The most impactful AI application I've seen isn't a chatbot, a recommendation engine, or an…

image generator.

It's a real-time captioning system for deaf and hard-of-hearing users.

Whisper + a lightweight LLM for formatting + a low-latency streaming pipeline = live captions that are more accurate than human transcribers, available 24/7, and essentially free at scale.

For someone who's deaf, this is the difference between participating in a meeting and being excluded from it. Between understanding a lecture and missing it entirely.

AI accessibility applications are profoundly impactful and criminally understaffed. Visual description for blind users. Sign language recognition and translation. Predictive text and communication aids for people with motor disabilities. Cognitive assistance for neurodivergent users.

Every one of these is technically feasible with current tools. Most are underbuilt because there's less venture capital interest in accessibility than in advertising optimization.

If you want your AI work to genuinely change lives — not just optimize metrics — accessibility applications are where the biggest gap exists between what's possible and what's been built.

The market might be smaller. The impact per user is orders of magnitude larger.