S1. Ep40 - What Do Anthropic and Google's New AI Models Mean for Your Business

Katie: Welcome to the AI Automations for Business Podcasts. We're your hosts, Katie and Null, and we'll be discussing how you can use AI for your business along with the latest news, updates and automations to help you stay ahead of the curve, allowing you to grow and scale your business more efficiently. Please be sure to subscribe and I hope you

Noel: enjoy this episode.

Katie: Hello and welcome back to another episode. Hi, hello. I'm Katie. And as all, I've got Noel here with me. Hi, Noel. Hello.

Noel: Yeah, look forward to getting into this one. It's been an incredibly busy week in the AI world.

Katie: I feel like you're very excited about this one because so we're going to be talking about the new AI upgrades and how they matter for like small service-based business owners. But I feel like you've been quite impressed by all of the upgrades and yeah yeah

Noel: it's been crazy so you know what one week someone releases the most cleverest model available you know the smartest out of them and then you know within a week that someone else goes well I've got one that does a bit better on this uh there's also our flagship so yeah it's yeah it's bit to and fro

Katie: so who who did the update first Anthropic or Google?

Noel: So Google, they were the first ones. So that was last, well, it would have been last week for us as a recording it. But they released Gemini 3. And at the time was miles and miles ahead of anyone else on all kinds of benchmarks. They, you know, whenever new models release, they love a benchmark. Yeah.

Katie: And then Anthropic kind of just went, hold my beer.

Noel: A little bit. Yeah, I don't think they quite beat Geminae in every use case. I think reading Anthropics release notes and all that sort of stuff, they were kind of cherry-picking the ones where they were really good. I don't think it fully covered everything. Whereas Gemini and Google, they were so excited. They just showed everything almost.

Katie: Okay. Well, let's go through some of the upgrades. and then what we want to do is obviously we don't just want to tell people about the upgrades which is great it's nice but we we also then want to share with you you know how this actually impacts your business how it helps your business so we'll be talking about that once we've gone through just some of the upgrades so no do you want to start with the Anthropic updates first? Yeah.

Noel: So their update is a released Opus 4.5. So if you're unfamiliar with the model levels with Anthropic, opus is like their top tier. It's like the best of the best. And then they go down to Sonnet and then Haiku. So Sonnet 4 was, sorry, Opus 4 was the best one. And then now they've upgraded that to 4.4. 5. So yeah, it's really, really clever. It's got loads of, it's really good for things like coding. And I always find that Anthropic has always been good at coding. So where I use Claude code quite a bit to do some bits and stuff and build some internal apps. And yeah, I'm looking forward to giving Opus 4.5 ago and just seeing how good that is because Sonnet 4.5 is brilliant. It just gets on and does a good job. It's a very good employee.

Katie: Well done.

Noel: Employee of the month.

Katie: It's also 66% cheaper as well than previous APAS models, isn't it?

Noel: It is, yeah. That's on the API, yeah. So that's really, really good.

Katie: That is. Why do you think they've gone with that?

Noel: I think it's kind of a general trend. So Gemini, most of their stuff currently, so all the Gemini 3 stuff, you can pretty much have a go with it for free. So although it's like incredibly clever and superpowered and all that sort of stuff, you know, you can just go into Google AI Studio and then just play with everything for no cost.

Katie: So I'm trying to keep up then with Google or like to kind of break the barrier a bit to make it more accessible to people?

Noel: I think so. It's kind of, yeah, I think it's part of that and also part, you know, the infrastructure's now there. It's getting better. It's probably getting cheaper because Open AI's models as well, you know, they've also been trending slightly down in terms of cost. So I think as we're going to go forward, I think, you know, we'll, we'll hit a point where it can't get any cheaper, obviously. But for now it's trending down to like a more. sensible level so yeah yeah i think we could yeah so they they all seem to kind of like just try and give as much away as they can't with i mean google i don't see how they're making any money with some of their releases but obviously they make so much money elsewhere in the business you know they can they can afford to give free access to things i

Katie: think that's the thing when these giant companies you know or are doing things like this lLM Like, they kind of have such a, like, I guess, like, advantage over a lot of the smaller companies because they're so well established. You know, they've got huge teams, you know, globally.

