Using AI Wisely in Communications and Design

Generative AI is a miracle technology which can increase productivity and quality in your organisation, today. But it’s important to understand how it can really help, what’s simply fun to play with, and what could do more harm than good.

What’s wrong with AI-generated content

Right now, large sections of AI-written text suffer from several subtle but important issues:

• Lack of substance. AI is particularly good at seeming like it’s saying something interesting, while actually being extremely vague and basic.
• No point of view. AIs have no values or perspective. They don’t care about the topic – and it makes dull reading.
• Generic and phoney style. AI content has a peculiar tone which, once you notice it, is easy to spot: excruciatingly balanced, servile and chirpy. Your content should sound like you, not a robot.
• Bias. Even setting aside harmful bias (like gender stereotying), AI was trained on the Internet, which is dominated by American content – just one particular flavour of human expression.
• Inaccuracy. AIs can produce falsehoods that sound true. They should never be the source of your information.

Ultimately, the upshot of these issues is that the content has no value. It may indeed discuss some topic, but in a painfully generic way that brings no new information, viewpoint or clarity to the reader.

And that’s a problem for your brand. You want your audience to engage with your content, feel enlightened, informed or inspired, and leave with a lasting impression. But that’s not going to happen with material that’s essentially identical to 10,000 other low-value blogs.

Of course, with extensive prompting, AI can be made to avoid some of the above issues. But you will likely spend more time getting AI to write well than it would take you to write better yourself.

Let’s compare

Consider the opening lines of Eric Reis’s much-revered business strategy book, The Lean Startup:

Stop me if you’ve heard this one before. Brilliant college kids sitting in a dorm are inventing the future. Heedless of boundaries, possessed of new technology and youthful enthusiasm, they build a new company from scratch. Their early success allows them to raise money and bring an amazing new product to market. They hire their friends, assemble a superstar team, and dare the world to stop them. Ten years and several startups ago, that was me, building my first company. I particularly remember a moment from back then: the moment I realized my company was going to fail.

Notice:

• The author’s personality shines through.
• It tells a story, because it’s his real story.
• The nuance is that the story is both romantic and common.
• It draws you in.

Now let’s see ChatGPT’s attempt:

I once believed that a well-crafted business plan was the key to startup success. I spent months refining financial models, mapping out market strategies, and perfecting pitch decks. And for a while, it worked. Investors nodded approvingly. Customers showed early interest. Growth seemed inevitable. Until it wasn’t. Despite the initial traction, my startups floundered. Assumptions I had treated as facts unraveled. Market conditions shifted. What had once felt like a solid plan became a set of constraints, trapping us in a failing strategy.

Notice:

• It sounds like nobody in particular.
• It’s just the basic ideas phrased in first person repetitiously.
• There’s no depth or nuance.
• You’re already bored.

How you should be using AI for content

Humans should write your content – but AI can be a valuable tool to aid the writing process. If you’re sitting down to write, use AI to help you:

• Brainstorm initial ideas, which might nudge you to come up with something better.
• Structure your ideas. AI can suggest an overall outline.
• Get past writers’ block. What could you discuss next? Is there a different angle you could take?
• Remember what’s on the tip of your tongue. AI can usually work out the idiom, news story or film reference you’re after with just a few details.
• Come up with examples to illustrate the point you’re making.
• Reword sentences you’re struggling with. But be careful before letting AI rewrite whole paragraphs for you (e.g. “make it more professional”, which usually results in “make it more generic”).

We can’t all be Shakespeare, of course. If you’re going to use AI to rewrite anything, just don’t forget the golden rule: never use any AI-generated text without reading every single word of it first. Ideally, edit it further to make sure it still sounds like you.

Used wisely, AI will increase both the quality of your writing and the speed at which you produce it.

What’s wrong with AI-generated images

Right now, AI-generated imagery should not be used without modification, as it suffers from several issues:

• Nonsensical elements. This can be obvious (like a dog with four eyes) or subtle (like dog with cat’s eyes).
• Generic look and feel. AI images have a peculiar quality which, once you notice it, is easy to spot: overly clean, pleasing, and often cartoonish. This ‘AI look’ has a bad reputation.
• Bias. Even setting aside more harmful bias (like favouring photogenic white people), AI leans toward predictable visual concepts and styles.

Ultimately, AI-generated imagery isn’t creatively intelligent – it just gives you what you ask for. And this is a problem because, if you think about it, creative intelligence is what makes graphics worthwhile.

Without creative intelligence, a logo isn’t a brand, it’s just a graphic you keep using; an illustration doesn’t bring content to life, it just puts some coloured shapes alongside it; and a layout doesn’t help communicate your ideas, it’s just an arrangement of elements so they don’t overlap.

Let’s inspect

Here’s ChatGPT’s attempt to depict a realistic scene.

Notice:

• The train track is partially duplicated onto the platform.
• Several people are little more than abstract shapes.
• The structures of the station (support beams, etc.) make no sense.
• There’s a sign (with gibberish text) floating in mid air.
• Some of the screens are bizarrely cloned.
• It’s 95% stereotypical male businessmen.

How you should be using AI for graphics

For now, making use of AI imagery requires some design skill – because you have to supply the creative intelligence, and you’ll need to edit the image.

• Make your requests to AI as specific as possible. Don’t just ask for “a woman standing in a park”, describe the scene in detail. This will make your image more unique, relevant and impactful.
• Ask for non-photorealistic styles. AI’s attempts at realism are prone to failure. Instead, ask for pencil sketch, collage, digital art, etc.
• Fix issues manually such as impossible elements, weird details or strange distortions.
• Know what the tool is good at. All current tools produce terrible logos. But with the right approach, they can produce decent illustrations.

Nothing AI currently produces is as good as human-generated graphics – but then, AI hasn’t managed to make money grow on trees yet, has it?

If you are willing to pay for graphics, great, but beware that a lot of stock image catalogues now include AI-generated images. Paid-for AI imagery both rips you off (it cost next to nothing to create) and rips artists off (whose real artwork, available for the same price, loses out on a sale).

Sadly, far too many companies have already begun deploying AI en masse, publishing mountains of 100% AI-generated slop.

And it’s creating a real societal problem. Increasingly, we can’t trust anything we read or see. The Internet is filling up with content that has no value. And the media we are exposed to is decreasing in variety, while the voice of real human expression is drowned out.

For now, both content and imagery produced by even reasonably-skilled humans is far superior to anything an AI can generate. But whether that changes or not, let’s use AI to create value, not noise.

Not a word of this article was written by AI.

Are you human?

Prove it by letting us know your thoughts over on: