SPEAKER_00 0:00

Google just put it in writing. If AI wrote it, a human has to check it before it goes live. Most agencies read that as bad news, and I read it as the best news this industry has had this year. I'm Troy Dean, and this is the Agency Our. In this episode, we're going to talk about the news. We're going to explore one big idea. I'm going to give you some homework and I'm going to take your questions. And the same thread as every week. Let's hand the delivery over to AI and keep the parts that only the humans can do. Hey, just a quick little sidebar here. If at any point you're listening to this podcast or hanging out on our YouTube channel or consuming any of our content and you just know the timing is right to take your agency to the next level, then go to the agencyhour.com slash headless and apply to work with us inside Agency Mavericks. We only take on a handful of clients each month, and so if that page is up and the application form is open, then that means we've got an available slot. It's the agencyhour.com slash headless, and we'll get you plugged into a great team of agency mentors, a great community of agency owners. We'll help you grow your recurring revenue, get your team, your processes, and your AI dialed in, and turn your business into an agency that is not only highly profitable, but an agency that you just love running. We've been doing this for a long time. We've mentored over 4,500 agencies from every part of the planet over the last 15 or 16 years, and we'd love to help you take your agency to the next level. So go to the agencyow.com/slash headless. All right, let's get back to the podcast. The feed. All right, this week in the feed, uh, open AI launched dots. Yep, they're always on personal agents with their own cloud computer. Now, this is a race for connectivity. Everybody wants you chatting to their AI to get work done, right? Everyone knows who listens to this podcast, I'm a big advocate of Victor. I use Victor every single day because it is super easy to connect it to all of my shit. However, I'm also using Clawed and Clawed code a lot these days because Victor is very expensive. So I'm using Clawed where I don't need it to be connected to all of my things. I think OpenAI here with Dots, they are racing for the same uh end game, which is your attention. They just want you to be using dots. Everybody is, and and frankly, ChatGPT have the biggest market share, right? So if anyone's going to win this game, it's probably going to be OpenAI. The value proposition with Dots is that it connects to your email, your calendar, and your apps, and it keeps working on a goal and it messages you when it has something for you to approve. And this is really interesting. It's not quite there yet. In fact, it hasn't been rolled out to all clans yet. Uh, it is being rolled out. They also have a Mac app, which I find really interesting. I might download the Mac app and take it for a spin. Anything that keeps me out of the browser or keeps me kind of offline, off social media, out of my emails is a welcome uh thing for me. So definitely worth taking for a spin and a very interesting development. Uh the same week, Meta launched Muse for Small Business with Shopify, Stripe, QuickBooks, and Clavio Connectors. Right? It asks for approval before it sends anything, publishes, or spends anything. Now, this is what's interesting. If any of your clients are in e-commerce, their customers will soon send an agent to shop, compare, and book. So you need to ask whether your client sites can be read and used by an agent, not just a person. All right, in the AI uh overviews, according to DemandSphere, who are an AI visibility software platform, AI overviews now show up on most branded searches. This comes from uh a combination of Search Engine Land and HRFs who reported on this. Demand Spheres say that AI overviews appeared on about 26% of tracked branded queries at the start of September and over 80% by the end. I'm not buying that growth in 30 days at all. I think there is definitely something distorting Demand Sphere's data there. But the point is that branded queries are now appearing in 80 to 90% of AI overviews. Okay, so what this means is that less and less people are clicking. Google haven't announced any changes. Um hrefs puts it at about 83% of branded queries. Uh demand sphere say that it peaked at about 90% on the 27th of September. The angle here is that when someone Googles your clients, Google's AI answers first. And we have heard that more than something like 60% now of Google searches end without a click. So do you know what happens when someone Googles your clients' brands? If you don't, you need to figure that out. No one's measuring the effect on clicks yet in in terms of like the last 30 days of data. So don't promise any clicks to your clients. We spoke about this on the episode uh of the podcast last week. Clicks are kind of done. If you're reporting on traffic, that's a that's a dying model, right? You need to be reporting on outcomes, which is a nice segue into uh the big four consulting companies that are now reporting that they are invoicing on measurable results instead of hours. So consulting isn't dying. It's just the business model behind it is changing. This comes from the Morning Brew. Uh Accenture's revenue was up 7%, and the stock jumped as much as 22% in a day. It also signed a