Noel: Yeah.

Katie: So I feel like actually to then let people use this for free is a lot. very, like, it's a very easy concept for them. Like, it's easy for them to let that happen.

Noel: Yeah.

Katie: Whereas some of the others, it's, you know, they actually need as many users, you know, paid users as possible.

Noel: Yeah, it gives them great exposure as well. I mean, I don't have an X account, but I'm sure every time they release stuff, it's full, you know, people trying it out and showing what it can do and all that sort of stuff.

Katie: Yeah. Google were always really active on X back in the day when I was. So, yeah, it wouldn't surprise me if they still are. Or even like if they're now I would say probably active on threads as well.

Noel: Yeah, yeah, absolutely. The other thing, going back to the Opus 4.5 models, it's also a lot better at a gentic tool use. And it might be thinking, well, what does that mean? But when we give, like, an AI agent, you know, a job within our automation, you know, we can assign it all kinds of different tools so it can access your calendars. It could access CRMs and all kinds of stuff. And one thing I'd always note with anthropic models is their tool usage is usually spot on. even with a really rubbish system prompt within the agent, it seems to always figure things out and do things better than what other people like OpenAI and their models can do. So, yeah, with 4.5, it's even better. I'm not sure how it could have got better, to be honest. But, you know, I think we built one in the school community, which has like a really, really basic system prompt for a task list and yeah the anthropic agent just nailed it the first time yeah whereas you put open AI in there it was like maybe I'll use that tool maybe I won't see yeah it's a big upgrade

Katie: okay so any other um anthropic upgrades that we need to be aware of

Noel: uh so they've also talked about like enterprise safety uh I did read the press releases a bit haulier as to what they mean by that but you know i'm guessing you know they're making things a bit more stricter on you know on the security and compliance side uh which is which is always good yeah and they also released a little side product as well um so if you're on the max plan uh you also have a chrome extension uh which will control your browser so it's a bit like where we talked a few episodes ago about atlas and uh AI browsers and things that. So this one plugs straight into Google Chrome and things that. And it can go off and, you know, you, they were, apparently they say you can use like CRMs and Slack to do tasks and things that. But yeah, I'm a bit worried about that.

Katie: Yeah, you don't feel very confident letting it run around in your CRM, do you know?

Noel: No, no, no, no, no, definitely not. though. I've got, you know, we've got, I would say like normal automations, you know, that's still king when it comes to doing stuff for the CRM. But I would still like to be clicking those buttons and then the magic happens. Rather than AI click the wrong button and then the magic happened somewhere else where it shouldn't have done.

Katie: Yeah. I feel like you've got trust issues with Anthropic Opus 4.5 now.

Noel: Just AI browsers in general really when it comes to controlling Okay.

Katie: So it

Noel: is nothing

Katie: personal against Vegas 4.5.

Noel: No, definitely not, though.

Katie: Okay, wow.

Noel: I mean, until it comes to Christmas shopping, it can go off and do that. That's fine.

Katie: Well, we're recording this on the 26th of November. So you've got about 28 days before you start yours.

Noel: Yeah, before I start the agent, yes.

Katie: Yeah, to start your Christmas shopping. Yeah.

Noel: Plenty of time.

Katie: And there's me, like, panicking because I haven't finished mine yet.

Noel: Yeah, very true. It's the same every year. You know what you're getting into.

Katie: Okay, so let's talk about Google and Gemini 3.

Noel: Yes. So this is an incredibly clever model. they again they talk about like a genetic tool use and how things have all improved when it comes to controlling agents which is great for me I've never really used a Gemini model in an agent yet so I'm probably going to give it a go and see how that measures up because yeah usually I get on so well with anthropic models I don't really get to have a look at anyone else but I think on this occasion I'll give it a go I find Gemini,

Katie: when you are asking it questions, it's very hit and miss on the accuracy. Obviously, I haven't used Gemini 3 yet, but I do find it's very hit and miss if it's going to tell me the truth or make something up.

Noel: Yeah, struggles with the hallucinations. definitely. But I think with three, it's, I would say, I've had a quick go with it, and I'd say it's better than 2.5. I mean, you're always expected to be better.

Katie: Yes.

