In recent months, we’ve seen a wave of new uses for artificial intelligence. Various industries, from healthcare and finance to manufacturing and logistics, are using AI in innovative new ways. But, when it comes to using AI for creative work, the situation is a bit more delicate.

We’ve written plenty about uses of generative AI for companies, including a client case study created with generative AI and creating AI videos that complement your content, concluding that AI isn’t where it needs to be to create compelling creative content from scratch…yet.

In both of those posts, we recommended a light touch that balances the time-saving capabilities of AI with the creative mindset that only a human being is capable of. But what exactly does that look like in practice? We put together a few do’s and don’ts on using AI for creative work.

How to use AI for work (And how not to…)

Do: Use AI for brainstorming and out-of-the-box thinking

If you’re looking for a general overview of a topic, AI tools like ChatGPT can be really useful. Beware of AI bias – algorithms aren’t as neutral as you might think! – but you can also take advantage of that and use it to generate viewpoints on a topic that are different to your own.

Example: 

Let’s say we’re writing about a topic that’s been covered time and time again. The prompt above generated ten different angles from which we might approach the subject. Just bear in mind that, given the nature of gen AI, these ideas might not all be as original as they claim to be.

Don’t: Rely on AI to creative completely original content 

Although written content and images generated by tools like DALL·E, Midjourney, and Claude can be impressive, they might also be thought of as a patchwork quilt stitched together from different sources. Unfortunately, such tools may include copyrighted content* in their output. 

Example:

Infringement and unlicensed content in training data is proving to be a HUGE problem in AI. The image above is part of a lawsuit against Stability AI filed by Getty, alleging that their images were used in training data without a proper licensing agreement in place.

Do: Have an understanding of the data your tool was trained on

We’ve previously written that AI is only as good as its ingredients, and that applies not just to prompts but to training data as well. Just like you wouldn’t trust a doctor who hasn’t been to medical school, you shouldn’t rely on a gen AI tool that hasn’t been trained on appropriate data.

Example: 

Microsoft’s chatbot Tay, launched in 2016, was trained on data from conversations with X (then Twitter) users. Perhaps not such a great idea in retrospect. Less than 24 hours after its launch, the bot was shut down when it started posting racist and sexist content…

Don’t: Publish gen AI content without checking it first

When forced to choose between generating a response and saying “I don’t know,” AI will always opt to generate something. Unfortunately, there are instances where that “something” is outdated, spurious, or erroneous. And that can have disastrous results.

Example: The number of AI disasters is already mounting up, but one notable example comes via McDonald’s. In a viral TikTok post, the brand’s AI-assisted automated system reverted to adding McNuggets (260 of them!) to someone’s order when it failed to understand a command.

Do: Work on techniques to refine AI output

Although the risk of AI hallucinations looms large, it can be mitigated by providing your chosen tool with context and/or examples and using caution when prompting. Without proper prompting, artificial intelligence is very likely to “play it safe” rather than create something specific or useful.

Example:

There are tons of guides for prompting out there, with advice varying slightly depending on the desired nature, style, and content of your output. Be sure to check out our post on making the most of prompt engineering for more thoughts on this delicate process.

Don’t: Assume that AI can replace creative teams

There’s no denying that AI tools can be helpful when it comes to research, creating outlines, or generating ideas. But as long as they lack the capacity for independent thinking, they can’t replace humans. We’ve seen AI in its current state likened to a magic trick and, the more AI output the general public is exposed to, the more people are starting to see behind the curtain.

Example:

Many businesses now use tools like QuillBot’s AI detector to identify content generated by artificial intelligence, with the tacit assumption that its quality will be poor compared with a human writer’s. The kicker? It’s powered by AI. The bots are officially ratting out the bots…

The future of AI for companies

McKinsey reported in October that nearly three out of every four businesses are using artificial intelligence in some capacity or another. Using AI to spark creativity or, say, repurpose your content is a smart move in 2024. In fact, businesses that fail to do so risk being left behind.

But remember, there’s a big difference between using AI for work and overreliance on it. As we’ve seen above, gen AI output is unlikely to be polished enough that you can just hit Send or Publish without tweaking it first or repeatedly reworking prompts to hit every point in the brief.

Artificial intelligence is not yet, for example, great at creating pictures with words in them – always a dead giveaway that an image was created by AI. Similarly, written content produced by AI often overuses words and relies on stock phrases. That can result in writing that feels emotionless, verbose, and…not very much fun to read.

Over time, the scope of LLMs (Large Language Models) will expand and the output of gen AI tools will undoubtedly improve. We’re already seeing that happen in real-time. But for now, and for the foreseeable future, the best uses of AI rely on accounting for its many limitations.

Taking all of the above into consideration, good and bad, it’s still well worth keeping abreast of developments in the AI space. The AI Resource, our AI newsletter for marketing and content leaders is the perfect way to cut through the noise and keep up with the latest “must knows”.