record number of big contracts. McKinsey and Ernst Young are moving to charging for measurable results instead of hours. Strategy decks are now under 20% of McKinsey's work, right? And so what this means for us, I know that we're not McKinsey or Ernst Young or Accenture, don't worry, but I'm looking at what's happening globally in the market that we can take our cues from. Advice still sells, but the hourly model is what's dying. So you need to be charging for measurable outcomes, not hourly rates, right? I still am a massive believer in paid discovery and selling paid discovery up front and selling for your advice, but they're not charging an hourly rate and having a block of hours as a retainer, right? Because frankly, clients don't care how long it takes you to do a thing, right? Like I'm not interested. I don't care how long it takes you to service the car, I just make sure the car's safe when I drive it. I'm paying you for an outcome. So, in this week, the deep dive is all going to be about what you should be charging for and what your model looks like right now. The deep dive. All right, on the 2nd of October, Google updated its AI content guidance. This is uh official from Google. This is a quote: it is critical to manually fact-check and review all AI-generated content for accuracy and trustworthiness before publishing. So Google has had this helpful content algorithm or this helpful content doc for a while now. And it says, it's it's now saying what quality raters are looking for, okay? And what it's saying is that effort is human work. Pumping out AI text without manual oversight or curation counts as little to no effort. And so that content, they're not saying that content's going to be penalized, but it's not going to be indexed in the helpful content index. And the same document now warns against fake author profiles, right? Fake author profiles. So all of this landed during the September spam update. What this means is that if you are adopting the approach of, hey, let's just turn it all on and publish everything, that's exactly the kind of content that Google is going after. And HubSpot experienced this. I can't exactly remember when it happened, but HubSpot reported an 80% drop in their inbound traffic from their blog because they were just pumping out hundreds and hundreds of articles that frankly weren't in alignment with the services and the software that they sell. And so they were just trying to publish as much content as possible to get as much traffic as possible. It was all written by AI. And eventually Google said, actually, this is not helpful content because you're bringing people in for a topic that's not related to your business. Right. So one approach that agency owners are now taking is, hey, just turn it all on and publish, you know, 10 blog posts a day. That's the content that Google are going after. The other approach is, well, AI is just too risky, so we're going to keep writing everything by hand. Well, guess what? Your competitor is using AI smartly and strategically, and they are cheaper. I've been talking to many agencies this week who are losing clients to cheaper options. So I'm not suggesting, by the way, that you need to drop your prices because you're using AI. I'm not suggesting that at all. I think if clients are leaving you because they're going for a cheaper option, they may not be the right client. But I do think that it's too late to ignore AI. You need to be embracing AI to do the repetitive computer work, like writing articles and publishing articles to websites and generating social media posts and generating emails for your clients to send, right? You just should not be sitting in front of a computer typing that shit out. You should be supervising it and tweaking it and editing it and overseeing it and fact-checking it. By the way, fact-checking doesn't mean copying and pasting an article from ChatGPT into Claude for a fact-check. It actually means doing some research. But the the point is now that it takes you 10 minutes rather than an hour and a half to write an epic blog post on a particular topic for a client. Okay. So here's the model that's working right now. AI does the typing, it drafts content, it does reporting, it does auditing, it checks ad accounts, it makes recommendations, it can design strategy docs, it can build strategy decks, right? As drafts. The human does the judgment, right? Is this true? Is it on brand? Is it what the client actually needs right now for the client to achieve their goal? That's a strategic decision that a human should make. Yes, AI is going to make suggestions. AI is going to make suggestions that you and I have never thought of because AI is trained on a lot of content. The large language model is being trained on years of data that you and I could never ingest or keep in our brain. So, yes, AI is going to come up with great suggestions, but you know your client intimately, and you are the one that ultimately decides is this the right thing to do for what the client has told us they want to achieve. Nothing ships without sign-off. Muse for small business by Meta works the same way. Approval before it sends, publishes, or spends any money, right? And the approval step is the part that the client's actually paying you for. Think about AI when it's delivering the work for clients. Think about AI as a team of juniors. If you hired