AI Is Amazing… and Also Exhausting: A Graphic Designer’s Perspective

Where AI Saves Time — But Still Needs a Human in the Loop

Is anyone else getting burned out on artificial intelligence (AI) yet? I’m not talking about the technology itself, which I find very useful. I’m talking about all the hype: news stories, tool-tip pop-ups, commercials, and the constant prompts. Will AI replace your job? Try our new AI assistant! Will the AI bubble burst? Switch to AI mode! At this point, I’m ready for an AI bot that can remove all of it from my line of sight!

I’m definitely not anti-AI, as I pretty much use it every day in some capacity. Many of our instructional designers find it very helpful as well (here’s a great blog by my coworker on how to harness its power most effectively). As for the larger implications and conjectures, those are better left to people far more knowledgeable than I. Still, I can’t be the only one who has experienced both the benefits and quirks of using this technology.

In that spirit, I thought I’d share some of the requests I get as a graphic designer and media developer at Innovative Learning Group (ILG), and how AI is helping — and, in some cases, still falling short. Many requests have little twists: some seem like major undertakings but take mere minutes, while others start as simple tweaks and end up consuming large portions of my day.

“Just one tiny audio change…”

A big part of my job is making videos for a wide range of clients and industries. Whether it’s a quick motion graphic, an explainer-style animation, or stitching together live clips to tell a compelling story, I’m constantly jumping between tasks. The advent of cheap and reliable AI audio has made adding voiceovers a breeze. As the tools continue to improve and more features are added, producing high-quality videos has become much easier and less time-consuming.  So, when I get a request to change one tiny word, you’d think it would be a quick task. Unfortunately, that’s not always the case. For example, while working with a Medtech client recently, they noticed the word “veterinary” was often mispronounced (they use vet-er-in-ary, not vet-tren-ary). Easy to fix? Absolutely! Quick fix? That depends…

Oftentimes, depending on the tool, AI pronunciation can be tricky to control. Some tools, such as WellSaid, have amazing phonetic controls, while others (Articulate Storyline) currently offer none. This can mean spending large chunks of time changing spellings, rendering, listening, adding punctuation or commas, rendering again, listening again, or even stitching together separate pieces of audio if the word won’t adjust correctly in context.

And even once the audio is corrected, new questions pop up: Does the word appear in multiple spots? Did the timing change, requiring adjustments to the animations? Have you already added closed captions, and do they now need to be updated? How about lip-syncing? How large is the video, and how long will it take to render?

It’s easy to see how a “tiny” change can balloon into many hours.

“Remove all labels…”

When I get a request to edit photos or diagrams, more often than not I’m dealing with a lot of “flattened” JPEGs or PNGs rather than an actual editable source file. While Photoshop usually makes quick work of chopping up pixels, it can sometimes be more difficult than it first appears; however, this is rapidly changing thanks to AI.

A colleague of mine was recently asked to remove text labels from several medical diagrams. It sounds simple enough, but the labels were a bit clunky and laid out on top of odd shapes with gradients; some were connected with thick lines that included drop shadows. This isn’t always easy to mimic or create from scratch once the labels are removed.

However, with the arrival of Photoshop’s “Generative Fill” AI tool, filling in the gaps and replacing  what’s underneath all those labels has become remarkably quick. With a simple prompt — and only an iteration or two — the labels and directional arrows disappear within seconds, leaving behind a relatively accurate representation of what should be in its place.

“Generate closed captions…”

Here’s a task I really, really wish AI could take over entirely (and almost has!). The majority of  tools I use nowadays for audio and video (After Effects, Premiere, Storyline, Rise) have built-in “generate closed captions” functions that work magnificently. AI lines up all the chunks of text  with the audio, spaces them appropriately, and displays the right amount of text on screen — it looks perfect! Oh wait… it spelled Sara incorrectly. Oops! It added a period where it should have been a comma. Hmm, actually that’s a weird line break halfway through the audio. Why did it spell the company’s name that way!? And on and on it goes.

This means I still have to comb through the entire set of captions, checking against the script to make sure no small errors have been made. It definitely saves time versus starting from scratch, but it’s more of a “you set ‘em up, I’ll knock ‘em down” situation rather than AI taking care of everything for me.

These are just a few things to keep in mind when developing custom e-learning courses or any kind of digital collateral. And of course, with the rate at which the technology is changing, it’s almost impossible to keep up. However, it’s important to stay current on the latest developments without getting overwhelmed by all the noise. There’s good in all the ruckus somewhere — I promise.

At the end of the day, AI has absolutely earned its place in my work — it just hasn’t earned the right to be left unsupervised. For now, it’s a powerful collaborator that works best with a human eye, a little patience, and a willingness to dig into the details. Maybe that’s the sweet spot: letting AI handle the heavy lifting, while we focus on the judgment, nuance, and creativity that still can’t be automated away.