A Case for Invisible AI

It’s sometimes easy to forget that SaaS powers industries outside of tech.

The Great AI Divide

We're living in two worlds right now.

In one world—the tech bubble most of us operate in—people are having daily conversations with AI assistants, generating images for work presentations, and automating parts of their workflow with AI tools. The usage statistics for tools like ChatGPT (180+ million users) and Claude seem impressive at first glance.

But step outside this bubble, and you'll find a vastly different reality. The broader consumer landscape reveals that regular, productive AI usage remains limited to a small segment of early adopters. A recent survey showed that while awareness of AI tools is high (nearly 90% of Americans have heard of ChatGPT), regular usage outside tech circles hovers around 25-30% for any AI tool, with even lower percentages for consistent, productive use.

This divide becomes even more pronounced when we look at small business owners and industry professionals outside of tech. While Silicon Valley races ahead with one AI breakthrough after another, many industries remain stuck with outdated tools and persistent problems that technology could easily solve.

The Wedding Industry: A Case Study in Tech Lag

In my own work at David's Bridal (big news coming soon), I was blown away to find that the leading "SaaS platforms and products" in the wedding space are remarkably limited in their utility and functionality. These are tools that millions of couples and wedding vendors rely on daily, and they don’t really help solve the problems that have been persisting for decades (or more).

Shortly after I started researching platforms like Joy and The Knot, I discovered that wedding vendors have been plagued by fake leads and questionable platform practices. While there are a few stand out solutions for specific needs like Honeyfund, the problem space remains mostly untouched by what I would consider current/modern technology. While I’m not an advocate for applying technology simply because you can, I am a huge advocate for solving problems in the best way possible for the customer/user.

This problem isn't unique to the wedding industry. With all the sophisticated technology we have for attribution, lead tracking, and verification, it's mind-boggling that better solutions don't exist. Then I remembered something important: many of these technical concepts are completely foreign to the average small business owner. They aren’t interested in the tech itself, and the tools that do exist are too complex and continue to get more so over time.

The Overload Effect

The initial state of being overwhelmed, anxious, and resistant to technology is something I call this the "Overload Effect".” I once spent time explaining to a local burrito shop owner how he could track coupon redemptions and social media conversions. As I walked him through basic screens in Google Analytics and Facebook Ads Manager, I watched stress and confusion wash over his face, and watched him immediately forget about his initial goal of understanding whether or not his advertising budget was working, and move on to things he felt more comfortable thinking about.

This isn't just about technical literacy. It's about cognitive load and priority management. The owner's primary concern was making great burritos and keeping customers happy—not mastering complex analytics dashboards. The technology that was second nature to me represented an intimidating learning curve and time investment to him.

It’s important to remember all of the bits of experience, knowledge, insights, and repetitions that going into being comfortable using a product. We have a lot of affordances when building for other tech savvy people they we need to be hyper aware of when designing and building for the less tech savvy groups — which still make up a massive chunk of consumer and business markets.

The Invisible AI Approach

This realization led me to a fundamental principle for our design approach: The end user doesn't need (or want) to know what technology powers your product.

Looking at products that are built on or heavily leverage AI, we see a lot of the same interaction models and a LOT of AI-centric marketing. What many of those teams failed to realize is that they are introducing friction if they don’t know exactly how their users will engage OR if the user isn’t sure how to engage.

While those of us who grew up with AOL Instant Messenger, HipChat, and Slack intuitively understand conversational interfaces, many people find chatting with a machine strange and uncomfortable, and don’t know how to think about what is or could be possible with the AI.

You may have one user who assumes it can just answer questions and a another who assumes it can (and should) do everything for them. There’s a serious amount of thought that should go into deciding how and when to surface the AI — or if you are going to do it at all.

This isn't just about generational divides. It's about meeting users where they are rather than forcing them to adapt to our technological paradigms. This is precisely why I've argued for technical knowledge for non-technical product managers—understanding the technology is crucial, but exposing it to users probably isn’t.

Principles for Invisible AI Implementation

Based on my experience building AI-powered products for diverse audiences, here are key principles for creating effective "invisible AI" solutions:

1. Focus on Outcomes, Not Technology

Users care about solving their problems, not the specific technology behind the solution. Frame your product in terms of tangible benefits rather than technological capabilities.

I have a post about “The Problem Obsession” where I emphasized the importance of identifying "sharp problems" with significant impact on users. This becomes even more critical with AI products, where it's easy to get distracted by the technology itself rather than the problems it solves.

Example: Rather than marketing a "ChatGPT-powered assistant," position your product as a solution that "creates customized marketing campaigns in minutes" or "simplifies your travel planning."

2. Design Familiar Interfaces for Unfamiliar Technology

Use interaction patterns that users already understand. DON’T use them if they create more work for the user or increase cognitive load. DO use them if they are the most obvious and recognizable choice for the user to do the thing they need to do to kickstart the magic.

Real World Example with Napkin AI

Yes, they have “AI”in their name but I could show this to my wife or my kids and they would understand how it works without having to think about it. You put some text in and then you click the magic button. That’s it!

I’m really excited to see new products emerge that make people rethink what is possible with a simple interaction. Are buttons going away? Probably not. Can you do more in a single click of a button than you could a couple years ago? Yes, exponentially more.

3. Collect Information Progressively

AI systems need data (or context) to be effective. Design your products to gather necessary information through natural interactions over time, rather than trying to get it all up front. Your core experience you should be great without any data, but as the system learns about the user, the experience should get better and better.

5. Introduce Features Progressively

Rather than trying to show the user everything the product can do, consider letting them discover some magic moments of their own once they are familiar with the core uses and utility of the product. I immediately knew how to set up Linear, but it took me a minute to figure out that there was a command bar AND a ton of helpful keyboard shortcuts.

The trick is designing them in a way where the user can discover them, but doesn’t HAVE to use them until they are ready. Linear has UI for everything I might need to do, but why would I keep doing something in a way that takes 3 clicks if there’s a way where I can do it with a keyboard shortcut?

I could use the command bar, card preview, or card detail ui to set the priority or I could simple hit “P” and then “1” if I want it to be a P1. It works on the task detail and it even works if the cursor is hovering over the task in the board view.

The Future of Invisible AI

As I've explored in my Rise of Generalists, we're entering an era where technology is enabling broader capabilities rather than more specialized ones. The most successful AI implementations will be those that expand human capabilities without requiring humans to become AI specialists themselves.

The next wave of transformative products won't be "AI products"—they'll simply be excellent products that happen to use AI. They'll solve real problems in intuitive ways, and most users won't know or care about the technology that powers them.

This represents a significant opportunity for product thinkers who can bridge the gap between cutting-edge technology and everyday user needs. By focusing on sharp problems outside the tech bubble and designing solutions that feel natural rather than technological, we can create products that truly improve lives beyond the AI-enthusiast crowd.

Have you seen examples of "invisible AI" done particularly well or poorly? I'd love to hear about your experiences in the comments.

If you enjoyed this article, you might also like my exploration of Deep Research and Prompt Design or how technical knowledge enhances product management.