Internal AI Systems · Willie's Grill & Icehouse
The restaurant
runs on autopilot
before you wake up.
WAIOS is the AI operating system quietly running behind Willie's — reading the numbers, watching the guests, and writing the brief every morning.
Live in production
21 locations
Built by Diego Velasquez
n8n + Claude
The problem
Every morning, the same scramble.
Sales sat in one system. Reviews in another. Industry news, weather, loyalty signups, app complaints — all scattered, all manual. Pulling it together ate hours every week and still left blind spots. By the time anyone had the full picture, the day was half gone.
6+
systems to check by hand every morning
20+
hours of manual pulling & formatting per week
1
person expected to see all of it at once
The shift
One system reads everything,
and hands you the answer.
WAIOS is not a chatbot. It is a network of autonomous workflows that pull from every source on their own, run the data through Claude for judgment, and deliver a single decision-ready brief. The work happens whether anyone logs in or not.
Collectsales · reviews · news · weather
→
ReasonClaude scores & synthesizes
→
Deliverone brief, every morning
What's running today
The live system.
Intake & synthesis
Intelligence Collector
News Aggregator
Tech Scout weekly
Mirus Sales Parser source paused
Daily Intel Email flagship
Weekly Digest
Guest experience
Tattle Guest Experience
Guest Experience Agent
Agent Orchestrator
Interactive & support
WAIOS Chat
App Support Bot
Punchh Loyalty Import
NEXUS Data Feed
Self-monitoring
Error Alert Handler
Weekly Health Audit
Monthly Prune
Master Brain memory
~15 live workflows plus infrastructure, all feeding a single daily output.
A day in the life
By 8am, the thinking is done.
3:15 AM
Industry & AI news pulled and de-duped against what's already been seen
6:30 AM
Overnight sales parsed into structured numbers
7:30 AM
Guest-experience agent reads Tattle reviews, flags at-risk locations
8:00 AM
The brief lands — KPIs, priority call-out, guest pulse, weather, yesterday vs last year, watch list
No one pulled a report. No one opened a dashboard. The synthesis was waiting.
The daily output
What actually lands in the inbox.
📋 The brief leads with a decision
- Claude reads the full picture and writes the one thing to pay attention to today
- Header KPIs: system-wide sales vs last year, open guest incidents
- A watch strip for anything trending the wrong way
🍽️ Then the supporting layers
- Guest feedback pulse from Tattle — yesterday's low-star reviews surfaced
- Weather, yesterday's numbers, period-to-date
- Industry stories, competitor moves, restaurant-tech radar
Goes to Diego and the team. One read, whole business.
Guest experience, watched daily
A bad review can't sit for a week anymore.
Every day WAIOS authenticates into Tattle, pulls the last 7 days of metrics and yesterday's low-star reviews, and runs them through Claude to name the real themes and flag which locations are slipping. It writes findings the daily brief reads back to you automatically.
7-day
rolling guest-metric window pulled every run
Daily
low-rated reviews surfaced, not buried in a portal
Auto
at-risk locations flagged by the agent, per location
Beyond the email
Ask it anything. It answers guests too.
💬 WAIOS Chat
- A web app that answers plain-English questions about the business
- Backed by the same live data the daily brief runs on
- Live at waios-chat.pages.dev
📧 App Support Bot
- Reads incoming Willie's app support emails
- Drafts and handles responses automatically
- Takes a recurring inbox chore off a person's plate
The value
What it's worth.
20+ hrs
manual reporting work removed every week
21
locations covered in one brief, daily
~3 mo
from zero to a production system
Every AM
decision-ready before the workday starts
$__
est. annual labor value of the hours saved
CONFIRM WITH DIEGO__%
faster response to a slipping-guest signal
CONFIRM WITH DIEGO
Solid numbers are verified. Dashed tiles are yours to fill with real figures before Tuesday.
How it's built
One repeatable pattern, run everywhere.
No engineering team, no custom platform. Every part of WAIOS follows the same shape, which is exactly why it scales to new data sources and new brands without a rebuild.
Triggerschedule or webhook
→
PullAPIs, email, sheets
→
ReasonClaude scores & writes
→
Storeshared memory
→
Deliveremail · chat · dashboard
Built on n8n for orchestration and the Anthropic Claude API for judgment. Self-monitoring, so it flags its own failures.
Where it goes next
The framework is ready. The data isn't connected yet.
Weather & LSM Intelligence
Ties local weather to local-store-marketing moves
BUILT · DORMANT
Ops Intelligence
Live ops signals — needs Mirus / PAR POS access
NEEDS DATA
Marketing & Loyalty
Deeper loyalty insight — needs Punchh direct API
NEEDS DATA
Vendor & Food Cost
Cost swings & margin watch — needs US Foods feed
NEEDS DATA
Facilities & Comfort
Dining-room comfort — needs thermostat APIs
NEEDS DATA
Each is a slot already wired into the orchestrator. Connect the source, the agent turns on.
The takeaway
Any brand with scattered
data can run on this.
WAIOS proves one operator plus AI can watch a 21-location business end to end, every day, and hand leadership the answer instead of the raw data. The pattern is not specific to Willie's. It is specific to how modern operations should run.
Running in production today
Diego Velasquez
D.Velasquez@williesrestaurants.com