How our clients are using AI to boost business results
Costa gives an update on the state of modern AI tools like ChatGPT and how to take advantage of the unique opportunity they're bringing for small business.
Andrew reminds Costa that he has managed to predict a few things about how ChatGPT and other LLMs would be used, and how the technology would progress.
Costa explains that small business is getting AI for the first time, and has both the most to learn and the most to gain from modern AI tools.
Costa lays out a strategic framework for using AI effectively. The first step is to identify whether your business's growth is constrained by demand or fulfillment. Then find the overlap between the vital work in your business and what LLMs (large language models) are good at, and use AI to accelerate those tasks.
A simplified explanation of how LLMs are trained on nearly all of the Internet's data to become "next word prediction" engines. This process means that if the model was not trained on certain data (like your company's return policy, FAQ, or SOPs), it simply cannot produce a good answer. Plus, a weakness: AI can always make mistakes.
The primary strengths of LLMs: Speed and breadth of knowledge. Costa explains this through a case study where a client reduced customer service time from 3 hours to 30 minutes per day by using a Custom GPT.
This segment provides an overview of the major players in the AI space, including Grok, Anthropic's Claude, Google's Gemini, and OpenAI's ChatGPT. It covers their specialties and why ChatGPT is still recommended for small businesses due to its large feature set.
The discussion moves beyond simple chatbots to the concept of AI Agents
which are taught to use "tools" (like web search or file writing) to autonomously execute complex workflows, such as scraping leads or building software.
A real-time demonstration where an AI agent builds a functional, single-page website for a freediving business from a simple text prompt. The AI creates a to-do list, writes the code, and tests the final product, which includes interactive elements like a "box breathing" animation.
The speakers chart the likely future of AI, moving from current LLMs to agents, then to Artificial General Intelligence (AGI), and eventually "superintelligence". They predict that widespread, human-like robots will arrive much later, likely in the late 2030s, due to manufacturing constraints.
A philosophical discussion on what superintelligence
actually means. Using an exponential growth model, Andrew and Costa calculate that by around 2035, AI systems should be capable of generating a full-length feature film, or fix Game Of Thrones Season 8, on demanding task that currently takes millions of human work hours.
