Agentic Operating System for Appointment-Based Services
Celoria
The Agentic OS that lets salon chains scale — reversible tasks run automatically, irreversible tasks get human approval.
March 2026 · Confidential
Every competitor adds AI
on top of legacy tools.
We built the system AI-native
from day one.
The Problem
Scaling a salon chain breaks on three things no tool has solved.
Employees Can't Use Tools
Most workers have limited education. Complex dashboards are too hard — they call us constantly asking how to operate. (We support multilingual UI.)
SOP Falls Apart
Huge efficiency gap between best and average stores. SOPs rely on memory, never stick. Three disconnected tools cobbled together.
GM Only Reports, No Insights
$120K/yr GM only lists data, no insights. Owner still has to prep dashboards for employee reviews.
The problem isn't bad tools. It's that traditional tools are the wrong paradigm for this workforce.
AI Agent isn't a better tool — it's a different interaction model.
Paradigm Shift
Day-End Reconciliation
Traditional SaaS
1 Open finance module
2 Export today's transactions
3 Check receipts line by line
4 Find discrepancies
5 Manually fix records
1-2 min (no issues)
5-10 min (with issues)
...or staff forgets entirely → next day's open SOP can't run
Celoria Agent
1 Agent auto-reconciles at close
2 Shows only 3 anomalies + causes
3 Manager confirms or flags
5 seconds (no issues)
20-30 seconds (with issues)
The Solution
Agentic, not Autopilot.
The system decides when to act and when to ask.
Reversible tasks (reconciliation, anomaly fix, marketing) → auto-execute.
Irreversible tasks (scheduling, appointments) → assist humans.
The system judges which mode to use.
GUI doesn't disappear — it evolves from operations interface → approval layer → audit dashboard.
How It Works
Push-Based UX: The system finds people
Every business event triggers an intelligent to-do pushed directly to the employee's device.
✅ Checkout: Lisa Chen — Gel Manicure
Total: $85 + Tip: $15.30 = $100.30
Card on file: Visa ending 4242
Execute Standard
Customize
Key Innovation
Employee assists the system, not the other way around.
80% of operations → one-tap "Do As Suggested"
20% edge cases → AI generates options
Result: Zero learning curve for new employees — the system tells them what to do.
Product UX
Think Cursor, but for salon operations
Two Modes, One System
🤖 Agent Mode (95% of daily ops)
AI pushes decisions → human confirms
Day-end reconciliation, scheduling,
inventory alerts, marketing triggers
= "Accept all" to run your store
📝 Editor Mode (5% configuration)
Traditional SaaS GUI as fallback
Add services, design promotions,
configure SOPs, complex scheduling
= Full control when you need it
The Cursor Parallel
| Cursor | Celoria |
|---|
| Agent Mode | AI writes code,
human reviews | AI runs ops,
human confirms |
| Editor Mode | Manual coding
for edge cases | Traditional GUI
for configuration |
| Trend | Agent coverage
keeps growing | SOP rules keep
expanding |
Key insight: The disruption isn't replacing GUI with voice/NL.
It's eliminating the need to "log in and operate" at all.
A manager's day: 5 taps on phone
instead of 2 hours at a computer.
Architecture
6+1 Agent Groups × 15 Sub-Agents
🏢 Multi-Store Orchestrator
↑ coordinates across stores ↑
🎯 Supervisor Agent (per store)
↑ aggregates & escalates ↑
⚙️ Daily Ops
📈 Acquisition
❤️ Retention
📦 Inventory
👥 Talent & SOP
🛡️ Compliance
Appointment · Finance · Marketing · Loyalty · Schedule · Training · Performance · Compliance · ...
80% rules engine (deterministic, $0 LLM cost) + 20% LLM (edge cases only).
Each agent has a toolkit of API actions — 72 of 106 actions already built.
Category Validation
YC is batch-funding "AI OS for [Vertical]"
Bravi
Home Services
AI front-office + internal copilot for installers
YC F25
Carma
Fleet Management
Automates fleet ops, same-day service bidding
YC W24 · $5.5M Seed
Sandra AI
Car Dealerships
Replaces legacy DMS with AI-native platform
YC batch
Tensol
Hospitality
Replaces legacy hotel ERP end-to-end
YC batch
Brickanta
Construction
AI-native pre-construction workflows
YC batch
Celoria
Beauty & Wellness
$100B market. Legacy Zenoti. No AI-native player.
← The gap
The pattern is validated. The beauty & wellness vertical is wide open.
Competition
In beauty, everyone is adding AI to old systems
| Platform | Approach | Architecture | AI Depth |
|---|
| Zenoti | "AI Workforce" (Sep 2025) | 15-year monolithic | 6 independent agents, no coordination, no SOP state machine, no learning mechanism |
| Vagaro | AI booking assistant | Legacy SaaS | AI receptionist only |
| NailSoft | Nail-vertical POS | Legacy desktop | Vietnamese market standard, no AI |
| GlossGenius | AI business tools | Mobile-first SaaS | AI marketing copy |
| Tepali | AI-native OS for medspas | YC W26, modern | Similar direction, different vertical (medspas) |
| Celoria | AI IS the system | Event-driven + Agent-native | AI operates, human confirms |
Ethnic segmentation matters: Vietnamese salons (NailSoft) vs Chinese chains (Celoria) — different workflows, language, supply chains. No crossover tool exists.
