Artificial Intelligence is a set of systems that aim to enable machines and software to perform human tasks. Although there are many ways to apply it, artificial intelligence is generally capable of performing very specific actions based on a description, which is known as a prompt.
We explain exactly what the prompt is and how it can be useful when using AI applied to marketing.
What is a prompt in AI?
In AI, a prompt is a set of instructions given to artificial intelligence to get it to perform a specific action. That is, a prompt is an instruction from a user to an AI through a dialog box to get it to perform a specific action.
What you need to keep in mind is that, to achieve the desired result, you must be clear and concise in your instructions and avoid ambiguities or double meanings. Furthermore, the more information you provide, the better the response will be.
Artificial Intelligence isn’t capable of reading our minds, so explaining very specifically what we want will prevent the model from having to guess information and going down paths we don’t really want.
How to write a good AI prompt?
To write a good prompt, in addition to making it clear, concise, free of double meanings, and as complete as possible, we must try to keep the information as structured as possible. This way, the AI will understand it better, just as a person would.
Put yourself in their shoes for a moment. When do you understand information best? When it’s structured, well-organized, and divided by content or topic, or when it’s all mixed up? The same thing happens to Artificial Intelligence.
Follows the ASPECT methodology
The ASPECT methodology (acronym for action, steps, persona, examples, context, constraints, and templates) is one of the best practices for writing AI prompts and will help you write well-structured and organized prompts that are more understandable for this technology.
This methodology consists of dividing the prompt into 7 parts, in each of which the AI is provided with a type of information.
1. Actions
First of all, the prompt should tell the AI what action you want it to perform at a general level and for what purpose. It’s important that this first part be more general and not go into detail, as we’ll dedicate that to the next section.
An example of what you could include in this first part of the prompt is: “Write an email inviting the contacts in my database to a webinar on how to improve organic content on social media.”
2. Steps
Next, the prompt should go into more detail and tell the AI the process it should follow to perform the action we’ve previously indicated. In other words, you should establish the step-by-step process you want it to follow.
Continuing with the previous example, this is what you could add to your prompt: “First, analyze the script we’ll follow for the webinar to get an idea of what will be discussed (you can attach the script along with your prompt). Then, write the email that explains the webinar’s topic in general terms to the audience, but doesn’t reveal anything specific. It should encourage the audience to sign up for the webinar and have a typical email structure. Finally, review the text and make sure it maintains an informal yet professional tone.”
3. Role
After indicating the step-by-step instructions, your prompt should specify the role you want the AI to follow when performing the actions you’re instructing. This will help the model assume a specific role that will help it perform the action in a more specific way, which will bring it closer to the expected result. Telling it that the action will be performed by an expert isn’t the same as telling it that someone with no knowledge of the subject will perform it.
In our example, this is what the prompt might contain: “I want you to perform this task as if you were an expert from a marketing agency. However, the email will be signed by the agency as a whole, not by a specific professional.”
4. Examples
One of the most common mistakes when writing a prompt is not including practical examples of what we want to achieve. This type of information is really valuable for AI, as it helps it more accurately understand what you want.
If you’ve previously performed the action you’re asking for or have found some examples that fit what you want, let the AI know so it can replicate it as accurately as possible, but adjusting it to your previous specifications.
5. Context
This is one of the most important parts of your prompt. It should provide context for your situation and needs. In other words, it should include details that are important to consider when taking the desired action.
For example, in the imaginary case we’ve been following, it would be important to provide information about your past results by sending webinar invitation emails. Tell the AI what your strengths have been (so it can replicate them) and your weaknesses so it can try to improve them.
6. Restrictions
Something very important about writing an effective prompt, and one that not many people do, is adding constraints. We must tell the AI not only what we want it to do, but also what we don’t want it to do. This way, we further limit its range of action so that it acts more specifically.
For example, when writing an email, we can tell it not to use certain words or to limit it to a certain word count. We can also tell it not to reference certain elements.
7. Template
Finally, most AI tools allow you to upload documents, which we can also use to complement the prompt. In our email writing example, we can upload a document with a template to guide the model in composing the email.
Of course, it’s not enough to just upload the document; we’ll need to indicate in the prompt that we’ve done so and how we want the AI to use it.
Prompt optimization tip for accurate responses
Everything we’ve told you in this article is part of what’s known as prompt engineering. Prompt engineering is the practice of designing and structuring specific instructions for generative AI with the goal of obtaining responses aligned with our needs.
As a complement to the ASPECCT methodology, in this section, we want to give you a very useful tip to improve the results of your prompt, and that is the use of the chain of thought or reformulation technique.
This technique involves asking the AI to improve or reword it as many times as necessary once it has generated the content you requested with your prompt. It’s very common that, despite having indicated exactly what we want, there are details in the result that don’t match what we’re looking for. To achieve the most perfect response possible, it’s a good idea to give it instructions afterward on what points to improve.
Finally, once you’ve identified which prompts have been useful to you during the reformulation technique, we recommend including them in your initial prompt. This way, you’ll improve it and create a much more useful prompt.
Below are some examples of the best prompts you can use during this reformulation phase.
The ASPECCT methodology applied to digital marketing
We’ve previously shared a general overview of the ASPECCT methodology, but in this section, we want to tell you how you can apply it to achieve different goals within your digital marketing strategy.
As you can see in the image, there’s an important change. One of them concerns the “Ideal Client.” As you know, in digital marketing, it’s very important to always consider the target audience when developing a strategy. Therefore, it’s important that in your prompt, you tell the AI what audience you want it to target. This doesn’t mean we no longer have to tell it what role it should follow; it’s important to continue integrating this into the prompt, but we must complement it with the audience we want it to target.
Applying this structure and writing prompts this way is one of the best tips we can give you on how to improve your use of AI in digital marketing. We often encounter clients who don’t use AI because they don’t get good results with it. And most of the time, the reason is precisely that they don’t write prompts well.
