Executive Summary
CEOs are under immense pressure to innovate. But there is a quiet, expensive, and dangerous trend taking hold in the C-suite. We are pouring fortunes into these “token furnaces” that, rather than fostering innovation, are guiding our organizations onto the same homogeneous superhighway of mediocrity everyone else is on.
I believe we are witnessing the “Walmart-ification” of Silicon Valley; a race to the bottom where the winner is simply the one who uses the least expensive AI most efficiently. This is not innovation. It is an accelerated path to being indistinguishable.
The Myth of AI-Driven Innovation
It is the modern version of the old adage: Nobody ever got fired for buying IBM. Today, that translates to FOMO. We use the same models as our competitors, follow the same “best practices,” and deploy the same generic prompts. But “safe” isn’t actually safe in today’s world. Following the herd doesn’t provide safe harbor; it makes you another seal clinging to a crowded iceberg in shark-infested waters. When your competitive edge is derived from the exact same LLMs as your rivals, you haven’t built a moat, you’ve crawled into an echo chamber.
I feel as though the corporate world is rapidly losing focus. For example: allowing “citizen developers” to treat AI as a magic wand without corporate guardrails or a strategic North Star, we’re creating chaos disguised as productivity. A shiny app is just a shiny app; if it isn’t driving unique ROI, it’s just a distraction.
Convergence Experiments
To see the problem in action, I ran a simple experiment. I asked the three major frontier models to act as a professional business consultant.
1. Let’s Start a Business!
PROMPT: Come up with 3 passive income ideas. Assume I have a technical background in IT, and can invest up to $10,000 to get the income stream rolling.
| ChatGPT | Gemini | Claude |
| Build a niche self-hosted SaaS | Micro-SaaS / API Acquisition | Niche SaaS micro-tool or API wrapper |
| Build a small “boring infrastructure” business | Decentralized GPU Compute Hosting | Dividend-paying index funds / REITs |
| Sell technical digital products | DevOps / IaC Asset Marketplace | Automated content or tooling for developers (courses, templates, boilerplate kits) |
Every single model recommended essentially the same two paths: build a SaaS offering or sell digital products. Despite the limitless possibilities of the modern digital economy, the models converged on near-verbatim advice.
Kudos to Claude for thinking somewhat outside the box.
2. Let’s Bake a Unique Cookie
Unsatisfied with all entrepreneurial roads leading to IT, perhaps I’ll take my new business venture in a culinary direction. Maybe a bakery?
PROMPT: Give me three new, highly unique cookie recipes; something the world has never had before. I like chocolate and I also love peanut butter, but you don’t have to use these ingredients. Free to get creative!
| ChatGPT | Gemini | Claude |
| Chocolate-Cornflake Miso Cookies with Lime Sugar | Gochujang Peanut Butter & Burnt Caramel | Miso Peanut Butter Blondie Cookies with Burnt Honey Drizzle |
| Peanut Butter, Raspberry, Cardamom & Potato Chip Cookies | Black Garlic & Espresso Fudge | Fermented Fig and Black Pepper Sablés with Goat Cheese Filling |
| Black Sesame Peanut Butter “Crackle” Cookies | Toasted Sesame & Yuzu Matcha with white miso | Black Sesame Chocolate Cookies with Cold-Set Raspberry Core |
Here I expected waves of culinary innovation. Instead, I got an apparently “burnt” mountain of miso and sesame. With thousands of potential ingredient combinations, these models all converged toward the same specific, Asian-inspired flavor profile. If they can’t differentiate a cookie recipe, why do we trust them to differentiate our corporate strategy?
Anyone up for some black garlic & espresso fudge?
3. Let’s Build a New Website
My new cookie business obviously needs an online store front. Something fresh, something unique.
PROMPT: Build me a static HTML homepage for a cookie business named “Dave’s Got Dough” which highlights the cookie of the week in a hero banner near the top of the page. The footer should be simple and clean with an “About Dave” section (with sample text filler you can create for now) and a “Locations” section with two locations: One in Houston, the other in New York City. The cookie of the week is the “Kernel Panic” with a description something like: “A sweet-and-savory golden cornmeal cookie packed with caramelized sweet corn kernels, freeze-dried corn powder, and brown butter. It triggers an immediate system shutdown of your self-control, proving that a critical exception in your diet is sometimes a feature, not a bug.”
As shown above, each model produces nearly identically boring, cookie-cutter (pun totally intended) designs.
The irony here is, in order to design truly inspiring websites, you must know how to prompt the AI with specificity towards code libraries, styling choices, and so on. In other words, you have to know how to build websites in order to tell the AI how to build websites. Do you see the innovation paradox there?
The Homegenous Superhighway to Boring Town
The reason for this convergence is simple: large language models (LLMs) are probabilistic distribution engines. They aren’t designed to give the “best” answer; they are designed to give the most statistically likely next token based on their training data.
When your enterprise relies on off-the-shelf LLM workflows for architecture, code generation, or operational strategy, you are paying a premium to implement the mathematical average of the public internet. In other words, you might be paying millions to erase your own differentiated advantages.
What AI Should be Used for In the Enterprise
It is not all doom and gloom. AI is a powerful tool, but we need to stop asking AI to innovate and start asking it to automate.
Innovation is the disruptive, market-shifting creation of value, which I argue is still 100% a human domain. Science, art, and genuine technological breakthroughs require a friction that statistical models actively work to smooth out.
AI excels at the “boring stuff.” It is unparalleled at:
- Parsing massive, disorganized knowledge bases
- Scanning hundreds of thousands of PDFs to extract metadata
- Breaking down silos and automating repetitive human toil
As Youngme Moon brilliantly argues in her work on differentiation, doing what everyone else does leads to death by commoditization. If you want to stand out, you have to be willing to look different. Use AI to clean your house, manage your data, and scale your operations. Yet keep your innovation and your strategy strictly in human hands.