AI in Healthcare: Lessons for Cultural Competence and Patient Care

I’m a board-certified dermatologist, practicing in Denver since 2010, with more than 20 years doing cosmetic procedures including injectables and lasers. One place AI has helped me is patient education around IPL/BBL treatments for a diverse group of patients with different skin tones, outdoor habits, and health literacy levels.

For example, I’ve used AI as a second-pass tool to simplify laser pre- and post-care language: avoid treatment on tanned skin, use strong sun protection, expect pigment to darken like “coffee grounds,” and understand that flaking can take 1-2 weeks. I still review every word clinically, but AI helps me spot where our instructions may be too medical or easy to misread.

The cultural competence lesson is that “clear” does not mean “clear for everyone.” If AI says “redness is normal,” that may not describe irritation well for patients with deeper skin tones, where swelling, tenderness, or pigment change may be more useful warning signs.

AI is best when it makes me slow down and ask, “Would this make sense to this specific person in front of me?” It should not flatten patients into averages; it should help us personalize the conversation.

One experience that stands out involved preparing for a follow-up visit with a patient whose medical history included records from multiple providers. An AI-supported clinical tool helped organize the information and highlighted trends in prior test results that deserved a closer look. It didn’t diagnose the patient or make treatment decisions, but it made the review process more efficient and helped ensure nothing important was missed before the appointment. That balanced approach aligns with guidance from the Agency for Healthcare Research and Quality (AHRQ), which emphasizes that AI should support patient-centered clinical decision making rather than replace the clinician’s judgment.

The biggest lesson for me was that AI can organize data, but it can’t understand the person behind it. Patients from different cultural backgrounds often describe symptoms differently, have varying levels of health literacy, or need information explained in a way that reflects their language, beliefs, or previous experiences with the healthcare system. Those are things that still require a conversation. I’ve found that some of the most meaningful moments happen after the technology has done its job, when I take the time to ask a few extra questions, listen carefully, and make sure the patient truly understands the plan. AI can make clinicians more efficient, but cultural competence, empathy, and trust are still what lead to better patient care.

One of the underappreciated problems in clinical AI is that most models are trained on datasets that don’t reflect the actual diversity of the populations they’ll serve.

They perform well on the demographic majority in the training data and less well on everyone else. That isn’t a theoretical concern. It shows up in diagnostic accuracy, in risk scoring, and in treatment recommendations. At QuanMed AI, we’ve made it a design principle to interrogate our models across demographic subgroups before deployment, not as an afterthought. In one case, we identified that a risk stratification model was underperforming for a specific patient subgroup because that group was underrepresented in the original training data.

We rebuilt the training pipeline with more representative data before the tool went anywhere near clinical use. The broader lesson is that AI in healthcare doesn’t automatically reduce disparity. It can entrench it if you’re not actively building against that tendency from the start.

Russell Sean, CTO, QuanMed AI / CEO, Quantum Labs

One experience that stands out involved a patient who came to our urgent care with a family member who was translating. The patient’s symptoms initially suggested a relatively minor illness, but several answers seemed inconsistent, and I wasn’t confident we were getting the full picture. Instead of continuing through a family member, we switched to an AI-powered translation tool so I could speak directly with the patient in their preferred language.

That change completely shifted the conversation. The patient shared symptoms they had not previously mentioned because they were uncomfortable discussing certain health concerns in front of a relative. They also explained that, in their culture, some symptoms were considered a normal part of aging rather than something worth bringing up to a physician. Having a direct conversation allowed me to better understand what the patient was experiencing, identify important details that changed my clinical assessment and prompted me to recommend an in-person evaluation, and develop a care plan based on complete information rather than assumptions.

The biggest lesson I learned is that AI can improve cultural competence by creating space for patients to communicate in their own words, but it cannot interpret culture on its own. Removing a language barrier does not automatically remove cultural differences. AI helped remove the language barrier, but it was still my responsibility to recognize cultural influences, ask respectful follow-up questions, and confirm the patient’s understanding before making clinical decisions.

That experience reinforced how I use AI today. I rely on it to strengthen communication and reduce barriers to care while remembering that trust, curiosity, and clinical judgment remain essential to providing equitable care for every patient.