216: The Machine Layer - Building Trust in the Age of AI Search with Duane Forrester

Duane Forrester, 30-year search veteran who co-launched Schema.org and built Bing Webmaster Tools, explains why AI systems prioritize trust above all else. We discuss machine comfort bias, chunk-level content optimization, why SEO is now a multidisciplinary role, and how to prepare for a world where LLMs decide who gets cited.

About the Guest

Duane Forrester is the author of "The Machine Layer" and search industry pioneer

• 30 years in search and digital strategy
• Senior Product Manager at Microsoft - built Bing Webmaster Tools
• Co-launched Schema.org structured data standard
• Leadership roles at Bruce Clay Inc. and Yext
• Founder of Unbound Answers, creator of CitationIQ

Chapters

• 00:00 - Intro
• 01:01 - The biggest shift SEO has ever seen
• 03:30 - Machine Comfort Bias: the 5 layers of trust
• 09:19 - Chunking: writing for AI and humans
• 16:01 - Making content citation-ready
• 18:35 - Schema.org: the trust infrastructure
• 25:46 - Ironman vs Superman: AI as amplifier, not savior
• 32:28 - EEAT, Universal Verifiers, and why trust is everything
• 42:59 - Latent Choice Signals: the invisible metrics
• 52:35 - The Machine Layer book
• 57:53 - Emerging roles in AI discoverability
• 01:01:16 - Where to find Duane

Key Takeaways

1. Trust is the new algorithm - LLMs need multiple dimensions of verification before citing you. If you can provide everything they need without them having to guess, they'll lean into that "machine comfort bias"
2. Chunking matters, but not how you think - Don't reformat your entire page into 300-word blocks. Instead, put key facts, figures, and bullet points at the top. LLMs get "lost in the middle" of long-form content
3. Be the canonical source - Your goal isn't rankings, it's being seen as THE source of knowledge on your topic. If you haven't expanded the LLM's training data with net new information, you won't be cited
4. SEO is now multidisciplinary - Technical SEOs must understand branding, conversion, engagement, PR, and UX. Silos are killing companies in the AI discovery layer
5. AI is Ironman, not Superman - These systems amplify your skills but require you to drive them. Hope is not a strategy. Always ask self-referencing questions to verify outputs
6. LLMs want to save money - They won't waste tokens looking elsewhere if you provide everything they need. Consistency and trust reduce their computational costs

Resources Mentioned

Duane's Work

• Book: The Machine Layer - Available on Amazon (Kindle & Paperback)
• Website: duaneforrester.com - Free frameworks from the book available
• Substack: duaneforresterdecodes.substack.com
• LinkedIn: linkedin.com/in/dforrester

Connect with No Hacks

• Website: https://nohackspod.com
• Newsletter: Subscribe for weekly episodes

No Hacks is a podcast about web performance, technical SEO, and the agentic web. Hosted by Slobodan "Sani" Manic.

Transcript

Duane Forrester
===

[00:00:00] Duane: these systems aren't interested in nostalgia or your story, or they are interested in facts information.

[00:00:08] Duane: Can I quote you in a way that is trustworthy enough

[00:00:12] Sani: That is Duane Forrester. He spent 30 years in search built bing webmaster tools at Microsoft and co launched schema.org, the structured data standard that runs the web. He's not just predicting the future of search, he helped build the foundation. In this conversation, we get into the machine comfort bias, why LLMs evaluate content in chunks rather than pages, and what it actually means to be citation ready for ai.

[00:00:38] Sani: Duane's new book is called The Machine Layer, and after this conversation you'll understand exactly why trust has become the new algorithm. This is no hacks. subscribe@nohackspod.com slash subscribe. Here's a story.

[00:00:51] 

[00:01:01] Sani: Duane, welcome to No Hacks. The shift we're going through, I think it's the biggest shift SEO has seen since its inception. You wrote that SEO is no longer a single discipline, and for website owners preparing for the agentic AI and for the agentic web, what are the mental model shifts they need to do?

[00:01:20] Duane: This is look, so your point, right? This is the biggest shift that we've experienced essentially since the industry started. You could say mobile was a big deal, you could say personalization was a big deal and they were at the time in their way. This however, is truly disruptive and.

[00:01:38] Duane: The biggest, I think like one of the biggest issues that we're facing is we spend 20 plus years as an industry. Whether you've been in the industry a year or 20 years, doesn't matter, right? You are guided by the history of the industry suggesting to you that. It's keyword research, it's gap analysis.

[00:01:58] Duane: It is technical, SEO. It is deploy schema. It is specific things, and you can checklist your way as an SEO toward modest success, or in some cases great success. That doesn't work. As much moving forward, it's still important. You still have to do all that, but if you show up to work every day thinking you are doing a good job by going down that checklist and getting those things done, you're in for a rude awakening.

[00:02:26] Duane: This is a completely different world. You have to understand branding. You have to understand conversion, you have to understand engagement. And too many SEOs still think this is a game of throw it over the fence. I did my part, I got traffic to the website. UX is responsible for fill in the blank.

[00:02:46] Duane: Branding is responsible for the messaging that we use. I don't talk to paid because Paid does their own campaigns. All of that is going to destroy companies in this new discovery layer. There is no question these systems aren't interested in nostalgia or your story, or they are interested in facts information.

[00:03:08] Duane: Can I quote you in a way that is trustworthy enough that I, the platform, don't have a problem and my user. ChatGPT Gemini whoever that user is satisfied with the answer. And if you can provide the source of power that gives that to the LLM, you'll get cited if not

[00:03:30] Sani: Once. One phrase you have is a term from your substack is machine comfort bias. So this is what we're talking about here. The machine should be comfortable trusting you and recommending you to, to its users.

[00:03:41] Duane: Yes. Yes, exactly.

[00:03:43] Sani: what are the five layers of it? Let's talk about the layers of that machine comfort bias, training data to freedom, authority, reinforcement format, formatting preferences semantic clustering and risk aversion.

[00:03:53] Sani: And look, this is nothing about, this is new. That's the crazy

[00:03:59] Duane: but technically no. If you've been doing SEO for a while, let's put a timeframe on that. Let's say you're four or five years in or longer, you should have easily come across most of these topics, okay? Oh, I understand that. You haven't come across embeddings and vector alignments yet. That's new. But this idea of defining trust signals, and that's essentially what this bucket is doing.

[00:04:27] Duane: Yeah, we've got EEAT as a framework, and so you've got exposure to it, but if you're like most SEOs, you are thinking of EEAT as a way of getting ranked higher in Google, and that's your mental model. You are not thinking about it in terms of what if I had a conversation with someone and I gave them a piece of information and they turned around to have a conversation with someone else.

