Where does value accrue beyond Open AI?
First came the foundation models. Then the enterprise tools. Next comes the consumer wave: millions of tiny daily decisions, habits, and purchases getting rerouted through AI.
But how do you build with a giant, dynamically generated, shadow looming over you?
OpenAI has a formidable head start. And it’s not just massive…it’s everywhere. ChatGPT is likely to reach over 1 billion monthly users before the end of the year. In other words, nearly 1 out of every 5 adults across the world will be be interacting with Altman’s chatbot.
Giants tend to win by default, particularly in consumer applications where gravity is critical for building network effects. In the early 2000s, Google steamrolled everything from AltaVista to Yelp to MapQuest. But even Google couldn’t crack social (remember Google+?) or build an e-commerce engine to rival Amazon (how about Google Express?). Dominance in one vertical (even a massive one like search) doesn’t guarantee success everywhere else.
The opening exists where generalists can’t or won’t go deep. Here’s where consumer AI founders should build now:
**Complex, regulated industries**
Healthcare, taxes, insurance, real estate: these markets don’t reward “good enough answers” — they demand liability coverage, workflow integration, and regulatory compliance. Take the FDA’s AI/ML Software-as-a-Medical-Device framework: it requires pre-specified change control protocols, lifecycle oversight, and algorithmic transparency. That transforms OpenAI’s “ship fast” advantage into a liability.
This regulatory moat is why vertical specialists can win even while using OpenAI’s models underneath. Yes, OpenAI offers BAAs and enterprise privacy controls now, but compliance infrastructure isn’t just about data handling, it’s about audit trails, clinical validation, and regulatory relationships built over years.
You probably wouldn’t trust ChatGPT to file your taxes or dispute chargebacks on your credit card. And that’s good, because OpenAI doesn’t want to handle those risky workflows either.
**Bridging digital and physical worlds**
The next consumer breakthroughs won’t live in a chat window. They’ll orchestrate contractors, manage inventory, coordinate delivery windows, and handle angry customers calling at 9 PM. This operational complexity is OpenAI’s kryptonite.
Owning the last mile creates defensibility because it absorbs messy, real-world problems that general platforms avoid. This applies to everything from travel and hospitality to healthcare, homebuilding, and sports.
Doctronics is a good example. You might look to ChatGPT for general health advice, but are you going to trust it to diagnose that strange mole or handle a crucial insurance claim? At first blush, the Doctronics experience feels like any other chatbot — but the genius (and the reason ChatGPT won’t be able to compete with them) — is in the final step: AI triage leads to a telehealth session and, maybe, even in-person care if needed.
**Hardware with conviction**
OpenAI clearly eyeing hardware opportunities. But hardware remains brutal terrain, even for tech giants. Google needed acquisitions like Nest and Fitbit to go deeper into physical products, and most failed anyway. Meta, Amazon, and Microsoft have all poured billion into hardware, but still lead with software.
The opening for startups is specificity. A focused team building opinionated devices around a single, obsessive use case has room to run. While OpenAI optimizes for general intelligence, great hardware demands specific intelligence — the kind that comes from understanding exactly how people will hold, carry, and live with your product. WHOOP and Oura show the pattern in personal health: a tight loop from sensor to insight to behavior, repeated until it becomes an unshakeable habit.
I recently learned of a hardware concept in hospitality that adapts a restaurant or hotel experience based on your personal preferences. I’m eager to see other ways that hardware and software interplay in this next phase of AI, and particularly excited to see how new entrants approach wearables and integrated devices to finally divorce us from our screens.
**Creative tools that amplify taste**
OpenAI is not creative. No LLM demonstrates genuine originality: they remix and recombine existing patterns. As much fear exists around AI displacing creatives, the reality is that AI will become the most powerful creative tool ever invented.
The opportunity isn’t in replacing human creativity; it’s amplifying it. This requires understanding not just what creators make, but how they think, iterate, and collaborate. Approximately 1,000 thought pieces have been written this year about the importance of taste (I know, because I wrote one); but it’s true — foundation models can’t learn taste from training data alone.
This is why we launched AIR, a Brooklyn residency for founders building at the intersection of design and AI. We believe creatives will need to take the wheel in imbuing AI startups with creativity. And, let’s face it, Sam Altman is not known for his creativity or design instincts (the naming conventions at ChatGPT alone decry a lack of brand know-how).
******
OpenAI might own the majority of the consumer AI experience, but there’s plenty of room for new entrants to succeed, The lesson from history is clear: giants harvest the averages, but startups win the edges.