five juniors right now in your agency to run ad campaigns for clients, would you put them on your best client accounts and let them off the leash and say, go for it? Of course you wouldn't. You would approve their work before they publish it. Right? It's the same with AI. If you hired a junior copywriter, would you let them write a blog post and publish it on your best client's website? No, you would approve it, or you would have someone else in the organization approve it before you published it. There's nothing new here. Right? The agencies that I'm seeing that are stuck just don't trust AI enough because they're trying to give AI too much autonomy. It's not autonomous. It needs a human in the loop to approve it. So how do we price this? Well, if you if I'll put the show notes in, I'll put the links in the show notes here, but if you look at the uh the report and the article from McKinsey and Ernst Young, charge for the outcome and the judgment, not the hours. The hours are shrinking. You shouldn't be spending as many hours, right? Stop selling four blogs a month and sell. We own your search visibility and we sign off on every word that goes out. We take that responsibility. We will get you more traffic and more visibility through search, and here's how we do it, right? So this is exactly how we do it when we set headless up for our agencies. Nothing gets approved. All the skill files that we write all put things in front of an agency only in Slack or Microsoft Teams before they get approved. We don't run autonomous agencies that are completely off the leash. I think it's just too dangerous, right? I don't think we'll ever be there. I mean, if you're just having uh AI do admin tasks that are never going to see clients, like just logging into your own Google Drive folder and reorganising things, that's fine. Maybe it can be a bit more autonomous and a bit more agentic. But if you're doing things that are client-facing or you're delivering deliverables for clients, you must, I believe, you must leave a human in the loop. The simple checklist is fact check, claims and numbers. You know, in Australia, if you're a health practitioner, you are governed by the Australian Health Practitioners Registration Registered Association, uh, the Association of Health Practitioners Registry of Australia. I can't remember what the friggin' acronym stands for, but APRA, they uh govern what you can and can't do as a health practitioner. One thing you can't do is you can't publish testimonials to say, hey, I went and saw this chiropractor and he fixed my back pain. You can't do that. What you can do is you can say, I went to this chiropractor and their practice was beautiful and their parking was very easy and they were lovely and I had a great experience, but you can't say they fixed my back pain. Right? There are regulations around that. So fact check claims and numbers. Make sure it's a real author with real experience. So don't, you know, don't let AI make stuff up. Um, something original the AI couldn't know, like a client's story, some data from within the organization, from within your client's organization, or an opinion from your client, the owner of the business, the chiropractor, maybe they've got a particular way of doing things. They can write blog posts about their technique. They just can't use social proof that say that that particular technique fixed back pain, right? Check titles, meta, and alt text. Google's update calls these out. So Google is specifically looking for things that get missed, like titles, metadata, and alt text that gets missed by AI. So, you know, back to basics. This is this shouldn't come as a surprise to any of us, uh, but make sure that you've got a human in the loop checking the juniors' work because you that's how I think about AI is that they are a highly productive, highly skilled, but very inexperienced junior. Done beats perfect. All right, this week on Done Beats Perfect, I want you to pick your top five clients, Google their brand names right now, screenshot the AI overview, and then ask yourself, is this accurate? What sources does it cite? Is the client's own site one of those citations? Fix what's wrong at the source, in other words, the about page, the Google Business profile, the reviews, the directories, and send each client the screenshot with one line on what you're fixing. This is the easiest account management win that you'll have this month. If you're not doing this, I promise you your clients are. And if they're not happy, they might not say anything to you. They might just quietly disengage and fire you and go to another agency. Okay? So pick your top five clients, Google their brand name, screenshot the AI overview, fix a couple of things, send them a screenshot telling them that you're fixing it. All right, well, the fan mail feature has been very popular with one person. Thank you to Dave in the school community who tested it for me. That's right. Uh, he went to theagencyhour.com slash fan mail. There's the link at theagencyhour.com slash fan mail. And he uh clicked the button and he left me a voicemail to test that it works and it did. So thank you, Dave. I get it. Maybe you guys are a bit gun shy and you might feel a bit embarrassed if I play your question on the podcast. Uh, but remember, if you're if you've got that question, someone else out there has that question