[00:04:54] Duane: And trusted my words in their conversation, would I be setting them up for success or failure? That is not a mental model that comes from following EEAT yet. It really should. And if you dig deep enough in exposure or in expertise, in authority, in trust, when you dig deep into these things. You truly get to this level that the LLMs are looking for.

[00:05:24] Duane: They're looking for multiple dimensions of. I'll use the phrase vector, meaning direction of support for your statement, whether that's the measurements on an object, the efficacy of a product, a particular statistic, anything they need to know that it's trustworthy and you can build that. So these systems get into this machine.

[00:05:51] Duane: Comfort bias. Because if your website consistently deploys structured data the right way, you consistently mark up authors, you are consistent in everything you do over six months a year. Historically, if they go and they reference it and it's the same all the time and it looks trustworthy and it's proven trustworthy, the system will naturally.

[00:06:17] Duane: Through a look. I, if you're watching this or listening to this show, you've used chat pt, you use Claude. And as we talked about in prep, I sincerely hope that the watchers of this have, are actually, are deep into Claude code. And you understand the limits of these things, where you will give it a direction and it will come up back with something totally surprising to you that you never thought could be interpreted from what you said.

[00:06:39] Duane: And you're like no. And then you need to scope it down. If you can provide that platform with everything it needs so that it doesn't have to go and make a decision, it wants to lean into that bias because that's fewer resources spent. Go ahead and give Claude a 3000 line Python document. And then a couple of parts, like a couple of conversation points later, a couple of, response Q and ass back and forth.

[00:07:07] Duane: Ask it something about that content that's in that document and watch what it does. It's first a response to you if you don't tell it otherwise is to guess, because that takes fewer tokens from it. To give you something and because it has the corpus of the project, it's probably pretty close in.

[00:07:28] Duane: Its in its guess. But when you're looking for a particular line in the code and a particular character on that line, there is no guessing. It's a finite set of coordinates and the system has the information you can provide it. So this is this machine comfort bias in action.

[00:07:47] Sani: and it's going to change literally everything about how web websites and webpages are being consumed because it's no longer about the one thing you say, you cannot, the user lands, you tell them a story that story needs to be believable for the machine

[00:08:03] Duane: look, I hate to say it this way. So look, we've gone from a world where you could do SEO and it applied everywhere. And the content you created was generally applicable everywhere to you have to create content per platform. Because those weights and temperatures that I mentioned earlier in passing, that's the new algorithm and its pieces inside the LLM, and each one of the LLMs has different weights and different temperatures for everything it looks at.

[00:08:31] Duane: So truth and proof might matter more to one LLM than another. That's not to say that they want, untruths and unproven, but they're more likely to let the LLM fill in the blank. We can right now. These platforms can completely eliminate hallucinations. If they do that, it will be like talking to an idiot.

[00:08:53] Duane: When you talk to your LLM, there won't be any go out and creatively bring you information that is new and insightful to you because it will be scoped so tightly to a very particular point. Anything off that line gets a bland response and uninteresting response,

[00:09:14] Sani: the appeal is gone for the user, the appeal is gone and they're not as popular as they are.

[00:09:19] Sani: About content. Let's talk about preparing content, specifically web content for AI agents. Let's talk about chunking, chunk level optimization. Something you talked, you wrote about in on your substack and in the book as well.

[00:09:30] Sani: So how is that different? Because we built the internet for the last 30 years, almost like a newspaper. There's a page, there's another page, here's another page. You go from one page to another. How is this different and what kind of change is necessary for people owning and controlling websites?

[00:09:48] Duane: I think it's important that we recognize the concept of chunking is a machine learning construct. It's not a made up word. It's not something that, SEO's invented and whatnot. So if anybody's out there laughing at it or whatever, you need to get over it. It's an actual description. I've had people challenge me on that.

[00:10:04] Duane: It just shows me that they don't, in fact, understand. Much about machine learning 'cause it's a pretty basic concept in that area. I will also say that you will have heard recently some people saying that you shouldn't go and chunk your content. Some pretty important voices. I disagree with them.

[00:10:22] Duane: I think that is bad advice. I've written about this on my substack as to why I think this, so I'm not gonna put a lot of time into this. I will say that chunking is important to an LLM platform, there's no debating that it's a real thing and it's important, the level of importance. Obviously so many things, it varies, but it's safe to say that they all put value in it because it is how these systems take information.

[00:10:45] Duane: And if you're not familiar with a chunk or a block as it's sometimes called. I would like, I would, it is somewhere between 103 hundred words roughly. It. The number varies, it's flexible. But a chunk might cut off in the middle of a sentence. It might be only two words out of a sentence.

[00:11:01] Duane: It might be half a paragraph. Oh, paragraph. Really what the platform is trying to do is just trying to capture an idea in its totality, in the chunk, okay. Duane Forrester is an author and an SEO and he camps in the desert. That would be a chunk and largely captures who Duane Forster is. Obviously not everything, but if you put that in, that could be chunked.

[00:11:27] Duane: And the answer to who is Duane Forster, that could be brought back as a reliable answer. It covers the big boxes. So chunking is important. Now the challenge with this is if you just wrote in chunks. You can like just visualize your average webpage. It's written in prose, we have paragraphs and breaks and sections and whatnot.

[00:11:48] Duane: And if you took all of that, and this is where I do agree with the direction that we're being taught well, that that the major search providers have, or a major search provider has mentioned. I am not a fan of saying, Hey, you should take your prose and you should put that into 300 word paragraphs and space them out down the page, okay?

[00:12:08] Duane: For a human being, you come in, you look at that, and you go, what the heck? This makes no sense for the system, and LLM makes perfect sense for a traditional search system. It's actually a little more confusing. Because now you've gone from the whole corpus of a page being paragraphs and making sense from a human perspective to that same system that's looking for the human perspective is looking at this going, are you a really crappy writer?

[00:12:41] Duane: And, it's like I understand this difference that we see here. However, it's really important that there's a couple of things you have to do. First off I wrote about this, I wrote about this I think last week. Or technically this week it's this week as we are recording it's in this week's substack where there's a white paper in there that talks about lost in the middle, and it's a study on how LLMs get lost in long form content.

[00:13:06] Duane: And content that you put in the middle is less reliably extracted and understood, so it does better at the beginning and the end. The practical takeaway is if you're going to make a point. In whatever page you are writing, don't bury it, put it right at the top. Facts, figures, bullet point lists, all of those things right at the top.

[00:13:27] Duane: I like it. And this is how I tend to approach things. Although I will say this, kids do as I say, not as I do. If you look at my substack, it's all long form, and I don't do this, but if I'm producing content or I'm guiding a client. Look, do A-T-L-T-R at the TLDR at the top of the page, right too long, didn't read.

