 34% versus 33 and 33. it could be a close or it could be like no that’s 70% of our energy goes to that and then we divide out the next 30. Each company is going to have to treat that a little bit differently. But I would say if you’re not prioritizing generative engines, you’re probably in for a world of hurt coming up. And some of that has a lot of double dipping. Like I’m not saying you have to abandon your SEO strategies. A lot of the time I’m producing content that’s built to rank right now, but also to train LLMs, right? I’ll like format them to train LLMs.

Dave Davies: So I’m not saying abandon everything, but I would prioritize generative engines and in some cases that involves looking out to systems that are other APIs, other offerings. You could look at travel. I’ve talked about travel for example, and there’s hotel ads, we’ve all sort of seen hotel ads. They’re a pain in the butt. Anybody who runs like a vacation rental company or whatnot is going to be like, “Yeah, they’re annoying. You have to go through one of their partners. You can’t set it up yourself, Can you tell who’s had to deal with this in the past? And I happened to look it up a couple months ago. I’m like, it is still that way. They’re not letting just somebody who’s managing 300 properties just upload those, and manage your own listings. You have to go through partners and now you just like it’s going to cost a lot.

Dave Davies: Bing doesn’t do that. Bing will let you actually do that. And again, I just happen to have looked this Bing, for the most part, powers ChatGPT. ChatGPT just started, I’m sure you remember it because this was covered on Search Engine Land. That’s where I learned about this. just started showing product placements. Now, it’s not actually pulling directly from Bing right now, but ChatGPT and OpenAI are losing money, so we know they need to put ads in. that’s just they can’t support a free product, I pay them a couple hundred bucks a month because I use that level and then I have to pay their API fees as well. But that’s worth it for me.  But for most people, it’s not. My dad, that would be just useless, why would he possibly do that? Most people use the free one. Maybe the $20 one, but most people are using the free one. So, they need to monetize that. How are you going to monetize that? Well, of course, you’re going to start showing ads. That’s just the way that is done. So already I would advise people, start looking at Bing now.

Dave Davies: Even when you’re thinking SEO, start really looking at Bing because that’s where you’re getting a lot of the influence that’s going into ChatGBT and a lot of people aren’t using Gemini, right? They are still using Google and so Gemini is powering inside of Google their AI Overviews and stuff. So they are using Gemini, they just don’t know it. But ChatGBT has a lot of weight.  So, I would suggest looking at and I told people you need to get into Bing Hotel if you are a property manager in that get over there because they’re going to have to do this. And it’s a low lift thing. So, do it. Maybe Dave’s wrong, but if Dave’s right, you have a big lead over your competitors. If I’m wrong, it didn’t take you very much time. It took you a couple hours. So, get that done. And if I’m right, it’s a big win. If I’m wrong, it cost your developer a couple hours, this isn’t a big deal. And you could probably use that for Google once they get their crap together, open up Hotel Ads to just everyday people.

Dave Davies: So that’s sort of the sort of like way I’m sort of thinking about that. So generative engines first, start thinking about what’s feeding the generative engines and start thinking about the opportunities there is sort of my main thing. And then I’m treating sort of more traditional organic as almost the means to the end with a pleasant byproduct called traffic.

Danny Goodwin: At SMX Advanced in Boston, June 11th to 13th, you are kicking off Friday the 13th with your keynote, which is Agents of change: SEO in the era of agentic AI. For people who are watching and maybe thinking about coming, or are coming, can you give us a little preview what you’re going to be focusing on talking about in that session

Dave Davies: Yeah, agents are really really exciting to me. And there’s actually a little part which I’m going to be including in it is from a deck that I did back in 2018 and I was not expecting them to take this long to get here but we’re almost there. We’re still not there but we’re almost there. And the thing that’s really really exciting about agentic systems to me and so I’ll be defining that for our audience here’s takes a few minutes but sort of here are what agents are but the cliff notes for the folks listening right now is it’s systems built I’m going to placeholder some wrong terminology but it’s like I use words instead of tokens so we’re just going to the placeholder sort of pool use right where you build a bunch of different agents into a system each one with its own capabilities baked in right. I can check this I. I can run these search queries. I can do X, Y, or Z. You build in agents.

