Come one, come all! Everyone is a content creator now. Human, android, vocaloid, even your stay-at-home AI girlfriend! 

Okay, I’m kidding. But if prompt engineering is the new must-have skill for the workplace, then we all are writers now.

The barrage of reactionary media (and hype) around ChatGPT has begun to quiet. But in its place are more thoughtful discussions on how most of us will, or should, adapt to the presence of artificial intelligence as a new class of tools. 

And then there’s Reddit: the internet’s foremost mix of measured considerations beside elitist or alarmist takes on generative AI, depending on your preferred cup of joe.

I’ve been a casual reader of AI subreddits for a while. One topic has been especially grating on the community: prompt engineering, the latest in you-must-upskill rhetoric. Some claim prompt writing isn’t a skill, it’s basic English grammar. Others claim the opposite. Following the upskill hype around becoming “data driven” and having an “Agile mindset,” it would be easy to cast “prompt engineering” into a similar category. But I don’t agree that prompt engineering is just a trendy word for jotting down instructions. 

It’s tempting to discuss people who use English at work when its not their first language, or people who approach AI chatbots with all the same manners of a human conversation in the South (*raises hand*). But the root of the issue is that prompting an AI is writing. And writing well has not only never been easy, but what qualifies as good writing varies significantly depending on the context. 

Prompt writing is a creative skill 

Miro turned the whiteboard into a multi-billion dollar software success by creating a digital tool that provided humans – as individuals or in groups – with an intuitive, useful medium to think with. Especially noted for its collaboration functionalities, Miro has become a staple in the workplace. Skillsets around using Miro, especially on how to be an effective workshop or meeting facilitator using Miro, have become prized. 

In the same way, generative AI is a tool – and new skills are going to be expected by its adoption. Everyone is bracing for the radical changes to junior and mid level IC roles. Creatives, like photographers or manga artists, will be hit first. But no career is untouchable; lawyers and programmers are also facing an AI transformation. 

In the coming years, most digital media and content production will contain – to some degree – machine assistance. With new workflow processes, new creative feedback loops between human and AI, there’s an opportunity for endless subsets and endless variations. I share the belief that AI will not replace humans, but humans who can utilize AI will replace humans who can’t (or won’t). 

Under such a rapid pace of adoption, though, the most critical area for anyone in any industry to “upskill” is not just identifying when to use AI at work but how to use it effectively. For many people, whether at work or at home, prompting has become their first tangible Human-AI interaction. When we talk about how generative AI is “revolutionizing” culture and work, what we really mean is that anyone can synthesize media – and this is happening through prompt-based learning. 

Yet, some workers report feeling embarrassed of making use of these tools. Asana’s Work Innovation Lab found that less than 30% of UK workers were using generative AI tools weekly, compared to around 45% in the US. Asana found that workers felt anxious about their own value, creative competencies, or whether their work will be seen as lazy or fraudulent. It doesn’t help that major companies like Verizon and Apple have blocked ChatGPT at work. 

This isn’t good for workers. While companies are blocking Open AI, they’re building internal systems integrating new AI tooling and testing their own LLMs. Meanwhile, workers are plagued with anxiety – and rightly so. Creativity is something we’ve long ascribed as uniquely human. But now machines are doing (what appears to be) creative things.

Writing well to the LLM 

For us to cocreate a world where AI is to the benefit of the workers, then – like any technological tool – workers need exposure. We need opportunities to troubleshoot, explore new working processes, trial and error. Generative AI will evolve by us interfacing with it. 

In “Why Johnny Can’t Prompt,” researchers found a tendency to overgeneralize when prompt writing. Another research project found that the “open-ended nature of text as interaction” was another barrier, as being able to input anything and generate infinite variations caused a reliance on brutal trial and error. 

I’m trying to find a sweet spot where I feel augmented by AI in the same way that I feel augmented by the internet or Miro. So far, refining my prompts can sometimes take as much time as the first step of the task I’m trying to accomplish. Beyond the brainstorming stage, I’ve found that reliance on generative AI tools can block me from entering a state of flow. On the other hand, I’ve found that a quick and lazy prompt can actually help break through procrastination or a creative block.

In order to not get lost in a rabbit hole of my own prompt-refinement, I’ve turned to tools like WhyBot to get started. I’ve also limited myself to just two or three variations before moving on and having a go myself. There are a few guides, tips, and tricks backed by research on effective prompt writing. My favorite is role-playing.

But I suspect that what makes prompting effective – as in, yielding strong results and offering a meaningful enough output to reduce the labor or enhance the creativity of the prompter – is as dependent on context as writing itself.