too, and they're also too scared to ask. So go to theagencl.com slash fan mail, click the button, leave me a voicemail, I'll play it on the episode next week or the weeks after, and I'll answer your question. I'll leave that feature up there. Um, if you guys don't use it, I'll just stop talking about it. What I will say is the number one question that is coming up in our internal channels and in the school community, which I'll put a link to uh to that in the show notes as well. Uh, it's a totally free school community. We're almost at 400 members now. We've only been open a few weeks, so it's growing. It's great. I'm loving uh answering questions in there and hanging out in there. And I'm also using that school community as inspiration for new content and new videos that I'm going to be making. Uh, and one of the big questions that's coming up is clients are saying to agency owners, hey, I'm just gonna use AI to do this myself, right? And so there's a couple of things here. One, they're probably not the right client for you, right? Your best client is time poor and has money to pay you to do a thing. If your client is trying to decide whether it's worth paying you to do something or whether they're gonna use AI to do it themselves, they're not the right client. I've had that conversation recently with someone. I said, no, I'm not gonna hire you to do that thing. I'm gonna use AI to do it myself. I am not the right client for them at the moment. I might be in the future, but I'm not right now. The second point is it's all, and I think I mentioned this last week on the on the podcast, but it's all very well to say, I'm gonna use AI to do this. But the question is, how do you know what you should be doing in order to achieve your goals? And hopefully that's where your experience being in the agency game for however long you've been here and the experience that you've had across all of your different clients' business models and challenges and the constraints that they've got, hopefully you've built up some strategic muscle and you are in a better position to help your clients figure out what to do next in order to achieve the goal. I can promise you something. If a client comes to you and says that they're gonna use AI to do something themselves, what they're doing is they're just gonna copy a tactic that they've seen online, that maybe they've seen their competitor do or that they've seen someone else in a different industry do, and they're gonna go all in on this thing and use AI to help them do it. My question for them is always how do you know that that is the next best thing for you to do to achieve your goals? And what do you have to say no to in order to spend the time to use AI to do it? Right? Maybe they're gonna, maybe they're saying they're gonna publish three blog posts a day because they heard that that was a good idea on a podcast. And you can say, well, actually, Google's helpful content index is now saying that that's not helpful. Don't do that because you'll get penalized. So the question is not can I use AI to do it? Of course you can. The question is, what should I be using AI to help me do? What is the right next thing for me to do in order to achieve my goals? Talked about this on the episode last week. Most business owners just don't know what their goals are. So your job is to sit down with them and help them figure out what their goals are and then help them design a strategy, a plan to get closer to those goals. Okay. All right, just to recap some of the things that we've spoken about in this week's episode, AI does the work just like a junior employee. You sign it off. That's what clients pay for now. They're paying for measurable outcomes. I think they expect that you're using AI to do the job. Quick little sidebar as well. YouTube is cutting reach for channels that re-upload barely changed clips. This has been a strategy uh for a while now. YouTube is just wanting to put more content out there, uh, uploading a clip that is barely changed from the previous week. YouTube's cutting reach on those. So the algorithms are getting smarter, uh, penalizing content that's not unique, that's not helpful, that's not original, right? So even though AI is here to help us produce a lot more content, the truth is that we still want to know what you think. We still want to know what your opinion is, and everyone still wants to follow someone who's a few steps further ahead on their journey. And that's your job, is just to be a little bit further ahead, okay? Hey, next week on the podcast, I've got a guest, a special guest coming in. He's actually one of our clients, Darren Rayner from Magnify Media in Canada. Uh, he's a video production agency. He works with some very big clients like Lululemon, Red Bull, Meta. Uh, he's got a great uh process and a great um uh framework for helping brands produce video and get more reach on social. It's gonna be a fantastic conversation. Stick around for that. So, Darren Rayner from Magnify Media coming up on the agency hour next week. All right, my name's Troy Dean. I hope you're enjoying the new format of the agency hour as much as I am. Go to the fan mail, uh, send me a message, uh, leave some comments, do the thing, share it with anyone you think might benefit from the agency hour, and I'll look forward to hanging out with you again next week. I'm Troy Dean. Let's get to work.