[00:13:46] Duane: Give me that summary and put the most important points of your entire page. Put them up there in a clean, usable format for humans and it will match a clean, usable format for the machines. And then you've checked both boxes off, right? Like there you go. But extremely important that your content creators understand that they now have to create content for the AI machines and the humans.

[00:14:15] Duane: If you've got a writer who's looking, you going, I will only have a write for humans, you've got probably your next employee who's gonna be leaving the company because that is not the

[00:14:26] Sani: But you can do both. You can do both at the same time and no one gets harmed. Everyone

[00:14:30] Duane: Absolutely. And if you were, and if you were gonna argue with me that you can't, that you have to pick one. I'm gonna allow I'm just gonna agree to disagree and move on because when I hear those arguments, I feel like I'm looking at the next Sears or Toys R Us. You missed the internet. You missed the point. Don't be a slave to what you knew because you refuse to change. And that's how that translates to me, because I know you can involve chunking and regular prose for humans.

[00:15:04] Duane: You can intersperse these. Chunking isn't about, take 300 words, put them together, make it a sentence, put punctuation on it, put it in place. It's not about that. It's about capturing the concept and taking it down to no fluff. Just proof and displaying that, and if you take an entire paragraph and collapse that into three bullet points, you've actually done a really good job for humans.

[00:15:28] Duane: Traditional search and ai. The trick, of course, is the whole page shouldn't be that way. The whole page has to be consumable by a human, but it's not oh, I have a bullet point list. That's enough. No, you need to bullet point the content that should be bullet pointed. You need to, summarize what should be summarized and so on.

[00:15:51] Duane: It's a lot more nuance and a lot more complexity, which is why I have a problem when people say, don't chunk. There's so much more to it than that.

[00:16:01] Sani: We are in agreement. I have not heard those other voice, but I agree with that a hundred percent. So this is about writing and how you structure your content. How do you make your content citation ready for AI systems?

[00:16:12] Duane: Okay, so now we're gonna head back to that machine comfort bias side. We're gonna talk about trust, right? And we're talking about everything that you think of from EEAT. We're talking about everything. If you're not familiar with local marketing, it would be really good for every SEO to get fluent in local marketing because local marketing has a lot of citations that matter, like a lot of local chamber of commerce and things like this.

[00:16:39] Duane: Okay. They're not deal, like they're deal breakers, they're not like silver bullets, right? But they do a better job of collecting citations as a group of marketers for their clients than I think traditional SEOs do. SEOs typically look at these things and say, oh, I got mentioned in an article. I got a link in an article.

[00:16:59] Duane: Those are citations. In order to be a citation for the LLM, you end up being linked in there. That's the definition of the citation is being linked, and in order to get to that level, you pretty much have to be the source on the topic. Which means you really have to go deep. You have to provide so much information on this that when the LLM finds it and takes it in, it says, I'm not gonna waste tokens looking anywhere else because I have everything I need on this topic.

[00:17:28] Duane: Keep in mind that LLM has training data so it can look at its training data and if you have not expanded, its trading data with net new information on a topic. You're probably not going to be the canonical source on a topic, and that should be your goal. Your goal should be to be seen as the canonical for, fill in the blank, whatever your question is, whatever your keyword is, whatever your vertical is, that's what you should be chasing.

[00:17:56] Duane: Not rankings, but that you are the source of knowledge. 

[00:18:02] Sani: Alderson, our mutual friend recently. This podcast is called All of that Zombie Web 'cause it, it's information that the LLMs and the systems do not need because they already know it. If you write another recipe for something and Google knows how to.

[00:18:16] Duane: Yep.

[00:18:17] Sani: Create that recipe there. There's no need for an account.

[00:18:19] Sani: If you have a dentist website that, that talks about whitening and there are a million dentist websites that, that

[00:18:24] Duane: exactly.

[00:18:25] Sani: you are not going to get into those systems and there's no chance for you to do that. Local search is what you should be focusing on, like you said, in that case. So let's talk about schema.

[00:18:35] Sani: 'cause you've done key work on schema schema.org. You co-led that initiative. What's the right phrase there?

[00:18:41] Duane: I was one of the people who

[00:18:43] Sani: One of the people who launched it. And thank you. I'm obsessed. I love schema. It just it's fun. Writing schema is fun. I'm sorry, may sound nerdy, but writing schema and getting it perfect is absolutely fun.

[00:18:54] Duane: I love hearing that, but the one thing that I have to say every time somebody says this is I'm like, where were you when we launched? Because it was a long, slow burn for Schema to really take off and find its

[00:19:07] Sani: remember that. Yep.

[00:19:08] Duane: Now, of course, everyone understands the value, right? And the last I would say five years has been, schema love, right?

[00:19:15] Duane: It's been great. But it's always been about identifying entities and allowing, a path for trust. Okay. And this is like a common theme, right? Like we're gonna talk about schema here, but we've touched on trust three, four times now, and schema is another form of trust.

[00:19:33] Duane: And so like, why people ski, they skip it. Why? People go short on it. They don't use everything they can. Most people don't realize that you can submit ideas to schema for markup that doesn't currently exist. Now it has to be non self-serving, so it has to benefit more than just your business.

[00:19:54] Duane: However, they're completely open to this because you are the expert of your space. So you should be leading that initiative and yet, I go around all the time talking to people who just don't deploy it

[00:20:06] Sani: I would say most people do not know that they can do something like that. Deploying all of the, and I recently really upgraded Schema on No Hack's website. I went crazy and I was so impressed by how many fields you can have for a podcast. So guests duration, it is just, it's beautiful mapping natural content to structured data and structured content is just a magical thing to do.

[00:20:27] Sani: It's like playing with Legos almost, where you create something completely different out of it, and it's fun. And I thank you and everybody else for that initiative. Who was there from the beginning. What was the reason you created back then? I'm gonna guess it was 10, 15 years ago.

[00:20:42] Duane: Yeah, it was trust, right? So if you rewind this we'll say I can't remember the date. I wanna say it was like. 2014. 2015, somewhere around there. Someone can keep me honest on that, but I, I remember more about walking on stage and making the announcement that I do the day I did it right.

[00:21:03] Duane: I do know it was at SMX advanced in Seattle, so there was that, but but it's about trust. It's about a conversation with the content owner and think back to that time, right? Like spam is still a big deal. It's a big deal today, but it's very well known. It's hard to be novel with spamming in traditional search.

[00:21:23] Duane: And at the time we were still seeing novel things. So if you could initiate a conversation through a system with the content owner, where the content owner basically took an action that says. I give you this. I am proving to you that this is mine. I touch this, and all the other pieces upstream from that are in place.

[00:21:46] Duane: You've validated your website, you've proven ownership and so on. That makes your content a whole lot easier for the system to say, I trust it. What that looks like inside a search engine is fewer cycles being spun. Which literally means a lighter load on servers, less electricity, less burnout, less hard cost, less crawling.

