The Best AI Is Invisible AI

When we visited an indoor padel facility in San Diego, the first thing you noticed was all the wires.

Camera systems running across the walls. Infrastructure everywhere. The technology was obviously there, and it took something away from the space.

So when we designed the courts for Padel Society, we made a different decision. Run the wires underground. Keep the surfaces clean. Same cameras, same tech — just hidden.

The experience you want to have on a padel court shouldn’t include noticing the infrastructure.

I’ve been thinking about this a lot while building AI systems. Because the same principle applies.

What Great AI Implementation Actually Looks Like

I’m currently building an AI system for Arena Hall, a new members club opening in Austin. One of the features we’re designing is how to make sure the right members meet each other.

The problem sounds simple. In practice, it’s complicated.

You have hundreds of members with different interests, goals, professional backgrounds, and social circles. Every event, the right introductions need to happen — but you can’t rely on a staff member knowing everything about every member at every moment.

Here’s how the system works.

Every member interaction gets captured — meetings, event check-ins, conversations. Profiles are built and updated continuously. Before any event, an AI agent cross-references who’s attending against the member database, identifies the introductions worth making, and surfaces them in a briefing for the hospitality team.

Staff reviews the brief. Shows up prepared. At the right moment during the event, they find you and introduce you to exactly the person you should know.

From your perspective as a member? The staff just seems incredibly attentive. Perceptive. Like they really understand the room.

You have no idea that AI compiled that briefing thirty minutes before you arrived.

That’s the outcome worth building toward.

The Problem With Visible AI

A lot of AI deployments get this backwards. They make the technology prominent — the chatbot widget, the “AI-generated” label, the automated response that feels obviously automated.

Sometimes that’s fine. Sometimes people want to know they’re talking to an AI.

But in a premium service environment, visible AI can actually subtract from the experience. The moment a guest thinks “oh, this is automated,” the warmth of the interaction disappears. It stops feeling like a curated experience and starts feeling like a product.

The goal in most service businesses is the opposite: the person should feel seen, recognized, taken care of. If AI can create that feeling — great. But they shouldn’t see the machinery.

The Principle Applies Everywhere

Think about the best restaurants you’ve been to. You don’t think about the kitchen logistics, the reservation system, how the staff coordinates on who’s going where. You just notice that your water gets refilled before you ask, your order comes out right, the experience is smooth.

The infrastructure is doing a lot of work. You’re just not seeing it.

This is the standard to aim for with AI in service businesses.

Another example: a client I worked with who runs a membership club in Austin was spending enormous amounts of time on meeting preparation. Hours of digging through emails, calendar notes, and documents to be ready for each conversation. We built a digital chief of staff that now handles all of that automatically — meeting context, attendee backgrounds, key discussion points. What used to take twenty hours a week now takes fifteen minutes.

Does his counterparty know he has an AI briefing them? No. He just shows up better prepared. That’s the experience they get.

What This Means for Building AI Systems

If you’re deploying AI in your business — for customer service, for operations, for relationship management — the question worth asking is: will people notice the technology, or will they just experience the result?

Both can be valid. But in premium contexts, the invisible option is almost always better.

A few things that help:

Start with the outcome, not the feature. What should someone feel when they walk away from this experience? Then work backwards to what the AI needs to do.

Centralize the data first. Most AI systems underperform not because the AI is bad but because the underlying data is messy or scattered. Before building workflows, get your data into one clean place. Once the data is right, the automation is the easy part.

Audit what’s visible. Every AI touchpoint is either adding to or subtracting from the experience. The ones that feel clunky or robotic are worth redesigning, not just improving.

The padel court doesn’t show its wires. The members club doesn’t explain its AI. The meeting prep system doesn’t announce itself.

They just create the experience you wanted them to create.

If you’re thinking about where to start with AI in your business, the 4-Day AI Sprint walks through the practical implementation — from first workflows to more sophisticated systems.