Candidate prototype | Synthetic operations data

Tortilla Close-Out Review Copilot

A transparent 17:00 triage screen for store teams. It highlights ingredients that remain above even the strongest comparable recent evening demand, then presents evidence for a human decision.

Download structured review JSON
Demonstration boundary: all inputs are synthetic. This screen does not connect to Tortilla systems, change preparation plans or create surplus listings. A store manager remains responsible for any action.
Stores scanned3
Components assessed9
High-priority flags2
Monitor flags1

Prioritised review queue

A high-priority flag appears only when remaining portions still exceed the highest evening demand observed in the comparison shifts by both the volume and share thresholds.

Demo Store A | Chicken

High priority
30 portions remain at 17:00. Highest demand in 5 recent evenings was 17, leaving 13.0 portions above that baseline.

Manager check: validate live orders, walk-in demand and food-safety constraints before deciding whether any intervention is appropriate.

Demo Store A | Guacamole

High priority
16 portions remain at 17:00. Highest demand in 5 recent evenings was 8, leaving 8.0 portions above that baseline.

Manager check: validate live orders, walk-in demand and food-safety constraints before deciding whether any intervention is appropriate.

Evidence explorer

9 components shown
StoreComponentRemaining at 17:00Recent mean demandHighest recent demandConservative floor vs baselineStatus
Demo Store AChicken3015.85 shifts1713.0High priority
Demo Store AGuacamole167.65 shifts88.0High priority
Demo Store BRice3324.85 shifts267.0Monitor
Demo Store BGuacamole129.05 shifts102.0Low
Demo Store BChicken2119.05 shifts201.0Low
Demo Store CChicken2017.85 shifts191.0Low
Demo Store CGuacamole108.45 shifts91.0Low
Demo Store CRice2320.85 shifts221.0Low
Demo Store ARice2220.45 shifts220.0Low
No components match the selected filters.

AI and integration boundary

1. Deterministic screening

Python validates the snapshot and calculates every flag from visible comparison data. No model assigns risk.

2. Structured handoff

The same results are emitted as JSON for a future BI, Qlik Cloud or MCP-connected workflow.

3. Optional Claude brief

An LLM may rewrite manager-facing narrative only after classification, with required human-decision guardrails.

Manager briefs

Demo Store A

Demo Store A: review possible surplus for chicken and guacamole before close. Chicken: 30 portions remain, highest demand across 5 recent comparison evenings is 17, leaving a conservative surplus floor of 13.0; Guacamole: 16 portions remain, highest demand across 5 recent comparison evenings is 8, leaving a conservative surplus floor of 8.0. Before acting, check live orders, in-store demand and food-safety constraints. Any surplus listing remains a manager decision.

Demo Store B

Demo Store B: no high-risk flag, but monitor rice. Rice: 33 portions remain, highest demand across 5 recent comparison evenings is 26, leaving a conservative surplus floor of 7.0. Before acting, check live orders, in-store demand and food-safety constraints. Any surplus listing remains a manager decision.

Demo Store C

Demo Store C: no material surplus flag at the 17:00 checkpoint. Before acting, check live orders, in-store demand and food-safety constraints. Any surplus listing remains a manager decision.