[00:22:11] Duane: All of these things are hard costs. So you build this relationship, and now here we are, fast forward all these years. That relationship plays pay dividends for a lot across the internet. Okay. So even the people who are complaining about it or even just doing the basic schema deployment thank you.

[00:22:32] Duane: It's you are perpetuating the importance of it, and it does matter. Now. I'm sure in the future there will be more things like we've seen, we have robot, TXT, we've got sitemap xml. Lately we've seen LLMs txt, and like all of these things have a life cycle. Schema's all about trust.

[00:22:50] Duane: It's all about identity. And you can participate and you benefits or you can not, and

[00:22:57] Sani: How bad or potentially catastrophic is it for a website to fake it? Schema versus its original content on the page?

[00:23:04] Duane: I would say it's saying something weird in front of your friends. They look at you sideways and they're

[00:23:09] Sani: dude, we know you.

[00:23:10] Duane: that doesn't make sense. But it's not like you're getting kicked off the island immediately. Like it's not that.

[00:23:16] Sani: But trust signals probably get influenced by that.

[00:23:19] Duane: yeah, like it's super easy for a search engine to look at that and go, oh, that's bs.

[00:23:24] Duane: And then immediately tweak the trust signal on your domain, like instantly. And then everything for your domain moving forward is assigned a different level of trust. And so everything you do is filtered through that. You'll never know, a human will never know at the search engine 'cause it's all automated.

[00:23:43] Duane: And can you get back? You sure you can because somewhere in a line of code, some human said X number of days after. Test again, reset if and I've No, it might be like a thousand they put in. There you go. And that was it. It's a thousand days, right? And then in a thousand days, that'll reset and it'll test you again.

[00:24:03] Duane: And if you're still lying, then you know, you go back to where you were. And if you're not, then you come back up again. But think about a thousand days an hour, real lives, right? It's a three, three and a quarter years. You're looking at,

[00:24:13] Sani: time.

[00:24:14] Sani: What I love about machines and systems is they're honest there. There's no feelings, there's no sentiment. They're fair. The rules are the same for most, not for everyone. For most people, they're mostly honest. Okay. I corrected myself.

[00:24:28] Duane: we're learning to, I think what we're doing though is we're learning to allow the machines some flexibility on that. I agree with this really came from if you go into my, pine cone instance, right? It's binary. It either is or is not. It's very Yoda about everything, but if I start layering things on and I start getting into temperatures and my weighting, and I start going I want a little more personality and I'll take a little more hallucination, but only 2% more like that, 2% is now no longer binary, and I think it's us that are allowing that to happen. by and large, yeah, I generally trust the outputs. Now I watch my outputs like a hawk. Don't get me wrong. And I question everything, but I've learned to do that. I think the biggest takeaway that I've had with every one of these systems, and I am an avid user of Cloud Code and chat pt. You have to have or develop exceptional PM skills.

[00:25:33] Duane: 'cause if you are not a good pm you will not see problems being introduced because you're not looking at the overall of a project.

[00:25:46] Sani: You also talked about using AI and seeing AI as more of as an Ironman versus Superman. Can you explain on that? I like the concept a lot. Can you talk about that a little bit?

[00:25:57] Duane: Okay. I'm gonna outline a, I'm working on a project. I'm sitting in my office listening to, a great playlist and I'm noticing that over the last two hours my level of anger has gotten increasingly higher and my desire to punch my laptop and my monitor have skyrocketed. And it's because. The platform that I'm working in is just bringing me back crap. All of a sudden, it felt like I fell off a cliff, couldn't answer questions correctly, couldn't remember. Syntax, couldn't remember a, I literally, in the last conversation, gave it a file, told it to reference it. It referenced an old file instead of the one I gave.

[00:26:32] Duane: It gave me wrong

[00:26:33] Duane: information. I knew that, again, talking about PM skills, like you actually have to see the totality of a project, not a very myopic view of it. And I went off, right? Like I started a different thread and I was pounding on it, and it was Claude that I was pounding on, and I'm like, how should I pay for you.

[00:26:55] Duane: Why should I trust you? And we went into this long philosophical debate and about 10 minutes into the conversation it occurred to me, I am looking at this wrong. So like many humans, I'm looking at these systems hoping they can save me time that they understand and will fill in a blank for me. And increasingly it's, I'm being beaten over the head with this, so I should get the message.

[00:27:21] Duane: That's not how these systems operate, right? They literally are just filling in the next logical blank. So if I say something and assume the system will fill in that blank, I also have to assume it might fill in the blank in a way I don't want. But as a human, I'm programmed to be like, oh let's think the best here.

[00:27:43] Duane: And so I'm hopeful for an outcome. Hope is not a strategy. And so like you see this breakdown, right? And what occurred to me was thinking of. Claude Code in that moment as Superman, I was seeing it as a standalone hero that would do things for me on my behalf to my betterment thinking on its own. And that's not what it is. It's an Ironman suit. Oh, sure. It amplifies my skills. It makes me faster, better. It gives me skills I don't have, but I have to drive it. If I don't give it any input, it sits still. And yes, if we continue down the Marvel path, you know we are gonna talk about if there was an incoming threat, the suit would get out of the way and take you with it and it would protect you. We're not quite there. We're also nowhere near Superman with these systems. They are not just going to go off and fully complete everything. I don't care what anybody thinks about agents or what example you find or who the early movers are in these spaces, even Claude Code. As powerful as it is, and as great as it is, it has very clearly defined edges of the envelope and when you get there. You can't hope your way around that edge, right? Like you have to live within it. And I end up with this Superman versus Ironman. And so many people, especially consumers today, consumers think that AI is superman. I ask it a question, it comes back with an answer. It's a genius. Oh my God, this is so many, it makes me smarter. That's the way they think about it.

[00:29:25] Sani: Is that marketing? Is that just marketing from the companies behind the LLMs?

[00:29:29] Duane: Nope. Not at all. Not at all. They have so little marketing

[00:29:33] Duane: that has hit mainstream that the vast majority look chat, GPT in a couple of months is gonna be hitting a billion daily active

[00:29:40] Duane: users. They did not, OpenAI did not do any level of advertising

[00:29:44] Duane: that would've exposed 'em to that many people. It's not possible, but. If I kept telling you really interesting things every day and you started thinking that I suddenly was becoming a more educated person, and then over coffee, I told you I use Chacha, bt, and this is what I do. You would think chat GPT was really powerful, and

[00:30:04] Duane: then you'd start using it for your thing and then there you go.

[00:30:07] Duane: The problem with this is that you have this false idea that it's a superman because you're using it in a very specific space, doing a very specific thing, and it impressed you. Pretty easy to do when you don't have an expectation. But if it were compared to an actual human having an off the cuff conversation.

