Personalising Your AI: Make Your Digital Assistant Smarter

Set Up Your AI's Core Instructions

To start, we'll give your AI assistant some foundational "custom instructions." Think of custom instructions as a permanent briefing you give your AI, helping it remember your preferences and giving it a starting point for tasks, much like prior knowledge helps a person approach a new situation. This is how you define its baseline behaviour.

Open your chosen AI assistant (ChatGPT, Gemini, or Claude). Look for your profile settings, which are often found by tapping your profile picture or an icon like a gear or three dots, usually in a corner of the screen. A menu will typically slide up with options. From there, select an option that sounds like "Custom instructions," "Personalise," or "Manage settings."

Once you've found it, you'll see areas to input how you'd like the AI to respond. In the box where you can tell the AI your preferences, enter some details that establish a core role for it.

You'll know it worked when the instructions are saved (look for a "Save" or "Update" button, then the menu disappears, or a confirmation message appears), and your AI will now consider this information in future conversations.

Apply These Instructions to a New Task

Now that your AI has its basic instructions (its "prior knowledge" or defined persona), let's see it apply this to a new task. This simulates how a person might use existing knowledge to tackle something fresh, ensuring consistency in the AI's approach.

Start a new chat with your AI assistant. You should see a blank conversation window, ready for your input. Now, ask it to perform a task that directly uses the "skills" you gave it in Step 1. Remember, it's now set to summarise and explain simply.

What happens next is that your AI will process your request, using its new "custom instructions" to inform its response. You'll know it worked when the AI provides a response that includes a bullet-point summary and an explanation tailored for a beginner, showing it's applying the behaviour you defined.

Introduce the "Idea" of a Tool for New Abilities

Advanced AI systems can often "use tools" to extend their capabilities, much like we use a calculator for maths or a search engine for information. While your personal AI assistant might not have actual external tools built into its free version, you can instruct it to simulate having one.

In the same chat, or a new one, ask your AI assistant to perform a task that implies the use of a tool. Imagine you're giving it a new gadget to work with. For example, if you want it to "search" for something, you can ask it to generate a search query, or act as if it's found information from a search.

You'll know it worked when the AI responds as if it has "used" a search tool, providing relevant information about vegetarian lasagna ingredients and preparation. This shows it can integrate the concept of a tool into its task execution.

Refine the AI's Approach with Feedback

Just like a person improves a skill through practice and feedback, you can refine your AI's responses by giving it specific instructions. This helps the AI adjust its approach, leading to better outcomes in the current conversation and giving you a sense of how AI can adapt.

After the AI has provided its response from Step 3, give it some targeted feedback to improve its output. This helps the AI adjust its approach.

You'll know it worked when the AI re-generates its response, incorporating your feedback to provide simpler, more detailed cooking steps and addresses the preparation time, demonstrating its ability to learn and adapt based on your input.

Update Custom Instructions for "Memory"

A key aspect of advanced AI is the ability to retain information and "learn" over time, maintaining a form of long-term memory. While consumer AI assistants typically reset their memory after a chat ends (unless using specific paid features or ongoing custom instructions), you can simulate this persistence with clever updates to your custom instructions.

Go back to your AI's custom instructions (as you did in Step 1). You can add specific facts or preferences that you want the AI to always remember, acting as a simple, persistent memory that carries across different conversations.

You'll know it worked when you confirm the custom instructions have been successfully saved (look for a "Save" or "Update" button, or a confirmation message before the settings menu closes).

Test Your AI's "Long-Term Memory"

Now that you've updated your AI's custom instructions with new preferences, it's time to see if it remembers them across different conversations, simulating a simple form of long-term memory that influences its general behaviour.

Start a new chat with your AI assistant. This is important because it ensures the AI begins with a fresh context, relying only on its custom instructions for persistent information. Then, ask the AI an unrelated question that might give it an opportunity to use its new "memory."

You'll know it worked when the AI's response subtly includes a reference to your favourite colour, nature, or suggests a nature-related activity, showing it has "remembered" your preferences across different conversations and woven them into its reply.