Slicematic — AI operations copilot for a pizza outlet
An order management system that doesn't stop at tracking orders — it reads them and tells the owner how to grow.
Client
Independent build — single-outlet pizza business scenario
Role
Solo builder: product scoping, full-stack build, AI insight design, deployment
Timeframe
2026
Stack
Next.js · React · AI insights · Vercel
01 — Situation
Small food outlets run on point-of-sale data they never look at twice. Orders come in, orders go out, and the questions that decide profitability — what sells together, when demand peaks, which items underperform — go unanswered because the owner has no analyst.
Slicematic treats a single pizza outlet as a client engagement: take orders, track them through fulfilment, and turn the accumulated order data into plain-language business insights the owner can act on.
02 — Constraints
- !The user is a shop owner, not an operator of dashboards — insights had to read like advice, not analytics.
- !Order-taking flow had to stay fast enough for counter use; intelligence could never slow down operations.
- !Zero-budget infrastructure: everything runs on Vercel's free tier.
03 — Stakeholders
Outlet owner
Profit levers: what to promote, stock, and schedule
Counter staff
Taking and tracking orders with minimum friction
Customers
Order status without asking the counter
04 — Architecture
- 01Next.js app with an order capture flow, live order-status tracking, and an owner dashboard as separate surfaces for separate users.
- 02Order events accumulate into a dataset that an AI insight layer summarizes into recommendations — demand patterns, item performance, growth suggestions.
- 03Deployed on Vercel with preview deployments as the de-facto staging environment.
05 — What shipped
- ✓Live order-taking and order-tracking flows.
- ✓Owner-facing AI insights on orders, oriented around growing revenue and margin.
- ✓Production deployment: slice-matic-lime.vercel.app.
06 — Outcomes
1
production system live on Vercel
3
user surfaces: counter, customer, owner
AI
insight layer turning order data into decisions
07 — Retro: what I'd do differently
The consulting lesson: the owner dashboard got better the moment I started writing insights as recommendations ('promote X on weekends') instead of charts. Framing beats data density.
Next iteration: evaluation of the insight quality itself — an insight engine without evals is an opinion engine.
Questions about this engagement?
Ask the copilot →