New technologies are rarely understood & AI is no exception

New technologies are almost always misunderstood. Even when society senses something big is coming, it is rarely prepared for the real uses, benefits, and pitfalls that may follow.

Early depictions of the internet showed teenagers dialing in from their parents’ home phones, then instantly hacking nuclear silos, altering the weather, or getting rich overnight. We now recognize how absurd those predictions were.

We also consistently underestimate the time, money, and experience required to achieve practical results with any new technology. This is especially true with today’s artificial intelligence. Humans must still learn how to clearly communicate ideas to these systems — and how to turn the output into something useful.

The steam engine was conceptualized and physically built long before it powered coast-to-coast travel. The real delay wasn’t mechanical; it was in figuring out its formidable real-world applications.

From the technology’s perspective, AI must first “understand” what is being asked of it. That understanding is far less advanced than most people assume. Neither humans nor the systems fully grasp their own limitations yet.

Millions of miles of phone lines, security protocols, and disconnected machines stood between a teenager in Boston and a missile silo in Montana. Similarly, a massive amount of metallurgy, engineering, and control was required before steam could reliably turn giant locomotive gears. Technology only becomes truly useful after we master the directing of it.

Humankind faces parallel limitations now with AI: We must first clarify and organize our own thoughts, then translate them precisely for a completely new system to be able to understand. After that, we still have to learn how to interpret and apply the results. If our intentions are specific, the AI must be carefully guided to think about the problem and present information in a usable way. This all takes a great deal of time to refine and learn. Not just for us, but for it.

All human innovations ultimately require human direction. With AI, we must explicitly tell it how the outcome should be applied. Our own assumptions about AI may be the biggest barrier to its timely and practical adoption.

Lessons for the imprint industry

In the imprint industry (screen printing, embroidery, engraving, vinyl, DTG/DTF, etc.), we have always received poor-quality graphics from clients, and it requires time to clean them up — removing backgrounds, vectorizing raster images, and tailoring them for production. AI image generators are powerful new tools,
but they do not change this fundamental workflow.

Current AI tools can generate a graphic from a vague idea with shocking ease. The real challenges emerge when adjusting the output for production processes, while also managing customer expectations. It’s like playing a high-stakes game of “telephone” across multiple languages. Literally “language models.” A customer describes a concept (often referencing an existing logo or mascot), you understand it your way, then you prompt the AI, or they prompted it before sending the idea to you, and the result rarely matches what’s needed for production, let alone expectation. Language models are only part of the problem. Understanding by all involved (including the AI) is another large part of the problem.

Full stop: I am not against AI. It is an amazing new tool with near-limitless potential. What I am saying — without hesitation — is that the time required to train these systems with the right prompts, data, and description so they consistently deliver your needed results is still enormous compared to using proven, established methods.

In short: Learn the basics of graphic design. You don’t need four years of art school or expensive new hardware. With a modest learning curve and reliable tools, you can often produce what you need more efficiently than wrestling with today’s AI.

I’ve worked with hundreds of imprinters who struggle with AI when a simple, proven workflow would solve their problems so much faster. It’s no surprise that AI can also stand for Adobe Illustrator — in my opinion, a second-best option for our industry. The true champion, however, remains CorelDRAW. For decades, Corel has tailored its software to the specific needs of engravers, embroiderers, and millions of printers worldwide. Features like native Pantone support, superior vector tracing, and industry-specific AI enhancements (such as background removal) deliver more direct, production-ready results for us.

Too often, people chase AI simply because they were told to. Consider the source: Is it a hundred-billion-dollar institution that is massively invested in it taking you years of formal training to learn? Someone heavily committed to one narrow approach who might resist better alternatives? Or a veteran business owner who knows how to get efficient, high-quality output from CorelDRAW?

AI is years away from matching the hype — not because it lacks raw power, but because it and we still need to learn what “metals” will best forge and direct that power. Like a steam engine. Embrace the technology (it’s too late not to) but treat it as exactly what it is: a powerful new tool that demands significant time and learning to yield useful results.