What a AI Smart Office Management Tool actually does
Predicts office occupancy patterns and suggests optimal desk locations based on meeting calendar and peer proximity, reducing friction in hot-desk booking.
The hot-desking SaaS market is dominated by Envoy ($109+/mo per location), Robin ($3/user/mo), OfficeRnD, and Skedda. These platforms handle desk booking beautifully — responsive UIs, mobile apps, visitor check-in workflows. But they lack the AI intelligence layer: predicting 'tomorrow will be 60% occupancy, book Conference Room A now' or 'You have a 2pm meeting with Maria in product — book the desk next to her usual zone for better collaboration.' This implementation focuses on the AI sidecar: (1) occupancy-pattern forecasting (Prophet / NeuralProphet on historical booking + badge-entry data), (2) smart-desk suggestions ('You have a team meeting 2–4pm; book a desk near the collaboration zones'), (3) environmental-sensor anomaly alerts (CO2 thresholds trigger escalations), and (4) meeting-room demand forecasting. The white-label angle: workplace-ops consultants can integrate this AI layer via Skedda / Robin API, charge $500–$1500/mo for the AI service, and keep the booking UX clean (customers still use Skedda/Robin as primary). COGS is <$100/mo even at 5 locations.
AI capabilities involved
Occupancy-pattern prediction for capacity planning
Smart-desk suggestion from calendar + collaboration patterns
Visitor-pre-arrival info pre-generation (brief for host)
Environmental-sensor anomaly alerts (CO2, temp, humidity)
Meeting-room demand forecasting
Who uses this
- Workplace-experience consultants managing 5–30 office locations for mid-market / enterprise clients
- Corporate real estate teams optimizing space utilization across distributed offices
- Coworking operators running a side business selling workplace intelligence to their memberships
SaaS alternatives on the market
Real products you can sign up for today — with current 2026 pricing, honest pros and cons.
Envoy
Enterprise companies (500+ employees, 5+ locations) wanting a polished desk-booking + visitor-management platform.
Free trial for 14 days
$109/mo (Standard per location)
Custom pricing for 50+ locations
Pros
- +Beautiful mobile app for desk booking and visitor check-in.
- +Real-time occupancy dashboard (see occupied desks on map).
- +Integrations with Slack, Microsoft Teams, Outlook.
- +Visitor management + badge printing integration.
Cons
- −No occupancy prediction or smart-desk suggestions — pure booking UX.
- −Expensive per location ($109+/mo); scales linearly with office count.
- −No white-label for consultants; designed for end-user enterprises.
- −Limited customization of desk-booking workflow.
Robin
Mid-market companies (100–500 employees) with flexible seating and want an affordable per-seat model.
$3/user/mo (per employee with booking access)
Custom pricing for 500+ employees
Pros
- +Cheap per-user model scales with team size.
- +Integrates with Outlook, Google Calendar, Slack.
- +Mobile app with Bluetooth for seamless check-in.
- +Room-booking + desk-booking in one platform.
Cons
- −No occupancy forecasting or smart suggestions.
- −Per-user pricing means monthly cost grows with headcount; unpredictable for hot-desking scenarios.
- −Limited visitor management compared to Envoy.
- −No white-label tier.
Skedda
Workplace consultants wanting to integrate their own AI layer via API. Best platform for white-label resale.
Free plan (up to 2 locations, limited features)
$79/mo per location (Standard)
$249+/mo per location (Premium with API access)
Pros
- +Affordable ($79/mo per location for standard bookings).
- +API available (Premium tier); allows integrations + AI sidecars.
- +Visitor booking + resource booking (rooms, parking, equipment).
- +White-label option available — consultants can rebrand.
Cons
- −Lower-touch UX compared to Envoy/Robin (more admin-focused than user-friendly).
- −No occupancy prediction or AI features built-in.
- −Premium tier ($249/mo) adds cost if you need API access.
- −Limited mobile app polish.
OfficeRnD
Coworking operators and flex-office providers managing memberships + day passes. Not ideal for enterprise desk-booking.
$179/mo per location (Core)
$349+/mo per location (Premium)
Pros
- +All-in-one: space booking + analytics + membership + billing.
- +Strong for coworking spaces (memberships, day passes, billing integration).
- +Occupancy analytics (shows usage trends).
- +Mobile app for seamless check-in.
Cons
- −Expensive ($179+/mo per location).
- −Occupancy analytics are historical; no predictive forecasting.
- −No API for AI integration; limited customization.
- −Overkill if you just need desk booking (includes coworking billing).
The AI stack
The core pipeline ingests Skedda/Robin booking history + badge-entry data (optional, from door-lock systems), trains a simple forecasting model (Prophet for baseline, NeuralProphet for fidelity) to predict occupancy, and on-demand suggests optimal desk locations based on meeting calendar + peer proximity. The architecture is deliberately lightweight: Prophet runs once per week (12 hours to retrain), suggestions are synchronous (generated per booking request). For a workplace consultant managing 5–30 locations, total compute cost is <$100/mo.
