Who Does AI Actually Help?
The technology works. The question is: works for whom?
By Andres Sepulveda Morales · Red Mage
Before I got laid off the last time, my manager pulled me aside and said, “Find somewhere to implement AI so we can seem more ‘tech-forward.’”
Not because there was a problem to solve. Not because anyone had asked for anything. Just so we could appear to be moving in the right direction. That was the whole brief. I tried to fulfill the request, but I knew that the end product didn’t really matter, as the organization had given up on purpose-driven work long before I got there.
I’m queer and Puerto Rican and non-binary and disabled; I’ve spent my career learning to read rooms fast, because most rooms are not intended for people like me. That’s become increasingly clear the more time I’ve spent in the organizations that prioritize appearance over depth, effort over impact.
For example, in that last workplace, a colleague once explained during a team meeting that cross-racial joke exchanges were fine as long as they were reciprocal. His Mexican friend made racist comments about Koreans. My coworker reasoned that, because he was Korean, he could return the favor—it balanced out. People laughed when he said this. Daily stand-up meetings turned into theological debates about Islam, Christianity, and Judaism. A framed portrait of President Donald Trump in fighter jet gear hung in our coworking space like a patron saint of the enterprise.
I noticed all of it. That’s what happens when you spend your career moving through spaces that weren’t designed for you: You develop an instinct for what the people around you actually value versus what they claim to. At my last job, what mattered more than using AI effectively was projecting to the world that we did. People performed innovation in the same way they performed inclusion: loudly, toward no one specific, with no one accountable for the gap between perception and reality.
The Same Failure, Two Different Places
The failure rate on enterprise AI projects is somewhere between eighty and ninety-five percent, depending on the study. The explanations are usually technical: bad data, unclear use cases, poor integration. Those may well be contributing factors, but they usually trace back to something more basic: Many companies pursue AI initiatives for appearance’s sake. It’s the same root failure as performing inclusion rather than living it out.
Remember when IBM and McDonald’s spent three years developing Automated Order Tracking (AOT) technology? When they announced their AI voice ordering partnership in 2021, they wrote in the press release that it would “tackle integrations including additional languages, dialects and menu variations.” Instead, the companies ended the initiative in 2024, and according to CNBC, inaccuracy when interpreting accents and dialects was one of the potential reasons. They had committed to solving for both, but never fully took into account who was going to be ordering food.
Now raise the stakes beyond the drive-through. Research published in Wounds International found that AI tools built for wound and skin assessment perform significantly worse on darker skin tones because training data skewed heavily toward lighter skin.
A separate study in the International Journal of Dermatology found that tattoos can interfere with a model’s ability to read a skin lesion correctly. But tattoos, skin color, hair: These aren’t edge cases but features of human bodies. For a person like me, who isn’t white and who has a panther inked across their forearm, these considerations are barriers not just to a useful technology, but also to a particular form of medical care.
That barrier seems to exist because no one in the room where the technology was made sat with the question “who is this actually for?” long enough to get a real answer. Maybe they’d say “our users” or “the community,” gesturing at an abstraction instead of thinking about an actual person in an actual doctor’s office.
Building Trust Means Asking First
After that last layoff, I started to wonder whether tech was a space where someone like me could make work that meant something to someone real. I founded my consultancy company, Red Mage, in an attempt to answer that question. Red Mage works primarily with nonprofits and community organizations to create technology that actually serves them and their needs.
I like to call my clients my “impact makers and barrier breakers.” They’re real people with real pain points and opportunities: a caseworker spending three hours a week on documentation when she could be spending that time with the people who need her. A small business owner who can’t afford to hire someone for intake but could use a tool that handles it at midnight. A community organization whose members speak four languages and have never seen outreach that felt tailored to their needs.
When you build with specific people in mind, you can make a much deeper impact with technology. You can also establish trust, something that’s badly lacking in the communities I care about most. Having been left out of a lot of lofty claims (voice systems that will work for everyone; diagnostic tools that will catch what doctors miss), they’ve earned their skepticism when it comes to big tech. Trust doesn’t transfer from one broken promise to the next.
That’s why community is non-negotiable for me in this work, and why I’d encourage you to do the same. I run Fort Collins AI for Everyone—an inclusive and diverse space for AI enthusiasts—and work with the Rocky Mountain AI Interest Group because I know what it feels like to be the person existing systems forgot. These groups were built around the premise that their members matter more than the tools they’re using. The question “who does this help?” isn’t a framework or a talking point. It’s just how the room operates, and that’s the antidote to AI slop.
AI is an extraordinarily capable tool. Whether it actually changes someone’s life for the better, that part is still on us.
Andres Sepulveda Morales is the founder of Red Mage, a technology consultancy for impact makers and barrier breakers. They are presently Treasurer of the Rocky Mountain AI Interest Group (RMAIIG), founder of Fort Collins AI for Everyone, Top 10 Puerto Rican Voices in IT & Tech (Favikon) and were named one of Contra’s Best Software Consultants to Hire in 2025-2026 for English and Spanish markets.
andres@redmage.cc · redmage.cc