Building a B2B office-supply store around one simple idea.

A business shouldn't have to figure out the same supply twice.

Where the idea came from

When I joined Randmar Academy, the obvious thing to do was build an office-supply store.

Randmar already had the products and distribution. Shopify could handle the storefront. The temptation was to add as many products as possible and start selling.

Instead, I spent a lot of time looking at the catalogue and asking what kind of problem I actually wanted Maple to solve.

I kept coming back to printing supplies.

Buying toner isn't exciting, but getting it wrong is expensive and annoying. A business doesn't really want to shop for toner. Someone has a printer, they need the right cartridge, and they want to get back to work.

That became the idea behind Maple Office Supplies.

“A business doesn’t really want to shop for toner. They want the right supply and they want to get back to work.”

So, Maple

1. Find
2. Match
3. Order
4. Reorder

The second order should be easier than the first one.

A customer can search using the printer model or the MPN instead of browsing through a huge catalogue.

Maple checks the compatibility, takes them to the right supply, and once they've bought it, the Reorder Hub starts remembering the purchasing history.

That's the part of the business I became most interested in.

Instead of asking a company to search for the same cartridge again six months later, Maple can show what it purchased before, how often it purchased it, and let the buyer reorder it directly.

Building it

I didn't start this project as a software company.

But the buying experience I wanted meant going much further than changing a Shopify theme.

I used AI coding agents alongside Shopify to build the storefront, product compatibility system, Quick Order, structured printer data and the Reorder Hub.

The Hub became its own application connected to Shopify, using the company's actual purchasing context to identify frequently purchased supplies.

Two test orders for Xerox cartridge 106R03944 demonstrated the repeat-purchase pattern. The buyer could change the quantity and add the correct canonical variant back to the cart.

The technology is there to make something complicated feel boring.

What AI changed

AI made it possible for me to build things I would previously have needed several different people to build. But it didn't make the decisions for me.

There were plenty of times when something technically worked and I still rejected it.

A page could work but feel wrong. A compatibility match could look plausible but need manufacturer verification. A feature could be impressive but add friction instead of removing it.

I decide, the agent builds, and we test against the real system. I review it, and we change it until it earns its place.

Reorder / verified purchase history

“Maple should never ask a business twice for information it already knows.”

What I’d say I learned

The building workflow

1. I decide what Maple should do
2. The agent builds it
3. We test it against the real system
4. I review it
5. We change it until it earns its place

At the beginning, I thought the advantage might be having the right products at the right price.

Those things matter.

But the more I built Maple, the more I realized that the real opportunity is in remembering the customer's purchasing context.

For a business buyer, convenience isn't another animation or another recommendation widget.

It's knowing:

• This is my printer.
• This is the supply that fits it.
• This is what my company ordered last time.
• I need two more.

And then making the rest disappear.

AI has made the cost of building software dramatically lower.

The harder question hasn't changed:

What is actually worth building, and for whom?

For Maple, my answer became pretty simple:

Remove friction from recurring business purchasing.

Find → Match → Order → Reorder

No friction. Just supplies.