Noel: But I'm pretty sure there was a hallucinations benchmark that they showed in some sort of bar chart or scatter graph or whatever.

Katie: Is it going to

Noel: stop lying

Katie: to me?

Noel: Yeah, so yeah, I'm sure it showed that it was doing a lot better. than it used to and all that sort of stuff. But yeah, yeah, it's always been an issue. But, yeah, now we can obviously have access to Gemini 3. Obviously, if you use, like, Gmail or, you know, you're heavily invested into the Google apps and all that sort of stuff.

Katie: Yeah,

Noel: like docs and sheets and all that sort of good stuff. So, yeah, Gemini 3 is now embedded into those tools. as well.

Katie: Which is really cool.

Noel: Yeah, yeah, so I have a lot smarter than, yeah, the previous models. So yeah, it should, should, in theory, be a lot better. Yeah.

Katie: You know, I'm really excited about that integration into, you know, especially like Google Docs. I feel like that's a bit of a game changer.

Noel: Definitely, yeah. And also with Gemini 3, it's kind of more multimodal than the other models. So it can read documents, you know, it can go off and understand videos and audio from even images and things that. So it can understand all of that and bring that into its context. So, you know, that's actually really, really good. And I also find actually Gemini 2.5 is also really good at some of those bits as well. But, yeah, it seems to have got a lot better. So if you're kind of like on the visual side and you're like, well, here's a new. image, help me describe it so I can create something, you know, it'll understand it a hell of a lot better, which is really, really cool. But yeah, the other big things, obviously, they've got a different reasoning sort of model, so they call it deeper reasoning. So if you ask it a harder question, you know, it will dynamically think about it for longer and things like that to give you the better response, which is just pretty awesome.

Katie: Okay. Yeah.

Noel: If you ask it for a simple joke, it's going to take two seconds. Okay.

Katie: Is that where you're getting all your jokes from these days?

Noel: All my cracker jokes, yeah. Oh, yeah. Yeah. It's also with these models, they're a lot better at building user interfaces in that as well. So what on the code inside? So if you've ever coded a website or an app on things like Loveable or anything else like that, or maybe you've used OpenAI Codex or whatever, you'll find that they all kind of look the same because they're using the same libraries for the different parts kind of thing. So for me, it's very easy to spot when somebody's used AI to create a website. But now with Gemini, it's a lot better. And I've been building an app recently and I created it with 2.5, it looked terrible. And then as soon as Gemini 3 came out, I thought, oh, hold on me, I'll give it. Let's give it a go. Let's see what happens. And then it completely redid the UI and it looked brilliant. You know, whereas the other one looked really vanilla, really basic. Although it worked, which was great, it just didn't look amazing.

Katie: Yeah.

Noel: But yeah, now it's a lot better.

Katie: Can I ask, what are the telltale signs of an AI generated website?

Noel: So it's kind of hard to do. I guess, but they always seem to...

Katie: Is it just how it looks?

Noel: Is it just the visual

Katie: element of it?

Noel: Yeah. Most of them use a UI library called Shatsyan. And basically they all use the same components and it all looks very, very similar. So most people's hero images look the same. And then what it always does after a hero image is put six cards underneath. And those six cards always look the same sort of. style on different website.

Katie: Got you. So it's kind of a bit like when chat GPT spits out a caption for you, it will always follow the same format. Yeah. Yeah.

Noel: I mean, it lives very good. I mean, it is functional and it works very well. But yeah, if you want a bit of flare, it's not going to do it.

Katie: It's that same thing as a chat GPT caption. Like it's there. It's, you know, it works. But if you want a bit of flair, then you need to tweak it a bit.

Noel: Yeah, definitely, yeah.

Katie: Okay.

Noel: Yeah. That's very much, yeah. There's also been another, the Aunt Gemini 3, well, they're kind of are, but there's been other two other product releases last week as also. Nano Banana Pro came out, which is their image model. And yeah, that's a lot better than what it used to be, which is, well, I mean, not that it was bad, but.

