It’s easy to get started using ChatGPT to prep for your demos, but if you aren’t following these 3 rules when writing your prompts, you may be missing out on its true power. This is the first post in our series on how to use generative AI to make your demos more impactful.
We’ve had technology fads over the years, all with their own acronyms. NFT. PLG. IoT. But GenerativeAI feels different, not only because of its advanced algorithm complexity, but its versatility. You don’t have to be a technology expert to understand how to use it, and the use cases for it are nearly infinite.
Tools like ChatGPT are straightforward to use: you give it a prompt, it gives you a response. But the true power comes from “prompt engineering,” or the ability to write the correct prompt. To ask it a question that gives you an answer that’s not only valuable, but that satisfies your use case.
Let me start with this, and I cannot stress this enough: AI is not a replacement for humans. I am in no way advocating you just have ChatGPT do your work for you. That’s plagiarism. And you may not own the copyright to your work.
What I am saying is AI can be a helpful tool to deliver more personalized demos more quickly. To deliver more compelling demos. We should think of ChatGPT as a tool to help you prep for your demo. ChatGPT is not a tool to write the demo.
Let’s dive into writing ChatGPT prompts to prep for your demos There’s no “right” way to write a prompt, but there are certainly elements you can add to your prompts to make them more impactful. Here are a few things to consider when writing your prompts:
Don’t ask ChatGPT: “What are the benefits of my software?”
Why this is an ineffective ChatGPT prompt: It’s not specific. There’s no audience identified, making it difficult to specify benefits.
Instead ask ChatGPT: “I am a sales engineer preparing for a demo of cybersecurity software to a procurement team in the healthcare space. I am well versed in cybersecurity but my knowledge of healthcare, especially procurement in healthcare, is limited. What 3 areas should I focus on for my demo?”
Why this is an effective prompt for using ChatGPT to prep for your demo: It identifies your role (sales engineer), the intent (software demo), the software (cybersecurity), the audience (procurement team), and the sector (healthcare). And by specifically addressing knowledge of the space, ChatGPT would know how much detail to give (a lot on procurement and healthcare, not a lot on cybersecurity).
The more the AI understands about why you’re asking a question, the more the response will be tailored. Providing information like your understanding of the subject, how you intend to use the information, or the intended audience of your answer can all help elicit a more meaningful response. You may want to tell ChatGPT you’re prepping for a demo and not writing a blog post. Or you may want to tell it your demo is for IT, not business users.
Don’t ask ChatGPT: “What are the top 3 features of my software?”
Why this is an ineffective ChatGPT prompt: There’s a bit of specificity – top 3 features – but that isn’t enough to get a meaningful answer.
Instead ask ChatGPT: “Summarize the top 3 features of my software for a procurement audience and present them in bullet point form. I’d like to use this for a leave behind email, so keep the tone professional but light and write at a 6th grade level or less.”
Why this is an effective prompt for using ChatGPT to prep for your demo: It identifies how many results to return (3), the audience (procurement), the form (bullet points), the use (leave behind email), the tone (professional but light), and the complexity (6th grade level or less).
Anyone who has built or interacted with a database is likely familiar with the phrase “garbage in, garbage out.” Your prompts probably aren’t “garbage,” but the responses will always reflect the prompt. Broad prompts, for example, will return broad results. It may be better to say “top 3 points” to limit and prioritize what it returns. Or, in some cases, you might want a comprehensive response, in which case you can ask for one. You can also ask for a specific format, such as paragraphs or bullet points. When using ChatGPT to prep for your demos, the key is to be specific in what you want. After all, it’s smart, but it’s not a mind reader. Yet (cue the ominous music).
Don’t ask ChatGPT: “What do you think I should show for a cybersecurity demo for a procurement team? Give me bullet points.”
Why this is an ineffective ChatGPT prompt: It has context (cybersecurity demo for a procurement team), it has specificity (bullet points), but it lacks clarity. It’s not only a bit vague, it’s also framed in a way that asks for a subjective answer (what do you think?) without any “guardrails” around the subjectivity.
Instead ask ChatGPT: “I’m looking for an objective analysis. I don’t know much about what’s important to procurement teams, and I’m deciding what features to show. Here is my potential list: [Features]. Based on this list, what do you think are the top 3 features for the demo? Provide pros and cons of those features, the rationale for why you didn’t select the remaining features, and identify any assumptions you made.”
Why this is an effective prompt for using ChatGPT to prep for your demo: I’m still asking for an opinion, but I’ve given a significant amount of information narrowing the scope of the opinion. I’ve said that I want an objective analysis (theoretically all analysis from ChatGPT is objective, but its output can certainly be influenced by the prompt so it’s not bad to specify); why I want that analysis (I don’t know what procurement teams care about); the list of potential feature; specific requirements for the output (pros and cons); more context around why the other features weren’t selected and any assumptions made – which will help me assess whether or not I agree with what ChatGPT said.
Your ChatGPT prompt when prepping for your demo may ask for an opinion. Sometimes I do, but typically after offering my own opinion or hypothesis. For example, after creating an analogy I’ve asked ChatGPT “I’m concerned that this won’t land with a particular audience – what do you think?” I know the answer is subjective. I also know not to take ChatGPT as the final arbiter of my decision.
It’s easy to think of conversations with ChatGPT like conversations with a human. After all, there are plenty of humans with whom my only interaction is texting. But Turing test (an experiment to see if an AI can be confused for a human) issues aside, don’t forget that ChatGPT is software. And while its answers are “subjective-like,” those answers are based on patterns in the training data. Those answers are not based on personal experience or human emotions.
And yes, I actually considered whether or not I was going to hurt ChatGPT’s feelings in the previous paragraph. I for one welcome our new computer overlords.
When I first started using ChatGPT, it would often take me 5 or 6 versions of a prompt to get the answer I was going for. I’ve gotten better. I can usually get what I’m looking for in 1-2 prompts, but not always. As you use ChatGPT to prep for your demo, you may find it’s not giving you the information you want. In that case, you can try giving it an answer and reverse engineer the prompt. You may not get the exact prompt, but you’ll find you get a whole lot closer.
When used properly, generative AI applications like ChatGPT can be incredibly helpful when prepping for your demo. To be as effective as possible, keep context, specificity and precision in mind. And experiment – try different prompts and see what happens.
What are some of the tricks you use for your prompts? Let us know in the chat. And if you want to give you sales and sales engineers the skills they need to drive serious value in customer conversations, contact us.