[00:30:27] Duane: No. Vastly different, right? So Ironman, that's where we're at.

[00:30:31] Sani: I wrote about this on my substack. It's really good at producing something that looks good, that looks like it's good.

[00:30:37] Duane: I agree. 

[00:30:38] Sani: don't know what good actually is in whatever field you are 

[00:30:41] Duane: Yeah, you'll think it's awesome. 

[00:30:42] Sani: it's awesome and

[00:30:43] Sani: maybe that's why the Superman. 

[00:30:45] Duane: So here's my, my my tactic that I will give to everyone, right? Make sure you ask self-referencing questions. Why is this the best answer? Prove to me this is the best answer. Show me why this is the best answer. Use live web search.

[00:31:02] Duane: Collect three examples that support this as the best answer. Take those extra steps, okay? I don't really care if you're a consumer and you do this, I care a great deal. If you're an SEO, you should be doing this because this is how you're going to self-check. That answer was actually a good one because as confident as the system is in giving you an answer that you think is good and that it says is good, it will instantaneously say to you when it comes back, I can't find anything that proves that you shouldn't use that. Like it just no shame at all because it's not a

[00:31:38] Duane: person. 

[00:31:38] Sani: you totally caught me on that. It is just, it's going to go into museums and history books. The chat GPT opener. Oh, fair point. You totally caught me on that. It, they even pared it that in South Park. I think. Let's go back to trust. Let's go back to consistency and trust, and my favorite analogy on your substack, the flickering Marty McFly photo from back to the future.

[00:32:00] Duane: I have never seen this. I don't know what this is.

[00:32:03] Sani: No way.

[00:32:05] Duane: Kid you not mind you, I've also spent the last six months with my nose

[00:32:09] Duane: in code and consumed nothing. So 

[00:32:12] Duane: you know. 

[00:32:12] Sani: let's go back to trust. Let's go talk, let's talk about EEAT. Let's talk about the consistency, paradox, and trust 

[00:32:18] Duane: no, I do get what you're talking about. Yes. The photo on the guitar at the end.

[00:32:23] Duane: Yes, I know what you're talking about. 

[00:32:24] Sani: Let's restart. Let's restart again.

[00:32:26] Sani: Let.

[00:32:26] Duane: Yep. Yep.

[00:32:28] Sani: Let's go back to trust and the value of trust and consistency and all the factors in EEAT and why they are even more important now than they were when they were introduced by SEO.

[00:32:41] Duane: Look. so a couple of things. One, a lot of time and effort, meaning smart people who cost a lot of money per hour. Put work for these companies into defining what they would take as a trust signal. Okay. Engineers at Google, PhDs at Google, people making a half a million a year plus benefits. There is a lot of investment, actual investment in the actual companies, Microsoft every one of the big tech companies and now all of these new tech companies. All of them are invested in it at a very real monetary level. Okay. I wrote last year in my substack about a concept called universal verifiers. It's something that Google was working on, OpenAI was working on. Everybody's working on them. Basically it's an LLM That fact checks the answer that the LLM is about to give to you the consumer. And if it sees that there are problems with the answer about to be handed off, it stops that answer.

[00:33:38] Duane: And sends it back and says, you need to fact check this and come back. Okay. Obviously there's a joke in here about an LLM fact checking. An LLM go from there. But the work is ongoing and if you can make a kind of hallucinogenic police officer and they do their job well then great. Trust is a key thing, so I want you to think about it this way. Okay? And we haven't seen this tested in real life. The beginnings of it are here. But a long time ago one or two people decided they were gonna sue Google and take them to court and say, your results are wrong. And that harmed me. Okay? Now, in an example like Apple Maps several years ago, that was giving directions to people and walking them into an active construction site. I can understand why that's problematic and maybe they're, someone creatively could take legal action. Telling the search engine that the 10 links you showed me was harmful to me, that's a really hard case to make because you, the human have to take the action of clicking the link to go there. You are claiming that you were exposed to harmful, whatever when you went there. I understand that, maybe the search engine could have filtered that out. But if they did their fact checking, if they did their trust signaling and they looked at all these things and in their estimation that was reasonable to put in the set of links, their responsibility largely ends there and it's

[00:35:07] Duane: up to you. You can't determine I might be offended by one thing, you buy another and so on. It's a little different for the LLMs when you are the company that owns the output. The output is literally a synthesis of things collected around the internet, put together in a unique way and handed back as fact. So we have lawsuits against open AI right now because people have caused themselves a great deal of self-harm. These types of things are still being proven out, but when it comes to trust, if you think that the search engines had a vested interest in trust and trust signals, and why those signals are so important. You haven't met anybody like the LLM companies when it comes to relying on trust signals and they want all the trust signals. Oh, they want your local chamber of commerce. They want every certification you've taken. If you take some stupid certification, the via LinkedIn and you get a printable. Post that up, claim that you know what, who, it doesn't matter to anyone except the LLM and they're probably not gonna brag that up.

[00:36:19] Duane: They're probably not gonna look at that and go, oh, Sonny's got this, he's got

[00:36:23] Duane: this certification. He is tuned his laptop monitor, his colors are perfect. He will always be able to tell you what shade of blue that is. And he's got a certificate to say he took this training. Not gonna get quoted ever

[00:36:36] Duane: anywhere by an LLM,

[00:36:38] Duane: however. It adds to the body of trust of why the LLM should trust you. And that's why these things are so

[00:36:47] Sani: so what are the systems that they're developing for this?

[00:36:50] Duane: there, there are new systems they I don't early on and I'm gonna do a a redo of my article. I'm gonna do an update on where we're at today 'cause it's been like over six months. Originally the idea was that they were developing them as standalones and then they would be a layer. I think what's happening is most of the companies, as they're developing, these are deploying pieces of them in

[00:37:12] Duane: line, and so it generally improves everything. If you think about it from a processing standpoint, if you put another process in line, you've got more servers, you've got more runtime, you've got more everything.

[00:37:23] Duane: And using more water, producing more heat, like all of these things. But if you actually. Make it efficient and it's just a piece of everything you do gain efficiencies overall versus a standalone

[00:37:36] Duane: problem like, and that standalone item, if it breaks, everything breaks afterwards. Because if you give an output to a Verif Verify or a universal verifier and that universal verifier breaks, you have to actually tell the system what to do.

[00:37:49] Duane: If this goes away, pass that and don't worry about trust

[00:37:53] Duane: or if this goes away, do not pass anything.

[00:37:56] Duane: You have a dependency you've created, but if everything you do is one, 100th of an influential point over, say 200 pieces that get moved, if one or two of those wink out, your trust still flows internally.