Dave Davies: And then you build an overall system. So sort of I liken it to an ant colony. There’s a queen organizing here are your goals. Now go out and do that. And then they communicate amongst each other. They come up with the best way to support those goals and then they report back. That’s sort of the way I sort of like placeholder thinking about it. It’s the world we’re heading into Project Astra from Google, which they started showing guess about a year and a half ago at Google I/O is sort of almost more like a chat. It understands visibly what’s going on you can show it pictures including through cameras, right? They’re already starting to work with Samsung on some glasses. Google is, right. So that’s going to be really really cool. We’re sort of crossing over into that. Now all of a sudden I’m not needing to make decisions. I’m communicating.

Dave Davies: And this is what we’ll be talking about. And it gets to the caveat and the really interesting part for us as marketers in just a second. So in this world where I can just walk around and I’ve got glasses and I’m like, take me here.” just actually communicating with these glasses fully agentic systems just going like, “Okay, I got a wedding coming up, or it just knows in my calendar, “You got a wedding coming up?” “Do you have shoes?” and it knows my size It knows where I shop. It knows my styles. Bam. It can just get this done sight unseen to me, in a lot of cases. In some cases, I’d want to see those shoes, pop them up on my TV, pop them up in my glasses, let me just see that everything about my style and everything. So, in some cases I’ll be making just completely utilitarian decisions and I no longer need to make those decisions, right?

Dave Davies: I just offload to an agentic system that just understands what I need and that I don’t have anything coming up and I probably don’t need okay sorry I’m having to do some quick I don’t need a gallon of milk so I’m from Canada so I’ve got the metric system I’m like four liters nope that might not go across quite right right but I’m going off to Boston in a few days I’m out of milk but I don’t need right a gallon of milk I just need a few cups right. So ordering things in small understanding these and then just making those utilitarian decisions for me with me just offloading that I don’t need to do it. Now as marketers heading into that and we are the most recent Claude model for Anthropic is just built on this it has our long attention span in paying attention to what its task is what it’s doing what the other pieces of the agents are doing in there. So as we start to approach that who am I marketing for and that’s going to be one of the questions that I’m talking about and answering a little bit in this keynote. And some of it is like but wait there’s more and we got to stay tuned. On this one but if I’m now offloading a lot of my decision-making onto an agentic system are you even optimizing your content for me and is it even content anymore right where we’re in there. And you could look at some recent protocols that have come out the model context protocol also out from Anthropic, A2A protocol put out by which is agent to agent protocol put out by Google which I’ll be covering a little bit of those and showing what the agent cards look like. The MCP protocol is more here’s how agents communicate here’s how we pass ones and zeros and how we pass data. The agent to agent protocol is more exciting to me right from a marketing standpoint and I think of that it’s basically cards that you put for your agents it’s sort of like a definition card.

Dave Davies: You would build an agent, give it this definition card that goes here’s my capabilities, here’s the formats I support, here’s what I do, here’s a description, right? It’s just like a summary of here’s everything this agent can do and then you just drop it into this agentic system and the agents now communicate with each other to go I need to do these tasks. But they just understand how to communicate with each other to go, okay, here’s all the things that I can do. So now I look at marketplaces and go, we don’t even like I can just drop my agent into somebody else’s thing. So now maybe I build agents as a marketer and go, “I can actually just drop that in like we would with Looker Studio connectors or like we do with WordPress plugins and go now I can actually just drop my agent into Agentic Systems. Well ow do I now market that agent system? How do I monetize that agent? it just opens up a whole door and goes and I might actually be dealing with people that I actually just need to optimize that Agentic system.

Dave Davies: I’m not actually optimizing anything else because as soon as the person’s using that, all of those utilitarian decisions are being made for them. It might even be booking their trip to Hawaii. Because that is the top place and Dave likes it when it’s 100 degrees, right? So that’s off where we’re going and we’re just going to book that. Can you tell who’s planning a trip to Hawaii next year? Yeah, I think that’s what’s really really exciting and that we might not actually be dealing with content in the traditional way in a lot of spaces and that’s what the agentic world is coming up and that’s what a lot of the keynote I won’t be as verbose in describing it because it will be more rehearsed than just a question in a chat here with a friend.

Dave Davies: That’s I think part of what’s really really exciting and what is just going to fundamentally change with agents and yeah it’s going to be an exciting few years ahead for us and very tumultuous. But the winners are going to win big if you’re paying attention.

Danny Goodwin: Dave, thank you so much for joining us to talk about all of this stuff. Looking forward to watching you keynote as SMX Advanced and catching up with you in Boston. So, thanks for taking the time and, thanks to everyone for watching. Thanks everybody. Bye.