It isn’t the technology. It’s what we think the technology is.
Last month, I had an enlightening meeting with a very large hardware tech company — over $500 million in annual sales — to talk about AI. In the room were the head of product development and two of their senior dev managers. They asked me what I was seeing in the market, and when I said that everyone I talk to is struggling to find an ROI in their AI implementations, every head on the video call nodded.
That nod matters. In nearly every conversation I have with business owners about AI, the same thing comes up: they can’t justify the return on investment. If a $500 million company is wrestling with the exact same problem you are, you’re not behind. You’re normal. And that shared struggle is actually an opening — a chance to figure out where AI genuinely fits and where it doesn’t.
The current LLM models have been out for nearly four years now. The idea that we should still be struggling to find ROI is disconcerting. We are in an AI bubble — but not because large language models lack value for businesses. They offer real, significant benefits to organizations and the people who work in them. The problem is that most of us don’t actually understand what these tools are, underneath all the hype. And until you do, the ROI stays out of reach.
The objection I hear most
“AI is new and moving fast — it’ll take time to see real ROI.”
That’s true, as far as it goes. But there’s usually an unstated belief hiding underneath it: the assumption that ROI means AI fully replacing a human — doing the job cheaper, faster, and better, with no person in the loop.
If that’s your expectation, you’re going to be waiting a very long time. Today’s LLMs can’t do that, and honestly, they probably never will. The path to ROI isn’t waiting for the tool to become a person. It’s understanding what the tool does well, and what it doesn’t.
Try a different name
In a fantastic article in the Yale Review, Melanie Mitchell points out that back in the 1950s, as the idea of AI was taking shape, several scientists felt the term “Artificial Intelligence” was misleading — too anthropomorphic, too human-sounding. The name they wanted instead was “Complex Information Processing.”
We know which one won. And that single naming choice has caused an enormous amount of confusion.
Here’s a trick worth trying: every time someone says “AI,” swap in “Complex Information Processing” in your head. It’s remarkable how quickly the fog clears.
Because that’s what these tools actually are. LLMs are predictors — able to, you guessed it, process complex information. That alone makes them genuinely powerful, and it’s exactly why they’re so good at coding. What it does not make them is intelligent.
In his book Algorithms Are Not Enough, Herbert Roitblat describes how intelligent beings create tools to extend their own intelligence. Humans have been doing this for thousands of years — mathematics, algebra, calculus, the sextant, the computer. AI, or Complex Information Processing, is simply the newest one. It’s a powerful tool built by intelligent agents. It is not an intelligent agent itself.
Why so many companies are failing at this
The reason so many organizations haven’t found ROI is that they’re implementing AI on the mistaken belief that it’s intelligent and capable of true agency. It isn’t.
What companies should be doing is finding the specific spots in their workflow and technology where AI genuinely fits, and then integrating it with their existing stack — and, most importantly, with their employees.
Instead, many take the push-down approach. Believing AI is intelligent enough to replace people, they force it onto their teams in ways it doesn’t actually work, and tell those same employees it’ll eventually take their jobs. Then they act shocked — shocked! — when employees push back with genuine hatred for a technology that isn’t working for them.
Of course they do. Nobody embraces a tool that’s being aimed at their livelihood while failing at the task in front of them. Instead of using AI to make people’s work better, easier, and more fulfilling, these companies used it to threaten people. That kills ROI before it ever has a chance to show up.
There’s a better way
Integrate AI with your people and processes. Don’t bolt your people and processes onto AI.
That’s where the real return lives — Complex Information Processing solving very specific problems, and delivering value not just for a handful of employees, but for the entire organization.
And for cost savings, regulatory control, and better performance, the answer is Private AI — open-source LLMs that you own and control. I believe Private AI is the future. Why? That’s exactly what I’ll dig into, with real examples of doing this the right way, in the articles and videos to come.
For now, let’s further the conversation of sanity.
Trinsic Technologies helps organizations put AI to work where it actually fits — securely, privately, and in a way your team will thank you for. If any of this resonates and you’d like to talk through where AI belongs in your business, reach out. It’s what we do.