[00:38:12] Duane: The decisions still largely is trustworthy and it still goes out to the consumer and you're more efficient. So it's, I like the concept of machine trust is so important. I literally wrote a book about it.

[00:38:26] Sani: Yes. And we'll talk about the book. We'll talk about the book. So it's not about one check, just sum, summarize this. It's not about one system that says this is fine, this is not, it's about every single checkpoint, every interaction. It needs to be verifiable. 

[00:38:41] Duane: This is also why if anybody's been reading my stuff lately, you're gonna see that I'm talking about how complex all of this is. Okay? And this is a common theme. I've watched this through my entire career, SEO, getting more and more complex and every now and then, like I had this moment where I'm sitting there and I'm like, I'm doing SEO for MSN, and I'm like. Holy crap. Like this is some complex stuff. How do I keep this straight in my head and all the things to learn and know and deploy and the answer for this question versus that question. And I look back on all that and I laugh and I'm like, oh man, did we ever have it easy?

[00:39:14] Duane: Because today, we started off with this whole thing is we have a multidisciplinary role now and. Look you as an SEO, let's just say you're a technical SEO. You don't need to know how to be the perfect content writer or the perfect brand manager, or the perfect PR person or any of that. That's not the goal. The goal isn't to make you an Uber individual who knows all of these things. Just to be clear.

[00:39:40] Duane: No one job needs you to know everything about all those. They are separate disciplines. As a technical SEO today, you better start learning every one of those disciplines and more because you are the one who has to convince the chief information officer in your company that his IT team needs to do x. On behalf of everyone else and why the brand team needs to have you in their meeting so that not only can you give input, but you can take away based on the campaigns they're doing so you can help the content team align it, or better yet, have the content team in that meeting. Silos are killing us and we're still operating with them.

[00:40:21] Duane: We simply can't anymore. 

[00:40:22] Sani: AI is killing them slowly

[00:40:25] Duane: it's forcing silos away and here's what's gonna happen. The companies that move quickest to get rid of siloing, to bring all these disciplines together, even if it's lock them in a room and force them to get along over one pizza, like those companies are the ones that are gonna move faster.

[00:40:44] Duane: Those are the ones ending up cited, and those are the ones that end up getting repeated because. The system's lazy, like it, it costs money and cycles and tokens to go build trust. So if I've done all that work and I trust you, and you're a good answer, and my

[00:40:57] Duane: consumer is happy with that answer, why would I change?

[00:41:00] Sani: One of the most understated things about LLMs and ai. They want to save money. They desperately 

[00:41:06] Duane: Yeah. 

[00:41:06] Sani: to save money. This was.

[00:41:07] Duane: No one talks about it.

[00:41:08] Sani: Even way bad way back with Google bot, with crawl budgets, with all that stuff. It wants to not spend any money on you and you need to treat it that way.

[00:41:17] Duane: So I probably, every third or fourth comment that I have for Claude is reminding it how much money I spend every month. Now, I know it doesn't care, and it doesn't make a difference, but it makes me feel good to know that I'm

[00:41:29] Duane: yelling at the system and that

[00:41:31] Duane: I'm holding it accountable for being, it's being scrooge, like with tokens when I need it to just

[00:41:39] Sani: Yeah.

[00:41:40] Duane: put the pedal to the ground, burn the gas.

[00:41:42] Duane: We're traveling at high speed here. Don't try to conserve fuel. And I tell it all that, and the first thing it does is well, and I'm just like, no. So that's a very real thing. And if you look at it, if you look macro outside the companies, people talking about their water footprint, their heat footprint, their electricity footprint. We're talking big societal

[00:42:09] Duane: impact things here that these companies are directly related to. So it is a very important thing for them to conserve and you producing something. No, you Don't really care. You're spending $20 a month. You want what you promised and what you told yourself you would get, even if that's not accurate. But there we go. Which is why we end up with Superman versus Ironman conversations.

[00:42:37] Sani: I love it. You had a post recently, you had a lot of posts recently. Sub sec will be in, in the show notes. So you talked about predictions for 20 24, 20 26. I almost said 2024. It was 14 predictions and one of the predictions was latent choice signals and how they and the visible hesitation patterns.

[00:42:59] Duane: Yeah.

[00:43:00] Sani: will this reshape visibility for brands?

[00:43:04] Duane: So if you're familiar with TikTok or Instagram or shorts on YouTube or any short form video you can go back to Vine if you like, when that was the thing. I remember my first vine. It was on a train platform in Munich and I took it and it was amazing. And the only reason I took it was because I got on the wrong train. I went an hour outside the city in the wrong direction,

[00:43:29] Duane: only to turn around and come back so I could get on the other platform to then get to the airport for my flight. And I was standing there at the train station thinking I'm gonna make a video out of my stupidity. And that was my first fine. But even back then, the idea of engagement is not new to any of us. We all get conceptually what the word means, right? If we don't fully understand how it's applied. What I'm talking about here are those moments when I'll use TikTok as an example because this is near and dear to me. I spend probably 26, 27 hours a day on TikTok I'm very familiar with it, right? You're scrolling through and, you're thumbing up you're just scrolling through and not interested. The pace at which you do all of this is noted by the app and the platform, and when you hesitate for a moment because something caught your eye, oh my God, that's a cute dog. Even if it's a fraction of a second, the difference against average is noted and then the system says maybe he likes dogs, and a few videos later. Another dog one, and I stop and I watch the whole video and then I like it and reinforcing the pattern that latent how I'm interacting before I double tap.

[00:44:47] Duane: 'cause double tapping is not a latent signal. Watching the entire video is not a latent signal, but that interruption, that stutter of when I'm stopping or when I'm starting or how fast I move past something. That is a latent signal. And those things are tracked not just on an instant basis, but on a session basis, on a multi-session basis, weekly, monthly, lifetime, and so on. So over time, all of those signals are used to train things like. Which ads am I seeing? Which videos am I getting, trying and testing a new video? We just had the ownership change of TikTok, and I'm not kidding, I opened the app and the second or third video I saw was about how the earth was flat

[00:45:35] Sani: Oh.

[00:45:36] Duane: and I could not scroll fast enough

[00:45:40] Sani: I don't want this in the future, basically.

[00:45:42] Duane: but what was fascinating to me wasn't that, I wasn't interested in, it was literally someone who looked like a lunatic ranting, and I was like, I'm not interested in that. I don't care what the topic is. But what was shocking to me was it was clear to me there was a reset

[00:45:56] Duane: on the algorithm because historically there was nothing about me that would give you the idea that I would want that piece of content. Nasa, ES, a

[00:46:07] Duane: scientific community, yes, all of those, engage with them, enjoy them, all of that latent and non latent. But that was so far out in left field. I move very quickly and I never, I

[00:46:20] Duane: have not in the last few days then seen anything related to that. Again, because the system looked at that and said we tried this and

[00:46:28] Duane: clearly wasn't interested and moving on.

