From novice to pro: A layman's guide to mastering AI chatbots

by Tom Holmes

11 May 2023 ยท 6 minutes

When you first dip your toes into the captivating world of AI chatbots like Athena or ChatGPT, it's easy to make the rookie mistake of treating them like ordinary search engines. You might ask questions like, "What is the best tapas restaurant in Madrid?" or "Where can I get an amazing haircut in Singapore?" While you'd still get a decent response, you might not be fully tapping into the true potential of these AI marvels.

Considering that ChatGPT, which powers Athena, is based on a Large Language Model (LLM) with a training data cut-off of September 2021 (at the time of writing), it's crucial to keep in mind that some responses could be outdated. New information emerging after the cut-off might affect the relevance of the answers you receive.

But beyond that, there's a more profound reason to avoid treating AI chatbots like regular search engines: it's simply not what they're designed for! To truly harness the incredible power of an AI chatbot like Athena, it's essential to understand how it generates responses and how it differs from something like a search engine. Once you understand this it will take you a long way to understanding the usefulness of AI chatbots and unlock their full potential!

Using an AI chatbot like Athena as you would a search engine is like using a hammer as a can opener; you might get the result you want, but it is an inefficient use of the hammer.

A blank canvas

If you have played around with Athena you will know that as soon as you open the app for the first time you get thrown into something called a "New Topic." But what on earth do we mean by a "topic"?


We believe the best way to think about a new instance is like an empty canvas, similar to a new project in an app like Photoshop or Logic Pro, where you are at the beginning of a larger endeavor than just a one-off fire and forget-use like a Google search.

The reason for this is that Athena learns the more information and context you give her, so every time you send her a prompt and she comes back with a response that then adds to her understanding of that particular topic, hence the name. That is why when you ask a vague follow-up question like "summarize this in one paragraph" Athena still knows what it is you're talking about:


You may or may not be thinking that this is a trivial observation but this is in fact the key thing to take away from all this, that you are constantly building context for Athena the deeper you go into a topic. Thus you'll receive increasingly useful information the more you chat with her.

Productive power

Building upon what we now know about how we are effectively training (or "priming") Athena every time we send a prompt we can now dive into how we can take full advantage of this to make you a productivity powerhouse!

Luckily Athena comes with just the tool built-in to let you hit the ground running called the Prompt Gallery.


There are many different types of prompts in the gallery to kick things off, but to start we are going to focus mainly on the ones that prime Athena for a back-and-forth conversation.

NOTE: The curly braces {} are to represent replaceable text, the rest is just the boilerplate or "skeleton" of the prompt.

Okay so before we continue let's break down top to bottom exactly what this kind of prompt is trying to achieve.

The "system prompt"

Sometimes a useful initial prompt should be to tell Athena to act as something (or someone) that will help you solve your problem or simulate a person. This can let you "anchor" Athena for the rest of the conversation, and she will respond as such.

OpenAI actually has a term for this initial message: the "system prompt". This system prompt is very important as the deeper you go into the topic, eventually, the context of older prompts gets "forgotten" because of token limits, but not the system prompt. The system prompt will always be remembered so if you're planning on getting the most out of a particular topic, it is best to design your system prompt accordingly.

Here is an example of a good system prompt:

You are a fitness coach AI, and I am your human trainee.
You are to provide me with fitness advice based on my needs and goals.

Depending on what advice I am looking for, whether it's for a workout plan, nutrition advice, meal plans, or any other fitness-related advice you are to respond only as my personal fitness coach.

This will provide Athena with great foundational context and will drastically improve her responses and keep it relevant to the topic by outlining constraints.

Data is king

So far we have covered how to prime Athena for maintaining a response format but what if you what to tailor the response content based on data that Athena won't have access to such as your resume if you wanted her to write a personalized cover letter, or getting Athena to answer questions about a legal document that you do not have time to read?

Well, no problem as all you need to do is give Athena as much data relevant to your inquiry as possible and she can craft her response accordingly.

Here is an example system prompt:

Given an employment contract:
{Employment contract here}
Can you answer any questions I have about the contract?

Notice how the contract itself and the question are separated by triple dashes ---, or "fences"? This is helpful for Athena to parse your prompt and for her to understand the separation of the context and the instructions.

Now Athena has the base data and context to answer any questions you may have regarding it!

NOTE: Athena does not send any of your personal data to us, it gets safely stored on your own device.

Leading by example

Another great way to tailor Athena's responses is by providing examples of the type and format of response you want.

A great use case for this would be for extracting keywords from a body for text for something like search engine optimization:

Extract keywords from the corresponding texts below.

Text 1: Stripe provides APIs that web developers can use to integrate payment processing into their websites and mobile applications.
Keywords 1: Stripe, payment processing, APIs, web developers, websites, mobile applications
Text 2: OpenAI has trained cutting-edge language models that are very good at understanding and generating text. Our API provides access to these models and can be used to solve virtually any task that involves processing language.
Keywords 2: OpenAI, language models, text processing, API.
Text 3: {text}
Keywords 3:

This is essentially lightweight fine-tuning to help guide Athena to provide the right kind of response.

Final thoughts

There are many ways to prime Athena and get the specific kind of assistance you need, but depending on what you need some techniques will make more sense than others.

Like the analogy that was used earlier in the post, it is all about choosing the right tool for the right job.

Fortunately, Athena's Prompt Gallery has great pre-built prompts that can help you get started whether you're a beginner or a pro. You can sign up for the beta today and elevate your AI experience!