Katie: I have seen someone who. generated a image, they shared it in a Facebook group. So they, the prompt was, I mean, I'm probably going to butcher it a bit, but it was a, create a image of a coffee cup with my business logo on in a cafe. oh my goodness it was so good like so good like

Noel: that could be the new trend

Katie: yeah obviously you could tell it was AI but it looked so good it was like it actually looked crisp rather than how some of them the AI images tend to look

Noel: yeah they kind of lose a bit of sharpness, don't they?

Katie: Yeah, and it looked really realistic. Yeah.

Noel: I'm excited

Katie: to play around with that.

Noel: Yeah, I've not had much time to have that much of a deep dive with it, but the image editing in there as well is also improved as well. So, yeah, if you're like switching things around or adding people in, it's also a lot better, which is pretty awesome. And I guess the final product that they released was a tool called anti-gravity. which helps you code with AI. Now, I would say it's pretty good. I've had a go with it already. But what it does is, you know, you can ask it to build something or code something or, you know, to edit something that already exists or whatever. And then with the new agent sort of superpowers it has these days, it then creates itself a little to-do list. So it's like, right, well, to get to this end goal, I need to do all of these different jobs and it goes through them one by one and ticks them up, which is really, really handy. But in my experience, using the tools so far, is yes, it's great that it does that to-do list and checklists, but sometimes it can forget the context of what you're trying to do. So I was trying to get to build an iOS app. And I had a barebones structure of a X-Code project. so I'd already got Swift files in there and stuff and it completely ignored that and tried to build me a web app and I was like no no no no no we're not using that we're using Swift and it's like oh yeah oh yeah I didn't realize that I was like you spent all that time creating a to-do list but you didn't really look at what was going on so yeah I also finally get stuck in loops as well where it's trying to fix things so I would say it's very early I mean in six months time it's going to be incredible. I've no doubt. But yeah, right now it's a bit. Not sure.

Katie: Some tweaking still needs to go on.

Noel: Definitely, yes. Yeah.

Katie: As always, I think, with most of the things that get released.

Noel: Yeah.

Katie: They put them out when they think it's good and not when it's great, which I think is actually a really clever thing to do.

Noel: Yeah, definitely, yeah. I'm looking forward to seeing where it goes. Yeah.

Katie: So what do all of these upgrades mean for service-based business owners? Well, AI kind of stops being just a, like, tool, and it becomes part of your business. So it's going from kind of like an assistant to an operator. Would you say that now?

Noel: Yeah, definitely, yeah. It's moving more into like the apprentice role, isn't it, within your business, I think.

Katie: And I think it's moved on so much more. And there are still, I think, business owners out there that think, you know, AI is just a content creation tool. And, you know, we're so advanced from that now. You know, it's, you know, just doesn't, you know, come up with things that you can do. for your business but you know it actually does these tasks for you yeah so it is like you know the admin task it's the research it's you know going into your inbox it's multi-step workflows you know happening without you having to watch over it or be involved um but yeah like you say, I think it is very much like an apprentice now or like a junior employee. And I think the people who, you know, are just still seeing it as a content creation tool or like an assistant. I think those people are using like 1% of what these, you know, can do.

Noel: Definitely, yeah. There's so much more it can do autonomously these days. It's, yeah, it's pretty incredible. I mean, like you said, you know, you don't have to sit and watch it anymore, but there are times where it's like, you know, when I first set it up, it was like, I'm going to watch this at least 10 times, just to make sure. You know, they've still got that, you know, even though I'm pretty happy that it's going to do what I've asked it to do. You know, still got to sit there and watch it, but you would do the same with, you know, like either an apprentice or like a graduate that comes into your business, you know.

Katie: Yeah.

Noel: They could be incredibly clever. Like, I mean, any LLM is like PhD level. So, you know, they're incredibly clever, but they know nothing about the business and how it operates. So, you know, you need to take that sort of knowledge and superpowers from the AI. but, you know, handhold it as if it is like somebody, you know, brand new into your business. That's, I think that's kind of where some people go slightly awry with it. You know, they just think,

Katie: yeah, go for it.

Noel: Yeah, exactly. Yeah, it does need it. Yeah, it needs that handhold in. And then, yeah, you need to check back in with it and go, I wasn't quite happy with what you did there. This is how we want to do it. It goes, all right, yeah, sure, I'll do that going forward.

Katie: Yeah.

