Agent Portal · Overview
Fleet status

53% intelligence-ready across this cohort

System state first, agent configuration second — the real pipeline behind the Pennsylvania beverage-intelligence pilot.

System Health · run status, this interactive sample
53%
success98of 186
no_data0of 186
waiting32of 186
error56of 186
Attention
1,386

Restaurants flagged for human judgment, clustered by shared root cause — one decision can resolve several at once.

Open Judgment Queue
Recovery
32

Paused mid-run for a real environment reason (e.g. closed for the night) — resuming automatically from the last completed stage. Closed isn't failed.

Intelligence Coverage · every restaurant, every workstream
RestaurantQualifyPOSMenuCheckoutBundleImage
completewaiting (real env. reason)blocked / conflictednot started
Agent Setups · configuration, not system state
Online Ordering Validation
Active
5,279 qualified

Confirms each location has functional online ordering, excluding static menus. Real accuracy: 74% initial, retrained to 100% via full human validation.

From this pilot's real per-restaurant data (100% QA sample).

POS Platform Identification
Active
4,415 identified

Identifies the ordering system (Toast, Square, Slice, 50+ platforms) via checkout URL/DOM signature matching.

From this pilot's real per-restaurant data (5% cross-validated QA sample).

Beverage Assortment Extraction
Active
5,275 extracted

Captures every listed beverage — name, brand, sub-brand, price — including generic listings that resolve to a real brand via item variations.

From this pilot's real per-restaurant data (7% QA sample).

Upsell Detection
Not started
~19% of locations

Adds an item to cart and proceeds through checkout to see if a beverage upsell is prompted. Enterprise comparison: ~45%.

Reported in the pilot spec's own broader sample — not run on this 5,279-restaurant cohort yet.

Bundle Detection
Not started
~26% offer bundles

Identifies meal bundles containing beverages; of those, ~46% include Coca-Cola, ~36% PepsiCo, ~32% generic beverages only.

Reported in the pilot spec's own broader sample — not run on this 5,279-restaurant cohort yet.

Image Classification
Not started
~5% of items

Evaluates product images for presence, type, and content, across ~57,800 beverage products listed pilot-wide.

Reported in the pilot spec's own broader sample — not run on this 5,279-restaurant cohort yet.