Jason Dicken
Case study 01 · Restaurant technology / AI-powered operationsProduction · my own shop

Cobblestone POS + Daisy

A restaurant point-of-sale system I built and ran my own ice cream shop on, with an AI assistant, Daisy, who answers the owner's questions from live shop data instead of canned reports.

The problem

Off-the-shelf POS systems are expensive, and they're built for how a chain runs, not how a small shop does. Online ordering, delivery, menu boards and inventory each become a separate subscription, and none of them share data.

Even when the data is all in one place, an owner who wants to know "how did we do on Saturday against last Saturday" or "why is food cost up" has to know which report to run, and usually doesn't have time to run it in the middle of a shift.

The solution

A multi-tenant POS that covers the counter, online ordering, the merchant website, digital menu boards and delivery through a DoorDash integration, all on one data model. A bar inventory module depletes stock automatically from sales.

Daisy is the assistant inside it. The owner asks in plain English, and Daisy uses tools to query the live shop data and hands off to specialist agents for sales, food cost, inventory and mid-shift troubleshooting.

My role

  • RequirementsDefined every feature from running the shop myself: what the counter, the kitchen and the owner each need to see.
  • DesignSpecified the data model, the ordering and delivery flows, and the tools Daisy may call.
  • Agent directionDirected AI coding agents, mainly Claude Code, through the implementation, and reviewed what they produced.
  • TestingTested it in daily use at my own counter, where a bug costs a sale.
  • DeploymentDeployed and operated it in production at Cobblestone Creamery.

Technology used

  • Node.js
  • Express
  • Firebase Cloud Functions
  • Firestore
  • Next.js
  • DoorDash integration
  • REST APIs
  • Claude Code

Key features

  • Point of saleMulti-tenant, so one deployment serves more than one location or business.
  • Online ordering & merchant siteCustomers order from the shop's own site, on the same menu and data as the counter.
  • Digital menu boardsMenu changes show up on the in-store screens without a separate system.
  • DoorDash integrationDelivery orders come into the same system as counter and online orders.
  • Bar inventoryDepletes stock from sales, with supplier views, REST APIs and a catalog of about 600 drinks.
  • Daisy, the AI assistantAnswers sales, food-cost and inventory questions from live data, and walks staff through problems mid-shift.

Application screenshots

Screenshots from the running system are being prepared. This page will show real interfaces only, never mock-ups. Until then, the recorded demos below show the product in use.

Demo

Product demos of the POS are on the Cobblestone YouTube channel.

Development status

  • Working prototype
  • Active development
  • Internal production
  • Customer-deployed

POS: deployed in production at Cobblestone Creamery, the shop I built, ran and sold in 2026. I'm not claiming outside customers.

Daisy: in production inside the POS. She has no test set yet, so I make no accuracy claim for her answers.

Results

The POS ran a real shop: counter sales, online orders, delivery and menu boards on one system I specified and operated, instead of separate subscriptions.

The bar inventory module ships with a catalog of about 600 drinks and depletes stock from sales.

I haven't published revenue, time-saved or accuracy figures because I haven't measured them. The next step for Daisy is the same kind of test set I built for the support assistant.