Noel: Yeah. Never ask for a pay rise, though, which is always nice.

Katie: Yeah, true, true. So talking of employees, so I feel like these new models and upgrades can actually, like, save you money because you don't need such a, you know, a large team or even contractors. So I feel like it can also, I guess, like assist rather than it's not just replacing. But it could be assisting you. It could be assisting a member of staff. But, you know, it can now do things like maybe a researcher can do or a junior marketer, data analysts. But it all depends, I think, on what your business is and what your business model is as well. because for some people, you know, who can afford, say, a researcher, then, you know, then AI and LLM, AI is going to be like absolutely incredible and valuable for them. But then you might have a bigger business, you know, with a much bigger budget who'd go, well, I couldn't possibly get rid of my researcher because they find things that AI could never find.

Noel: true but then you could also arm that researcher with those ALMs to boost their productivity

Katie: exactly so

Noel: you don't need to hire another one you just need to you know make the existing one more more efficient because it's definitely when it comes to like searching through thousands of documents and things like within like a knowledge base like you know no sort of data analyst or, you know, researcher, human being is going to get to the answer quicker than AI. It's probably going to, it might even find stuff that they wouldn't have naturally found anyway. Yeah. So, yeah.

Katie: Yeah, because I think sometimes as well, like, we think it's, it's either a team member or AI. And I think we often forget that actually they work really well together.

Noel: Yes, definitely. So it's not one or the other.

Katie: you can have both.

Noel: Yeah.

Katie: But I just love as well, like now you're not limited by the size or your team. Like you're only limited to like how you use AI and workflows and automations.

Noel: Yeah, you're limited by your time to build automation and workflows, yeah.

Katie: Yeah.

Noel: Yeah. It's really, it's really awesome.

Katie: Yeah. I think so too. So now as well, I think you can delegate more important tasks. I mean, Noel, I don't know if you all agree with that point.

Noel: No. Well, yes, yes and no.

Katie: So I feel like now AI is reliable enough to do things like market research, creating proposals, maybe like financial protection, projections, helping with strategy, like so many things. I mean, I could be here for an hour, listening to how lots of things, couldn't I?

Noel: Definitely, yeah.

Katie: But a lot of these tasks that would take some people hours and hours and hours can take AI now just a few minutes.

Noel: Yeah, definitely. The speed, once you get it all integrated and, you know, up and run as it should be in working, you know, like the speed of it could blow people away a bit because, you know, you can be, especially when it comes like proposal, you know, generation and things like, things that are, you know, going to put you ahead of competitors. So you're, you know, you getting in there first with whatever, you know, proposal quotation or whatever. you know, getting in there into a business being the first one on the desk. I mean, that's invaluable. I mean, you know, if you're the businesses sat there and, you know, you're typing it out manually or, you know, you're taking weeks to do it, then somebody's probably already beat you and, you know, they've got it in on day one. Yeah. Yeah.

Katie: But I

Noel: feel like

Katie: it's not just about speed as

Noel: well so

Katie: it's not just about how fast that AI can do these things it's also allowing you to have like more time back so you can do more things achieve more things you know whatever that is you want to do with that time within your business

Noel: yeah yeah i guess in a way it kind of makes, obviously it makes people more productive because they can go off and do different tasks, but I guess from like a retention perspective, like, you know, if you're, if someone's in your business and, you know, they're doing lots of repetitive, really boring stuff, like, you know, and then a really fancy new job comes up and it's not much of a pay rise, but they know they don't have to deal with all this rubbish work, you know, they're going to, you know, go off to the other business. So, you know, I guess, kind of, it makes like the working environment, you know, a lot nicer, I guess.

Katie: Yeah.

Noel: You know, people are doing stuff that they really enjoy, you know, and that just, that just helps everyone. More time for a cup of tea and a chat around the water cooler. I don't know. Or the photocopier.

Katie: I would always be the photocopier or the printer. Because I could pretend, one, I could pretend I was in the photocopying queue. There wasn't a queue, obviously. I was just chatting or with the printer I was waiting for my stuff to be printed and I was just like waiting behind someone who was printing off a massive document

Noel: geniuses of doing nothing I

Katie: know I was such a great employee and people often ask why I quit corporate I was never at my desk

Noel: then you joined the Navy where hiding was an actual profession yeah

Katie: yeah yeah I feel like that just went up five levels like I thought I was quite good at like not being at my desk but I feel like I learnt a lot from being in the Navy and yeah yeah

Noel: yeah yeah Yeah.

