What AI Can't Do for Content Development -- And How to Fix It

Artificial Intelligence, and in particular the white-hot field of Large Language Models (LLMs) like ChatGPT, are clearly the wave of the future in marketing and communications.

Not only can they craft well-written articles and blog posts, but they are on the verge of taking on much of the work involved in graphic artistry, audio and video production, and even personalized interactions with customers.

Good, But Not Great

Despite these impressive accomplishments, however, they still cannot -- and in my opinion never will -- remove the human element from what is essentially a human-driven need to exchange information. And like virtually every other technology that has changed the world dating back to the first time a farmer attached a piece of iron to a wooden frame and hitched it to a pair of oxen, displacing the work of hundreds of farmhands, AI shows every indication of automating part of the communications landscape, but not all.

For those who are already using LLMs to craft content, does this seem familiar?

You tell your model to draft a response to an email, which you approve (maybe with a few tweaks) and send it out. Fine, no problem. It was only a few sentences. The same goes for a social media post, a few slideshow captions and other short-form content.

Ask it to write a lengthy blog post or an article, or a white paper, and it will do that as well . . . but not quite. The language is a little basic. It lacks flow and pacing. And it usually relies on outdated or inaccurate information, even if you tell it to use your own data.

So you ask for a rewrite, pointing out the problems with the first draft, and it will do that . . . but not quite. Maybe some of the issues are addressed, but not all. It still feels clunky, rather amateurish, and highly derivative of what's already out there. What it lacks is the kind of compelling, engaging voice needed to stand out -- and it needs to stand out because the world will soon be awash in AI-generated content.

Now, you may ask for another rewrite, and another, and another, but each time the improvement becomes less pronounced, until eventually you've spent more time trying to get a decent result out of your LLM than if you had just written the damn thing yourself.

It's All Just a Number's Game

This points to the fundamental flaw in relying on technology to achieve what is essentially a human endeavor: AI only deals with numbers, or more precisely, probabilities, not actual thoughts and ideas. Although it may seem intelligent and engaging, AI is only crunching numbers.

That means it will only give you what it calculates to be the right words based on what its algorithms recommend, which are influenced only by what you specifically ask it to create. This makes it very effective at maximizing SEO, fine-tuning content to finite market segments and ramping up the volume of communications.

But it cannot provide a written document that is enjoyable to read -- and this is the key to elevating your marketing strategy from merely saturating media with junk content to actively engaging with key markets in order to build relationships and brand loyalty. AI can help do this in other ways, such as chatbots, but not by crafting well-written content.

The Benefits of Experience

All is not lost, however. Since AI can automate many other aspects of the marketing burden, including the initial phases of long-form, creative writing, it only needs a little help to bring a more nuanced, thoughtful piece to the finish line.

This is where a professional writer, preferably one with lengthy experience in whatever field you happen to be engaged in, is essential. Once your model has taken the piece as far as it can go, a talented writer can give it that extra punch in the following ways:

• Provide engaging anecdotes and turns of phrase. AI can craft sentences with the best of them, and organize thoughts in a reasonably intelligent fashion, but it cannot fully engage with a reader because it does not really know what it's doing, or why. Again, it is only crunching numbers.
• Provide smooth transitions between thoughts. AI is very good at listicles and other forms of stilted presentation, and this can be very effective for marketing purposes. But these can be tedious to read, especially if they are everywhere.
• Drill down to what truly makes your product or service unique. AI algorithms are trained to provide the most common responses to any given query. You don't want your message to be common; you want it to be unique.
• Properly utilize idioms, similes and other turns of phrase. Common versions of these quirky elements of language can be housed in a database, but AI still has trouble figuring where, when and how to use them effectively.
• Incorporate irony, sarcasm and humor. Again, AI is very literal. When the actual meanings of words or phrases differ from the intended meaning, they don't get it.

Realize that the key benefit to artificial intelligence is its capacity to learn, but this is not nearly as profound as the tech industry would have us believe. After all, it wouldn't be the first time they've over-promised and under-delivered. As time goes by, we can expect AI to get better at all of these things, but it will get better for everyone. That means you'll need to find increasingly creative ways to make your content stand out in a market that is flooded with generic, derivative pablum.

An experienced wordsmith is the solution to this problem. With AI crafting the basic message, all it needs to stand out from the rest is a strong voice and a true understanding of the reader's goals and challenges.

Chances are, you already have someone like this on your staff. If not, there are plenty of freelancers who work fast, and at a fraction of the cost of writing from scratch -- although in many cases it might be better to do just that. It took me about 20 minutes to write, edit and proofread this article. I hope you found it interesting, engaging and thoughtful.