A while back, I examined generative AI and asked the question many were asking: Will this take my job? The short answer is no, but that doesn’t mean my work as a content manager didn’t change because of generative AI. Now, even if I’m the only person on the content team, I have an employee in the form of ChatGPT. And like any employee, I need to manage their work and ensure it’s up to quality and done on time.
Setting the Goals
You'll need to phrase your request correctly to get the best results out of ChatGPT. Most refer to this as ‘prompt engineering,’ but it’s not drastically different than sending a brief to a freelancer. Segment your target audience, define your goal, create an outline for what content should be included, and set expectations for the tone and style you want your article to have.
Defining the target audience and goals should be done internally before approaching ChatGPT. If you ask it, ChatGPT will be happy to make suggestions for you. But ChatGPT is happy doing anything. That doesn’t mean it knows how to do it well. The same goes for the tone and style. Decide based on what you know about your target audience and what style brings out the best response from them. That’s the human element you bring to the table.
Outlining can be done with ChatGPT, however. Use it to bounce ideas and have its suggestions expand on the points you want to make. ChatGPT will give you “the internet’s average” knowledge about a subject—and that’s great!
Optimizing ChatGPT's Results
You mention what you want to produce, what phrases you’d like included, who it’s for, and what style. The most significant difference is length: If you’re looking for a long piece of text, such as an article, it’s easier to break it down paragraph by paragraph. The longer the piece, the more likely ChatGPT will “forget” one or more of your instructions. For example, you might mention several keywords you’d like included, and ChatGPT might include them at the start of the piece and then leave them out of the rest. Save time by working it step by step.
Also, take full advantage of the fact that you’re talking to a robot rather than a human being. Use short, simple instructions and separate them using a period or a semicolon. A robot doesn’t care if presented with fragmented sentences. Keep your briefs short—you’ll get instant feedback and can start the back and forth immediately, so it’s more efficient to send a short instruction first and then sharpen it through minor fixes as you go.
Protecting Yourself
ChatGPT is the type of enthusiastic employee who never asks a clarifying question. If it doesn’t know something, there’s a good chance it’ll just make something up. In the AI industry, cases like these are called ‘hallucinations’—times when the model refers to a source that simply doesn’t exist.
Let’s be frank, not everyone on the internet cares about accuracy. Some companies are fine with spreading whatever content they can produce quickly and never bother with fact-checking their information. But most of us care that our thought leadership is backed by facts, especially if you’re in the B2B space, where members of your target audience are often experts in their field.
If you don’t want your articles exposed as nonsense, you will need to check ChatGPT’s sources. This is where your Googling and your research skills shine. You’re not looking to give an expert opinion, just to verify the text you generated. Since you can generate text at speed, it’s usually faster to replace rather than correct when encountering a mistake. Look for a different factoid to include.
There is another aspect in which you should protect yourself and your company when using generative AI—the legal aspect. But that deserves an article of its own, which will be up shortly.
An Ongoing Management
It’s common to say that generative AI technology is only going to get better from here. But no matter how much better the AI is going to be, one thing isn’t going to change any time soon—at the end of every generative AI creation, there’s going to be a human who gives the prompt. And that human will need to have the skills necessary to ensure the text, image, or video they generate is up to the task. In other words, they’ll have to manage the AI in the same sense as managing any employee, except at a faster pace.
Even now, you can see examples of bad AI management. You can already feel some articles have been written by ChatGPT, maybe even some emails. In some cases, it doesn’t matter. But when it does, it matters a lot. That means that anyone who works in Marketing or Product and is using AI will need to know how to create effective, engaging, and original texts while still enjoying the benefits of speed that generative AI brings to the table.
Comments