The Evolution and Potential of Chatbots
Chatbots aren't a new concept in computer technology. Before GPT and similar models, they just weren't as useful. For example, even though customer service chatbots are some of the most successful applications, they had to be tediously preset to handle all the frequently asked questions from customers. This often resulted in the occasional unexpected input that broke the application. Facebook Messenger chatbots, for instance, have been widely used by companies like Domino's Pizza and Sephora to handle customer inquiries and streamline processes. However, these chatbots were prone to malfunction when faced with unexpected inputs or queries that weren't anticipated during their setup (Emplifi).
Once large language models (LLMs) [What are LLMs?] were introduced to chatbots, the gap in contextual understanding significantly shrank. ChatGPT and similar models can adapt to the context of their role, the user's needs, and what they can do to help, with much less manual tuning than before. This improvement is mainly because the bulk of training for LLMs is done upfront by the service provider, simplifying the additional training needed afterward. It's clear why this old concept is becoming more relevant than ever before.
However, adoption takes time. When there are so many options, it can be hard to see how effectively it can support simple, everyday problems.
One way to narrow down its options is by creating AI Agents. AI Agents are software programs that use LLMs to take on specific roles, such as customer service representatives, administrative assistants, or data analysts, and help users with a variety of tasks. These agents can be customized for many different situations. The main obstacle is often having clear enough examples to show it how you want it to interact with users. Ideally, they are built around clear-cut, routine tasks. Once built, they're only activated when prompted, and the prompting can be designed to be a natural part of the workflow. Take Grammarly, for example. Its base version sits idly in the background, but when prompted by the user, it immediately checks for spelling, grammar, and general clarity, providing feedback instead of just making changes. This type of AI Agent can be considered a virtual writing assistant. AI Agents streamline workflows by assisting with specific tasks and separating tasks by agent.
The most basic way to create an AI Agent is by directly accessing an AI tool's API. Currently, one of the most generally useful ones for this type of work is OpenAI's Chat completions endpoint [link]. Even better, ChatGPT and Copilot can help you get started with this service, even with minimal coding experience.
Here's an example approach:
Identify and define the task you want the AI Agent to perform. Be as specific as possible.
Ask ChatGPT to generate a plan, including code for the completions endpoint, and step-by-step instructions for this task.
Use the provided instructions to integrate and run the code, while using ChatGPT to debug errors.
This approach, while requiring some effort, is more accessible than ever to create and use assistant-type support quickly and affordably.
Here are a few starter ideas:
Virtual Tutor: Create an AI Agent to help students with their homework by providing explanations and answers to common questions in subjects like math, science, and language arts.
Fitness Coach: Build an AI Agent that can design personalized workout plans, track progress, and offer tips on exercise routines and nutrition.
Recipe Advisor: Set up an AI Agent to suggest recipes based on the ingredients you have at home, dietary preferences, and cooking time available.
In summary, chatbots have come a long way from their early days to advanced AI like ChatGPT. By creating AI Agents, you can focus the general capabilities of AI on solving specific, everyday problems effectively. Making and using AI Agents is easier than ever before, so now is the best time to learn how to apply them to your business, job role, or personal life. Keep in mind that tasks such as fact-checking or dealing with important information should be approached with caution. Generative AI has been shown to hallucinate.
Interested in the topic but still want to learn more? Sign up for a free consultation on my website, and we can talk through ways to get you using AI Agents as soon as possible.