IRIS
Multimodal AI for Retail Store Operations
Retail's operating layer, running on WhatsApp: photo-based visual validation, voice-note ticket routing, and AI-answered Google reviews.
- Stores · 10+ brands
- 160+
- Photo to verdict
- <10s
- Google listings managed
- 196
The problem
Running 160+ stores and 10+ brands with a lean team means nobody can be everywhere. Brand standards degrade silently, maintenance issues travel through ownerless WhatsApp threads, and hundreds of Google reviews sit unanswered.
How it works
No app to install, no training: the store manager sends a photo or a voice note on WhatsApp. AI validates visual merchandising against brand standards in under 10 seconds, classifies the voice note and routes the ticket to the right team, and drafts on-brand replies across the group's 196 Google listings. Supervisors see everything on a dashboard with a health score per store.
Inside the system.
My perspective
IRIS is part of the software we build inside Yáneken. My perspective starts with the operation I lead: a store team needs a useful answer and a clear owner for an issue, in the channel it already uses. That is the problem the system is built around.
The design decision
Start with WhatsApp. Introducing another application would make adopting the interface a prerequisite for solving the operating problem. Photos and voice notes already travel through the stores’ working day; IRIS gives those inputs standards, routing and follow-through.
What the result shows
The project record reports a photo-to-verdict time below ten seconds and operation across the group’s store network. It also covers replies across 196 Google listings. Those figures describe speed and coverage. The accompanying essay explains why adoption is the measure that matters to the operator.
How to read the evidence
Coverage alone does not establish how often every store uses the system. This case does not publish a before-and-after adoption or compliance rate. Serious themes in negative reviews are escalated to a person rather than answered automatically.
The path of a decision
- 01
A store sends a photo or voice note through WhatsApp.
- 02
The system checks the photo against brand standards or classifies the issue in the voice note.
- 03
The right team receives the ticket and its SLA; supervisors see compliance and store health in a dashboard.
The thinking behind the work, in my essays:
What it does
- 01
Visual validation
Photo to verdict in under 10 seconds, against each brand's implementation standards.
- 02
Voice-note routing
Voice notes transcribed and classified into maintenance, checklist or commercial tickets, with SLAs.
- 03
Review agent
Replies in each brand's voice — 157 style profiles — with theme-based escalation of negative reviews.
- 04
Per-store health score
Weighted composite of visual, zonal and commercial compliance, with traffic-light alerts.
Deployment scope
In production across the group's stores in Chile, over WhatsApp — no app to install, no prior training.
Related projects
- ChispezaVoice AI for Retail Hiring and Sales Training
- CifraAI Cash-Flow Intelligence for Multi-Entity Retail
- DelphineML Demand Forecasting for Footwear Retail
Status and figures: August 2026
If you recognize this problem in your business, let’s talk. For an implementation with my team, the advisory work lives at Menlo & Oak.