One-bar queue dashboard
// What
Single PoE camera + tablet for the head bar, real-time queue length and refusal-risk score.
// Why
Proves the loop end-to-end on one decision point before you spend on the rest.
Change my answers
// Playbook · venue-events
Independent venues, bars & event spaces
AI in venues works best when it's giving humans better windows onto the floor — not replacing them. The wins are in queue management, capacity awareness, ticket fraud detection and post-event analysis. The losses are in 'AI bartender' and predictive staffing models that don't know about your local football fixtures.
// Starter projects · sequenced
Sequenced. Sized. Honest.
From a confidence-building first step to a transformative bet.
Capacity + flow heatmap
// What
Add cameras at entry/exit and 2–3 internal choke points; floor manager gets a live heatmap.
// Why
Lets you act on emerging crush situations before they're emergencies.
Operational signal fabric
// What
Cameras + EPOS + ticketing + weather as a single 'state of the venue' that drives signage, staff alerts, and post-event learning.
// Why
Once you have the substrate, every new question is days of work, not months.
// Casework that grounds this
Built it for them.
CASE / 01
Sentinel — bar-queue & ambient retail watch
Ayr Pavilion (1,500-cap event venue)
Edge cameras + small object-detection model giving bar staff a live read on queue length, refusal risk and crowd density.
CASE / 02
Pavilion Festival — one CMS, 30 dynamic screens
Multi-stage festival operations
A single event-state JSON drives 30 cinematic signage layouts across 6 stages — change a set time once and it cascades everywhere.
CASE / 03
Camera-vision-enrichment — vendor-neutral scene labels
Multi-site operators with mixed-vendor camera estates
A Viseron-shaped service that watches any RTSP stream and emits structured labels (people, vehicles, animals, parcels) for downstream automation.
// Risks worth flagging
- Auto-staffing models fail spectacularly on irregular events (cup finals, weather, school holidays). Forecast as input; humans rota.
- Edge inference dodges the GDPR conversation; cloud inference re-opens it. Choose carefully.
- Cameras pointed at customers will produce questions. Have an answer ready before the first one's installed.
// Watch out for in pitches
- Resist 'AI table-allocator'-type pitches; they don't survive a busy Friday.
- Don't pay per-camera SaaS fees that scale with your estate; build the substrate once.
// The honest take
The venues winning with AI are doing dull, narrow things very well. They're not chatbotting their bar.