"Adding a touchscreen to a Nokia doesn't make it an iPhone.
The architecture has to be AI-native from Day 1."
Unit Economics
Hybrid Architecture = Industry-Leading AI Margins
$35-48
AI cost per store / month
~12%
As % of $299 subscription revenue
Industry average AI SaaS: 15-30% of revenue
Why so low?
80% of events → rules engine
= deterministic, $0 LLM cost
20% edge cases → Claude Sonnet
= ~40 LLM calls/day/store
Flywheel: every human override
→ new rule → less LLM reliance
→ cost per store trends down over time
Defensibility
Three Compounding Moats
🧠 SOP Engine
Salon operations encoded as configurable state machines. Domain knowledge from 100+ real stores.
Anyone can call OpenAI. Nobody else has this rulebook.
🔄 Override Flywheel
Every time staff rejects AI → we learn a new rule → less LLM → lower cost → better accuracy.
Competitors can copy features, not 50 stores × 365 days of corrections.
🔒 Operational Dependency
Employees rely on Agent for SOP execution. Leaving the system means they don't know the full process.
The more you use it, the harder it is to leave.
Market Size
$300B+
TAM — Global appointment-based
service industry revenue
$1.1B
SAM — US 180K+ stores
× ARPU $523/mo
$226M
SOM — Chains with
2+ locations
Expansion Path
Wedge: Nail salon chains
Next: All beauty chains (spa, lash, hair)
Then: All appointment-based services
(med spa, fitness, clinics, auto repair)
Agentic Commerce protocols —
SOP engine is industry-agnostic by design.
Traction
Live in production. Not a prototype.
26K+
Transactions processed
Solo-built. Zero outside funding. 2 months from first deploy to 35 locations.
Agent layer ships on top of a battle-tested platform — 1,133 API endpoints, enterprise multi-tenant architecture.
Business Model
Two Revenue Layers
SaaS (including AI)
$299/location/mo
Payment Processing
2.3%/tx
Per-Store Unit Economics
SaaS revenue: $299
Payment margin: $224
Gross revenue: $523/mo
AI + server cost: -$35-48/mo
Gross margin: ~91%
Team
Eric He
CEO / CTO
MS Design (NYU) + MS CS (Georgia Tech)
Meituan exposure — Services AI at scale
Yonyou internship — Enterprise AI agents
Solo-built entire Celoria platform:
1,133 APIs, 4 frontends, AI integrations
Joey Fei
Co-founder / COO
18 years beauty industry
QQ Nails — NYC's largest nail chain
35 locations live, first customer
Equipment supplier to 100+ salons
= Built-in distribution channel
Daily ops transitioning to management team,
progressively shifting to full-time Celoria
Advisors
Advisory Board
Amazon Senior Developer
— weekly architecture reviews, 6+ months
Senior Developer @ major SaaS company
— weekly engineering mentorship, 6+ months
Industry veterans from Joey's network
— product validation, GTM strategy
First hire post-funding:
Founding Engineer
Go-to-Market
Built-in distribution, not cold outreach
NOW
QQ Nails 35 locations live — system proven in production, AI agents MVP in 8 weeks
NEXT
Joey's supply network — 100+ salons, warm intros, zero CAC. Expand NYC metro.
THEN
NYC → LA → beauty verticals (spas, lash studios, med spas). Replicable playbook.
Agent layer is the wedge for enterprise chains.
No competitor offers "your new store runs itself from Day 1" — that's the pitch that gets 20+ location chains to switch.
Execution
Agent MVP in 8 weeks. Demo in 4.
Week 1-2
Foundation
Event Bus
SOP Engine
DB Schema
Socket.IO channels
Week 3-4
First Agent
Appt→Payment flow
Push UX on iPad
"Do As Suggested"
← VC Demo ready
Week 5-6
Day Start/End
SOP checklists
Auto reconciliation
Tip distribution
Anomaly detection
Week 7-8
Supervisor
Daily briefing
Owner dashboard
Escalation engine
Full loop complete
Only 6 new APIs needed — 72 of 106 agent actions already built into the existing platform.
The Ask
We're raising to build
the AI operating system
for beauty & wellness.
Engineering
Hire CTO + 1-2 engineers.
Ship Agent MVP in 8 weeks.
Go-to-Market
Expand beyond QQ Nails.
Joey's 100+ salon network.
AI Infrastructure
LLM costs, SOP engine depth,
multi-store orchestration.
Your next store shouldn't need
a new manager to run it.
It just needs Celoria.
eric@celoria.app · celoria.app