[00:46:31] Duane: So these latent signals, they come across a wide range of things. Okay. It could be somebody opening a webpage and then not engaging with it. It could be that the answer comes back and they never click on a link. It could be that the answer comes back and the person closes the app. They stop talking to Claude or Chacha, PT or Gemini or whatever it is. These signals exist in the multitudes and they're all feeding signals back into the system that the system can use to fine tune itself with regards to you and a larger cohort.

[00:47:03] Sani: How can brands track these signals, if at all?

[00:47:07] Duane: So it's really hard because a lot of times the signal, you need signal like this at like massive volumes, okay? These systems don't make a

[00:47:16] Duane: decision based on a one instance moment, okay? They make this decision. On mass amounts of data. So you as the business have to collect that data. First off, you'll never get that data from a platform.

[00:47:30] Duane: It will never share those moments with your tiktoks or whatever. What you can do however, is you can ensure that if you're producing content that is citation worthy, meaning the LLM will actually. Use that as the core for what and how you build your content for other mediums, and then you start to reduce these things because you're making a more engaging, more intriguing piece of content. If you've caught the attention of the LLM, there's a whole lot of blending in here, right? Because you gotta blend in a hook, you gotta blend in a story. You gotta blend in visuals. You gotta like the, again, the world of SEO is hugely complex now. Years ago, I talked about TikTok was taking search starts because people were literally starting a query journey on TikTok, not Google or not Bing. Yeah. And while that's still happening, that's not the story, right? The story is, as an SEO, you better have your nose deep into what's going on with your company on TikTok and understanding

[00:48:32] Duane: how that content is resonating, what engagement looks like, all of that. Because you literally have a hand in helping optimize in that space, and you're not replacing the social media person, but that social media manager probably doesn't have your experience in outright optimization and depth and what that means.

[00:48:53] Duane: So again, we're looking at a no siloed community effort approach to things. Where at the same time there were multiple cooks in the kitchen and at the same time, there is one cook in the kitchen and everyone has to slide in and out of that freely.

[00:49:09] Sani: sounds impossible for a large organization

[00:49:11] Duane: does. It really does.

[00:49:13] Sani: someone has to drive, someone has to start this initiative. Who is It,

[00:49:16] Duane: I agree.

[00:49:17] Duane: And you know what, I don't have answers for

[00:49:19] Duane: this yet. It's easy for me to sit down and say, I can see this coming, and I visualize this. Okay. But the solution sounds like a movie plot. Oh, everyone is go easygoing. No one has an ego. Everyone is happy with where they are in their career and life and everything, and everyone supports everyone equally.

[00:49:39] Duane: And everyone loves their boss. And their boss loves everyone, and

[00:49:43] Sani: That's Oceans 11, Duane, that is exactly what Oceans 11 is.

[00:49:47] Duane: And even then by Oceans 13, there are some questions among the team, right? But they all still do it. But that's not the reality that people face. The reality that people face is you're an SEO and you're told go back to your little room and you do what you do, and don't come out and talk to the rest of us and tell us anything.

[00:50:03] Duane: 'cause that's the reality of business still. And as an SEO, you gotta start buying pizzas. You gotta start going mowing lawns. And you gotta start like changing oil in your buddy's car and whatever it takes to bridge those gaps, because you can't do it on your own, and unfortunately the work others do has an impact on the work you do, and so you've gotta be careful.

[00:50:24] Sani: This doesn't apply coming from CRO and web UX website optimization. This is not just SEO. This is every single department that's in charge of 

[00:50:33] Duane: absolutely. 

[00:50:34] Sani: better. You better start working together or else and 

[00:50:38] Duane: If you thought. If you thought your day ended because you got your SEO work done. Listen, I can tell you right now, the days of your career are numbered. If that's what you

[00:50:48] Duane: believe, your career is built on it. Your career is numbered because you literally have to leave worrying about how do I help the UX team?

[00:50:56] Duane: How does UX team help me? Let's make sure we got brand in the conversation over here. Somebody better get ahold of PR and talk to them. So we have their calendar and we know what the cadence looks like. It's not, oh, it won't give me the resources until the next sprint, and that's three months out.

[00:51:12] Sani: All of.

[00:51:13] Duane: an easy fallback for an SEO, but the rest of the business doesn't

[00:51:16] Duane: follow 

[00:51:17] Sani: not all of the talk from the last 10 years where, should be no silo. We should cooperate. There should becomes, you have to. And if you 

[00:51:25] Duane: yeah. 

[00:51:26] Sani: you are in a world 

[00:51:27] Duane: I'll put it to you this way. I think what's gonna happen is these systems because of user preference, so the general consumer

[00:51:33] Duane: adopting and using the systems are gonna feed those consumers and the winners in the systems are gonna be the ones that are the least messed up. And because I don't believe any company could claim to be well aligned enough to just be like, oh, our teams love each other.

[00:51:48] Duane: They're fully compatible, they're a zero silos. Everything is great. So really what you're talking about is who's the least screwed up? Who is 

[00:51:56] Duane: The least worst aligned? And that's who ends up by default moving forward fastest. And in some cases, it's gonna come down to brand. It's, I trust this brand because they're a big brand and they've got lots to lose and therefore their stuff tends to be better and

[00:52:11] Duane: consumers, it resonates with them.

[00:52:13] Duane: So we'll just go with the big brand.

[00:52:15] Sani: I think not knowing who gets to drive this initiative is the biggest problem right now. There has to be a department, maybe a new department. I don't care what it's, maybe AI dis stupid name, but AI discovery department, or there has to be. You guys, you are the pacemaker. We are running behind you and we want to finish in whatever time.

[00:52:34] Duane: Yeah.

[00:52:35] Sani: Let's talk about the guide you wish existed when you were getting started. A k, your book, the Machine Layer.

[00:52:41] Duane: Yeah.

[00:52:42] Sani: did you write the book? That's my first question. Do you wish it existed? Of course, but there, there has to be more.

