We talk a lot in design about how "the best interface is no interface." But what happens when you apply that same thinking to AI? What if, instead of creating yet another dashboard or stream of alerts, you built an assistant that worked quietly behind the scenes—only stepping in when it had something truly important to say?

This is the story of designing exactly that: a back-pocket AI whose entire job was to keep people from losing money, without asking for their attention. The goal? Less stress, fewer missed opportunities, and a whole lot less noise.

Back in 2015, designer Golden Krishna wrote The Best Interface is No Interface, a book that challenged the idea that more screens, more clicks, and more notifications made better products. At the time, the idea of invisible tech—tools that quietly work for you in the background—felt almost radical.

Fast forward to today, and suddenly that philosophy fits perfectly into the world of AI. We're finally able to design assistants that genuinely take work off people’s plates, without demanding attention in return. This project was one of those moments: a quiet, back-pocket AI built to track supplier rebates, save people from losing money, and only speak up when it really mattered.

🧠 How We Designed a Back-Pocket AI to Catch Lost Revenue

A while back, I worked on a project that asked a simple question:

What if AI could quietly keep an eye on things—no dashboards, no digging—just a silent partner making sure important stuff didn’t slip through the cracks?

For the folks we were designing for, this was about protecting revenue. They were managing dozens of financial programs, like rebates and incentives, all with different rules, deadlines, and thresholds. And it was way too easy to miss something.

This is the story of how we designed an AI that worked in the background, helping people stay ahead of the chaos—without adding more noise to their day.

The Problem We Were Trying to Solve

Imagine you're juggling 30 different rebate programs. Each one has its own requirements, deadlines, and dollar amounts at stake.

Now imagine tracking all of that in spreadsheets, emails, and mental notes.

The risk of missing something is high. And every missed rebate? That's real money lost.

The people doing this work weren’t asking for another tool. They just wanted a way to know when they needed to step in—and otherwise, not have to think about it.

Ill first explore all the user needs and pain-points, Ill use Claude to help me write and explore user stories.

Ill then map out the users journey and map out possible touch-points

Approach: Make It Invisible

We kept coming back to this idea of a "back-pocket" AI—something that's always there, quietly watching over things, but only taps you on the shoulder when it really matters.

No new interface to check.
No constant pings.
Just.. peace of mind.

How We Made That Happen

We focused on three big things:

Only Speak Up When It’s Important
No one wants another stream of low-value notifications. This AI had to be smart enough to stay silent until it had something truly meaningful to share—like, “Hey, this big rebate is about to expire. You should look into it.”
Blend Into the Tools People Already Use
Why make someone log into yet another system? We pushed updates into the places they were already hanging out—like email or Slack—so the help felt seamless.
Make the Work Visible (When You Want It)
Even with an invisible system, people need to trust what’s happening behind the scenes. So we added optional status updates and activity summaries—just enough insight for someone to feel confident the AI was on top of things.

What It Looked Like

Voice messages through the users mobile devices can be used to start a conversation with the copilot and delegate tasks or set up alerts.

A simple smartwatch notification can be set to alert users when a claim is getting close to meeting the requirement or when they are about to expire.

Users can text the Copilot to send claims or draft communications from their smart devices.

What Worked

The magic wasn’t in the interface (there really wasn’t one).
It was in the calm it created.

People stopped worrying about missing deadlines.
They spent less time manually tracking details.
And over time, the AI just became part of the environment—something they trusted to quietly have their back.

What I Learned

Not every problem needs a big, shiny UI.
Sometimes the best design is one people barely notice.
Trust is everything when it comes to automation—especially when money’s on the line.

If I had to sum it up?
The best kind of AI feels like a safety net, not another task.

References:

Krishna, Golden. The Best Interface is No Interface. O'Reilly Media, 2015.
- Golden Krishna's groundbreaking book challenges the traditional notion that more screens and interactions create better user experiences. His thesis advocates for a shift toward simpler, less invasive interfaces—something that became a central inspiration in the design of our back-pocket AI system.
Norman, Donald A. The Design of Everyday Things. Basic Books, 2013.
- Norman’s work has been foundational in the field of UX design, particularly when it comes to the concept of user-centered design. His philosophy of making products intuitive and seamless aligns with the invisible AI design we sought to achieve. It’s all about designing tools that “just work” without demanding your attention or overloading you with complexity.
Hassenzahl, Marc. Experience Design: Technology for All the Right Reasons. Morgan & Claypool, 2010.
- Hassenzahl's insights into how technology impacts human experience were instrumental in considering how users interact with technology in the background. He emphasizes the emotional experience of technology, an angle that guided our thinking about designing AI that’s supportive and doesn’t overwhelm.