The Cheesecake Factory Problem in AI (Why Doing Everything Kills Trust)

Last December I flew to San Francisco to spend a morning with the Lindy team at their office.

They were walking me through a big internal decision: pivot from a general-purpose no-code automation platform to something much more focused. Basically just an AI work assistant for email and meetings. A bespoke product built around two specific jobs rather than a flexible canvas where you could build almost anything.

They wanted my read on it.

My read was: this is exactly right. And I told them why using the Cheesecake Factory.

The Menu Problem

Here’s how I think about it.

Hand someone the Cheesecake Factory menu and watch what happens. 250 items. Thai food, Italian, pasta, sushi, steaks, burgers, all in one place. You can get anything. And your immediate reaction — even if it’s unconscious — is: can they actually do any of this well?

That doubt is real. It’s the cost of breadth.

Now think about a Chinese restaurant with 12 items. You don’t question whether they’re good at Chinese food. You just go. The narrowness is a signal that they know what they’re doing.

This is the exact same psychology at work with AI tools right now.

Most of the AI tools people encounter are the Cheesecake Factory. They can do so many things — write emails, summarize meetings, generate images, build automations, search the web, run agents — that you end up trusting none of it enough to depend on it for anything serious.

And the irony is that the breadth, which is supposed to be a selling point, becomes the reason people don’t commit.

Where This Shows Up With AI Tools

I’ve been teaching AI workshops since February 2025. By now I’ve seen this pattern play out dozens of times.

Someone comes in excited about AI. They’ve heard you can do everything with it. They spend a day in the workshop, build a working executive assistant in Lindy that drafts their emails and sends them a schedule every morning. Real, working software. By the end of the day, they’re kind of amazed.

Then I ask if they’re going to customize it further on their own.

Almost everyone says no.

Not because the tool is too complicated. Because they’re afraid to touch it.

And part of that fear comes from not understanding what the thing actually is. If a product can do 500 things, how do you know what you’re changing when you poke at it? How do you know what you might break? The mental model is too big.

A tool that does one thing clearly is a tool you’re willing to poke at. A tool that does everything is a tool you leave alone after someone sets it up for you.

The Mailchimp Precedent

When I explained this to the Lindy team, the comparison that came to mind first was Mailchimp.

Mailchimp started as just an email newsletter tool. That’s it. Not a CRM, not a landing page builder, not a marketing hub. Just: here is a tool for sending email newsletters.

That narrow focus made them easy to understand and easy to pitch. Customer acquisition was straightforward because the value prop fit in a sentence. Over time they added landing pages, CRMs, automation, everything else. But the growth engine was the one thing done really well.

HubSpot. ConvertKit. Same story. They all started narrow. The breadth came later, after trust was established.

You can’t ask people to think their way into trusting you. You’ve got maybe five seconds of attention, especially in the current environment where everyone is evaluating AI tools. You need to be crystal clear: we do this one thing, and we do it well.

The Lindy pivot was right because it follows this principle. Instead of “here is a canvas where you can automate anything,” the new positioning is “here is your AI work assistant for email and meetings.” Most people can immediately understand the value. Some of them will discover later that it can do more. But the initial trust is easier to earn when the promise is specific.

This Isn’t Just for Software

I want to be clear that this applies well beyond AI products.

If you’re a consultant or coach and you claim to do everything, you’re the Cheesecake Factory. The prospect’s brain immediately starts calculating: but are you actually good at the specific thing I need?

If you’re building a consulting offer and you specialize in one industry and one specific outcome, you’re the Chinese restaurant. The prospect can calibrate quickly and trust more.

I’ve watched this play out with my own work. The AI workshops fill up because the promise is specific: spend a day building your own AI executive assistant. Not “learn about AI” — that’s too vague. One day, one outcome, one tool.

There’s also the flip side. Once you earn the trust with the specific thing, people naturally start asking “can you help with X too?” That’s when you expand. But you can’t expand from zero trust. You need the beachhead first.

What to Do With This

If you’re evaluating AI tools right now, look for the ones that have a clear specific job they do. Not the ones that promise everything. The tool that’s laser-focused on one workflow and does it exceptionally is more valuable than the Swiss Army knife that does twenty things at 70%.

If you’re building something, whether that’s software, a consulting offer, or a new service, ask yourself: what is the one sentence that explains who this is for and what it does? If you can’t say it in one sentence, it’s not focused enough yet.

The Cheesecake Factory is a fine restaurant. But you don’t go there when you actually need something done right.

P.S. My AI workshops are built around exactly this principle — one tool, one workflow, one day. The Two Hour Workday course takes the same approach to AI-assisted work. If you want to understand how to actually put AI to work in your life rather than just learn about it, that’s where to start.