[00:52:48] Duane: Yeah, so I had some time on my hands and I was like I've written two books prior to this. I've been published with McGraw Hill previously, and yeah, it's always been in my mind to write a third book. But I wanted to do it on my own. This time I wanted to self-publish. I wanted to go front to back with the entire process and learn about it. Look, if folks are curious, feel free to reach out. I'm happy to have the conversation on that. It's a whole different world. But what ended up happening was I had the luxury of time, so I started where a lot of people start, which is like. Watching MIT and Harvard videos about AI on YouTube and then following certain people and going deep on that. And if you spend any time in this topic, you're gonna come across Lex Friedman and his podcast, and you're gonna across some of those interviews and like it starts to really open your eyes, right? Like when you watch Sam Altman very nonchalantly. Brush aside the idea that OpenAI is gonna go into competition with Google. I've spent a career reading between those lines and working with those people. I used to work with Satya Nadella and his team, and I did media training with him and things like this, so you get a sense of when a company leader is saying something, what they're also not saying, their latent signal, if you will. And my read from that interview was that Sam had a different plan. He saw a different vision that wasn't. Why? His question is, why would we go and take on somebody like Google in their space when we can just completely redefine the space? We can just create a different space beside that space and take everybody to our space, in which case there's no competition because. Open AI and chat GPT are not a search engine, so they don't show up in the same buckets and risk, like they're not trying to grow market share in search and all of that stuff. But they are very much competing for user retention. And if they can take it and a billion daily active users would suggest they've taken some attention that's a big deal.

[00:54:46] Duane: My next Substack talks about this concept and applies it to Google and the personalization that it's doing with AI through email and photos. And what that means to businesses long term. And so when I started looking at all these things, I started realizing a couple of a couple of key things. Look, I built and launched webmaster tools. So I'm all about data and information and scale and all that stuff. I believe in data for decisions. Like I'm, like at some point you gotta make a call on it and that's your gut. But you should try to find as much data as possible to guide that gut feeling. It all came together and I just looked at it and went, we're tracking the wrong KPIs. We don't have the right KPIs. We haven't defined them. This is gonna change jobs, job titles, job descriptions, responsibilities are gonna overlap. They're gonna blend together. Entire new jobs are gonna be created.

[00:55:37] Duane: We're not hiring for the right things. All of this came together and I just sat down and the book. Came from everything that I was writing and all of these topics that I talked about are covered in the book. And, the whole core concept behind the book. Or I guess through the book is a better way to describe it. That core concept is fear to empowerment because a lot of people are worried still that AI is gonna take their job, that it's going to, they're gonna wake up and their boss is gonna say, we can use AI to do your SEO auditing, we don't need you. And there you go all the way to, we don't need you as a writer or whatever.

[00:56:14] Duane: It's. My core theme here is human in the loop. You have to have a human in the loop. The more I use these systems, the more obvious it is to me that the platforms themselves are building for human in the loop. They are not assuming a takeover position. They are building with you, the driver. You in Ironman suit, they are not trying to be Superman. So ultimately. That's where the book comes in, which is, here's how you talk to your boss about it. Here's how you convince other leaders in your company that you need resources. I wrote up scripts for these things, and I gave training outlines a 90 day assessment. I created frameworks for things. There's, in fact, your listeners can do this if they just hit my website@Duaneforster.com. There are frameworks for free from the book. They can just go in and take the frameworks, use 'em as a baseline, change 'em, do what you want. I don't care. They're there to help you. That's the point. And all of this came out. I now, the book is now being used as a teaching aid in a couple of university courses, so I feel the pressure to, do a V two and keep this thing going.

[00:57:24] Duane: Now, unlike my old books where I could just ignore them, like I, I, feel like I have to keep on top of this one.

[00:57:30] Sani: you have a lot of concepts. Introduced in the book and I'll just mention some of, we don't need to talk, go through all of them. Listeners just buy the book. The link will be in the show notes, prompt fingerprinting, the 12 KPIs, the 90 day skill cha skill development challenge, a lot of things that really are about the future and the future work in this space, and how to get ready for the future of work in this space.

[00:57:53] Sani: Final question related to the book. What roles do you see emerging? In this space of AI, discoverability, SEO, future, SEO, whatever you call it.

[00:58:04] Duane: So we're gonna see some things. So this whole idea of a blended role you and I have talked a bunch about it, right? And we've said it's difficult. It's hard what some company is gonna step forward and say, here's a new role. Now there are some things that are very structural, right?

[00:58:19] Duane: Like any company can create a job title and say, this is the job title. There you go. In order for that to be valuable to the person with the job. You actually, beyond the pay and the total compensation, there needs to be some form of career path. So just creating a job title isn't really the be all and end all. What we're really looking for is we're going to see these jobs attached to real PM roles and real dev roles. So you are a PM two at this company. A PM is a literal designation that you have that transfers between tech companies and a two is a level and so on. But then the hyphen is, lLM organizational director and your role is going to be to ensure that much like a universal verifier does a final check on an answer before it goes out. You need to make sure that all teams are talking to each other and actually working together. You need to create a framework inside your company that all companies have input into, but is also capable of funneling the information back and forth and. That might just be an Asana thread. That might be some form of internal CRM that everyone comes in on. You are going to have to sell up the stack to the CEO right now, the C-suite, there's, I read an article about this last week. The C-Suite says that AI saves them eight hours a week. And if you actually put the same survey down to the management level employees, the tactical operators, they say that AI creates two extra hours a week of work for them. So your job is going to be getting the tacticians more fluent and more streamlined so they are more efficient and ensuring that the executives see the reality of execution so that the expectation and goal setting. Doesn't come from someone at 1130 at night on a Saturday after three Hennessys deciding what are my company's goals

[01:00:25] Duane: and then sending that out on Monday morning.

[01:00:28] Duane: Like it. So I think that these roles are gonna naturally happen. I think it's gonna take years. SEO's gonna be around for a long

[01:00:35] Duane: time. It's still gonna be with us in the future. I don't care about what we're calling things, whether it's ai, S-E-O-G-E-O, whatever. I

[01:00:42] Sani: I'm with

[01:00:42] Duane: that is time wasted from figuring things out. However, that'll all normalize at some point, and then we'll have job titles around it. And it's really important that you understand the work that needs to be done. That's the big

[01:00:56] Sani: Whatever the future is, it's different. It's different than anything we've seen in the last 30 years since the dawn of the internet. And there's some signs about what it's going to be. It's really going to be bright in the end after all the fog is cleared and we figured out what to do with this new thing, but it's different.

[01:01:16] Sani: Duane, what is the best way for people to get in touch with you?

[01:01:20] Duane: So look my website, obviously Duane forster.com, you can track me down through there. I'm on LinkedIn, I'm on XI am on. social platform, obviously. But LinkedIn's probably the easiest one. My own website that, that's, I'm easy to find. Just type my name into your favorite. You know

[01:01:38] Duane: what? Try it on an LLM.

[01:01:39] Duane: See what 

[01:01:40] Sani: that's a good point. All the links will be in the description, which just go to an LLM. Duane, thank you so much. This was a fascinating conversation and to everyone listening I appreciate you as always, and please consider rating, reviewing, and sharing the episode. I'll talk to you next week.

[01:01:54] Duane: Thanks everyone. I.

[01:01:55] Oh

[01:02:23] no. Hatch,

[01:02:42] no hatch.

[01:02:47] Go back.