Katie: So, no, I feel like we're not in the AI helps a little bit phase anymore. I feel like it's, you know, AI becomes part of your, your operations, your business. It's becoming, you know, your apprentice or your junior employee. How are you viewing it with, with all of these upgrades?

Noel: Yeah, I definitely agree that it's going to be, you know, it's kind of going to change how you scale businesses in the future. And although it's not, I know we talked quite a bit about it, taking your job in earlier episodes, but I don't think anyone's going to lose any work, but I think it potentially could get harder for people to get into employment. because, you know, especially with like start-ups or small businesses look into scale, they're not, they're probably not going to look to employ someone. They're probably going to look to, you know, use AI and then, you know, train that to do exactly what they want, you know, and they're never, there's no overhead costs with that either, which is, you know, also good, I suppose. But, yeah.

Katie: It almost goes back to, though, what we were saying earlier, where it's not replace it and it's not one or the other, it's almost giving your current employee AI tools to help in a system. So if that employee or contractor can, you know, really do the job well by themselves or with the help of AI, which I'm guessing, you know, it's going to be trained on your business, then that makes them actually like, you know, irreplaceable almost, doesn't it?

Noel: or really hard to

Katie: replace.

Noel: Definitely, yeah. Yeah, yeah. Because you can't really teach like a human, everything about your business. Obviously, the owner, if they've created it from day one, they'll have a good idea. But as it grows, you know, not everyone could know everything. But if you give AI access to, I wouldn't say everything, but a lot of it, you know, it will remember. and, you know, show you things that you've done in the past that you probably think, wow, okay, we'd never, you know, I completely forgot we did that, you know, but it has access to those files and documents and go, well, look, you know, this is what we did last time, you know, right, oh, okay, yeah, that's really, really helpful.

Katie: Yeah.

Noel: It does it in like two minutes.

Katie: No, is there anything else that you feel you would like to add?

Noel: I don't think there is, you know. I think we've covered a hell of a lot. You know, there's, you know, a lot of new tools now with Gemini. You know, as I say, on the AI, Google AI Studio, you know, you can give a lot of them a go for free. So definitely go and do that. You know, before you start implementing the APIs and the business and you start paying for them, you can try before you buy, which is brilliant. So, yeah, yeah, definitely start giving it a go and obviously, you know, start thinking about, you know, building those agents that are going to, you know, sit by you and whoever you work with and things that and, you know, are there to help you scale the business. And yeah, who knows, as soon as we, you know, hit end on recording this podcast episode, I'm sure Open AI will release something and they'll be like, oh.

Katie: I imagine, I imagine so.

Noel: Yeah. It's just crazy. Get it all out before Christmas.

Katie: Before the end of the year.

Noel: Yeah, exactly, yeah.

Katie: And if you would like any help with building our AI or automations for your business and you want some help, some handholding, then we would love for you to come and join our membership, which is called AI Business School. In there, we teach you step by step on how to create your first automation, how to create things like custom GPTs, along with a lot of other things, and we pair back all the overwhelming noise and information, and we tell you what you need to know and able to get the task done. And, of course, Noel and myself are on hand. If you've got any questions, need any help or any support. And we keep this updated as much as we possibly can, because as we know in the AI world, things get updated pretty fast just when you think you've got used to one thing and then wham they release new things oh it's hard to keep up but we would absolutely love for you to come and join that you can search it on school AI business club or we'll leave the link below for you we would absolutely love to join you everyone is very very welcome we've got people in there who are brand new beginners we've got people in there who are got a bit of AI and automation experience as well. So if you're not sure if it's the right fit for you, please reach out to us. We will only, you know, tell you if it is the right fit for you. If it's not, we'll point you into a different direction because we only want people in there that, you know, is a good fit and, you know, it's the right space

Noel: for you.

Katie: Thank you so, so much for listening to this week's episode and we will catch you next time very soon.