Beyond Chatbots: Why You Don't Need the Latest AI Model to Win

Hi folks. Welcome to another AI Briefing. My name is Tom, and we are back to talk about the latest and greatest in AI. Now, [clears throat] of course, today is the day after the Mythos and Fable, uh, model releases for Anthropic, and of course, there's been an awful lot of excitement about those models. Now, I'm sure they're great.

Um, I tried to do some stuff on it last night, um, that wasn't coding. I saw absolutely zero improvement, but, you know, um, fair enough. But, you know, um, you don't necessarily have to go charging and starting to use the next model. And the reason I say this is because I do a lot of sort of legacy platform engineering, modernization, and building of applications that require, um, distributed compute, uh, data engineering, that type of thing. And I use a lot of AI to help me, you know, come up with that type of stuff.

But you don't necessarily need the most cutting-edge model anymore. I'm less interested these days in the model capabilities, more like: where can I run that model, and how can I also interact with it in a way that is not a chatbot? Because I spend an awful lot of time just chatting with Claude and just, like, dealing with different aspects of life. Um, you know, and chatbots are fine, and you can obviously from that, uh, an agentic, um, programming perspective, um, set them off on long-running tasks and leave them to it. But in reality, AI isn't gonna be who wins the chatbot race, because that's just not a convenient way of interacting with an LLM.

And so from a product development perspective, and when I speak to organizations and companies who, uh, are interested in leveraging AI, the question is not like, "How can I im-implement my latest chatbot? " It-- The question is more like, "How can I save my team, my organization time by taking away the repetitive work, the stuff that doesn't need that interaction? " Or you can interact with it in a different way. So product design isn't, you know, isn't always the same. And so are there ways of being able to leverage LLM technology without having a user having to sit there, chat with it, talk with it or whatever?

And of course, the answer in a lot of these cases is yes. But people are so used to, uh, interacting with a chatbot via text that they don't, um, they don't, uh, end up leveraging it in any other way. Um, and that can obviously have a negative impact on product design. So at Concept to Cloud, a lot of the stuff we do is about how to interact with the data that underpins all of this, and you'll hear me harping on about this a lot, because an LLM is just pointless without the data that underpins it, because that needs to be able to be served up. And so, you know, what can we do to make LLMs more efficient, more reliable, more beneficial to you as a user, as a product owner?

Um, you know, and that's the type of question that is always pertinent. So if you're looking to, like, deploy something from an agentic AI perspective, from an LLM, few questions to ask yourself. Do you need to have the latest and greatest model? Because they're normally more expensive, and they're often slower. So can you use a Sonnet as opposed to an Opus?

Can you use a Haiku? Because, like, that scale of Anthropic models, and of course, OpenAI has the same, as do a lot of the other services, um, can vastly change the pricing and the speed that you can re, uh, reply to a, an instruction as an LLM if you're not using the heavyweight models. So do you need it? And then is there a better way of interacting with that model than typing stuff into a computer? Because I bet you eight times out of ten, maybe not nine times out of ten, but eight times out of ten, there is a better way of interacting with that model than typing some stuff.

That's all from me today. Uh, thank you for tuning in once again to the AI Briefing. Go use that AI. Investigate that tech-technology and figure out how to make a difference in this world. But just make some sensible decisions when you do.

And if you need some help, you know where to find us. My name's Tom. I'll be back soon. [upbeat music] Why hire when you can partner? Concept Cloud's leading engineers build your startup's prototype without the overhead.

Launch faster. Conceptcloud. com.