Last week, Google released Nano Banana Pro, their latest image generator. The demos looked impressive. I opened Gemini to try it.
Then I had a question I needed to ask. Something unrelated to images. And without thinking, I switched back to Claude.
Not because Claude is objectively better at everything—honestly, I have no idea how the benchmarks shake out this week. I switched because I *know* Claude. I trust how it thinks. I’m comfortable with its voice. Going back to Gemini felt like visiting a stranger’s house when I could just walk into my own living room.
This is supposed to be irrational behavior. The “smart” move would be to evaluate both tools objectively, run tests, compare outputs. But I didn’t do that. I just went home.
The 9x Problem Nobody in AI Talks About
In 2006, John Gournville published a paper in Harvard Business Review called “Eager Sellers and Stony Buyers” that should be required reading for every AI founder. His core finding: to get users to switch products, your new thing has to be *nine times better* than what they already use.
Not twice as good. Not “demonstrably superior.” Nine times.
Here’s why. Users overvalue what they already have by a factor of three—the familiarity, the muscle memory, the sense of control. And companies overvalue what they’re offering by a factor of three—because they built it, they know every feature, they see the potential. Three times three equals nine.
This creates what Gournville called a “mismatch of nine to one, or 9x, between what innovators think consumers desire and what consumers really want.”
AI companies act like their next model release will make users switch. They announce benchmark improvements like they’re declaring victory. “Our model is 12% better at coding tasks!” Cool. Is it nine times better? No? Then I’m staying where I am.
Netflix Won on Comfort, Not Content
Someone on Twitter recently pointed out that Netflix won the streaming wars despite not having the best content. HBO has prestige dramas. Disney has Marvel and Star Wars. Apple is throwing money at A-list directors.
Netflix won anyway. Why?
Because the experience is frictionless. You open the app, you start watching, and before you know it you’ve finished one show and started another. The recommendations are *good enough* that you rarely feel like you’re wasting time. The interface gets out of your way.
Netflix feels like a comfortable sofa. You sink into it. The other streaming services feel like you’re shopping.
This is the moat that AI companies keep ignoring. They’re obsessed with model capabilities—who has the best reasoning, who generates the most accurate images, whose context window is bigger. Meanwhile, the real competition is happening at the interface level, and most of them are losing
The Tools That Trap You (In a Good Way)
Watch what happens with AI coding tools. Someone starts using Cursor or Windsurf or Lovable. The first few days are awkward—where’s this button, what does this command do, why isn’t it understanding my intent?
Then something shifts. They learn the keyboard shortcuts. They figure out how to phrase prompts for that particular tool. They develop a feel for when to trust the suggestions and when to intervene. The tool starts feeling like an extension of their thinking.
Now a competitor launches with better underlying models. Marginally faster. Slightly more accurate completions. Does our user switch?
Almost never.
The cost isn’t the subscription fee. The cost is *relearning everything*. Where do they hide the buttons in this new interface? What’s the mental model here? How do I phrase things to get good results? That learning took weeks the first time. Who wants to do it again for a 15% improvement?
The same pattern plays out with foundation models. Once you develop an intuition for how Claude thinks—or GPT, or Gemini—you don’t want to start over. You’ve built up a working relationship. You know when to push back, when to give more context, when the model is likely to misunderstand you. That knowledge is valuable, and it doesn’t transfer.
What This Means for AI Companies
If you’re building an AI product, your interface isn’t a nice-to-have. It’s your primary competitive advantage.
Here’s what that means in practice:
Make the first five minutes seamless. Users decide whether they’re staying or leaving almost immediately. If they have to think about where to click, you’ve already lost. Netflix auto-plays. TikTok starts scrolling. What does your product do the moment someone opens it?
Build for muscle memory. The interactions people repeat daily are the ones that lock them in. Keyboard shortcuts. Predictable flows. Consistent patterns. Every time a user does something without thinking, you’ve deposited money in the switching-cost bank.
Let users customize in ways that make leaving painful. Saved prompts. Custom instructions. Learned preferences. The more users invest in configuring your product, the more they’ve built something that doesn’t exist anywhere else.
Stop obsessing over benchmarks. I know this is heresy. But a 10% improvement in capability matters less than a 10% improvement in how it feels to use your product. Users don’t experience benchmark scores. They experience interfaces.
The Uncomfortable Truth
Here’s what nobody wants to admit: the best model might not win.
The most useful model might not win.
What wins is the model that people don’t want to leave. The one that feels like home. The one where switching would mean losing something—not just access to features, but fluency, comfort, all those intangible things that make a tool feel like yours.
Amazon figured this out with Prime. Apple figured it out with the ecosystem. Salesforce figured it out by making itself so embedded in enterprise workflows that ripping it out would require an act of God.
AI companies are still acting like this is a pure technology competition. It’s not. It’s a competition to become essential—and staying power comes from experience, not raw capability.
Your moat isn’t your model. Your moat is whether users feel at home.
—–
For more on why users resist change and how to overcome it, read Gournville’s original paper: [Eager Sellers and Stony Buyers](https://hbr.org/2006/06/eager-sellers-and-stony-buyers-understanding-the-psychology-of-new-product-adoption).