Occupancy forecasting (time-series prediction)
Predicts % office occupancy tomorrow and next 7 days, enabling proactive meeting-room booking and 'capacity alerts' when occupancy exceeds thresholds.
Prophet (Facebook's time-series library, free)
$0 (OSS library) + ~$25/mo Fly.io compute for weekly retrainingMVP; straightforward occupancy patterns. 1–10 locations.
NeuralProphet (PyTorch-based; free OSS)
$0 (library) + ~$50/mo Fly.io GPU-enabled worker for trainingProduction; locations with complex occupancy patterns (e.g., 'Occupancy spikes on project-deadline weeks'). 10+ locations.
Simple moving average (DIY baseline)
$0 (1-liner in Python)Prototype / proof-of-concept. Replace with Prophet once you have 2+ months of data.
Our pick: Start with Prophet on Fly.io $25/mo ($0.83/day = negligible). Train weekly on Skedda booking data (# bookings per day). Once you have 3+ locations with 90+ days data, graduate to NeuralProphet.
Smart-desk suggestion (LLM + meeting context)
When a user books a desk, suggests optimal location: 'You have a 2–4pm meeting with marketing team — book desk near the marketing cluster (D-Row).'
Claude Haiku 4.5
$1 / $5 per M tokensDefault choice; balances cost and speed for desk suggestions.
GPT-5.4 nano
$0.20 / $1.25 per M tokensHigh-volume suggestions; acceptable to have 1–2% poor suggestions.
Rule-based logic (DIY)
$0Prototype; replace with LLM once you want to scale to diverse meeting types.
Our pick: Use Haiku 4.5. Cost per suggestion is ~$0.00002 (~50 input tokens + 20 output, 1/50K daily bookings = $0.00002/day × 5 locations = $0.0001/day). Over a month: ~$3 across 5 locations.
Environmental-sensor anomaly detection (CO2, temp, humidity)
Monitors IoT sensors (CO2, temperature, humidity) in each zone. Alerts facilities team: 'Zone B CO2 exceeded 1,200 ppm; recommend ventilation check.'
Isolation Forest (scikit-learn, free OSS)
$0 (library)MVP; simple anomaly detection.
XGBoost regression (self-hosted, free)
$0 (library) + ~$15/mo Fly.io computeProduction; locations with significant environmental variance.
Our pick: Use Isolation Forest for MVP. Tune the contamination parameter to 2–5% (expect 2–5% of sensor readings to be 'anomalies'). Once you have 30+ days of sensor data, retrain with XGBoost for better drift detection.
Skedda/Robin API integration
Pulls real-time booking data (who booked what desk, when) and meeting-room calendar for forecasting + suggestions.
Skedda API (Premium tier $249+/mo per location)
$249+/mo (Skedda premium tier) + free API calls (within rate limit)Skedda customers who want to add AI layer.
Robin API (contact sales for access)
Contact sales; typically included in enterprise plansRobin customers with enterprise contracts.
Our pick: Skedda is the most transparent and consultancy-friendly. Recommend Skedda Premium ($249/mo per location) + your AI sidecar ($500+/mo). Customer gets booking (Skedda) + AI (yours) for ~$750/mo total per location.
Storage + scheduling (Supabase + Fly.io worker)
Stores booking history, sensor data, occupancy forecasts, and desk-suggestion logs. Fly.io worker retrains forecasts weekly.
Supabase Pro ($25/mo) + Fly.io ($15/mo worker)
$40/mo totalDIY; default choice.
Our pick: Supabase Pro + Fly.io worker. Total infra cost: $40/mo (amortized across all client locations).
Reference architecture
The system pulls booking + calendar data from Skedda/Robin API daily, trains a Prophet occupancy forecaster weekly on Fly.io, stores predictions in Supabase, and surfaces desk suggestions via LLM (Haiku 4.5) when users book. Optional sensor integration (CO2 monitors, badge readers) feeds into anomaly detection. The main engineering challenge is multi-location support: each office has different occupancy patterns, so forecasts must be per-location and seasonality must account for office-specific events (e.g., 'HQ has company all-hands first Friday of month').
Daily: Sync booking data from Skedda/Robin
Scheduled Fly.io worker → Skedda/Robin API → SupabaseCron job (daily 2am UTC) calls Skedda API to fetch all bookings for past 24 hours. For each booking: (location_id, date, num_desks_booked, meeting_rooms_booked). Stores in `daily_occupancy` table.
Weekly: Retrain Prophet forecasts per location
Scheduled Fly.io worker → Prophet → SupabaseWeekly job (Monday 6am UTC) queries `daily_occupancy` for past 90 days, groups by location, trains separate Prophet model per location. Generates forecast for next 7 days (% occupancy by day). Stores in `occupancy_forecast` table.
On-demand: Suggest desk location when user books
Skedda booking webhook → Lovable function → Claude Haiku + SupabaseWhen user clicks 'Book desk,' Skedda webhook triggers. Function retrieves: (1) user's meeting calendar for that day (via Outlook/Google Calendar API), (2) list of team members also booked for same day (from Skedda), (3) desk cluster locations (e.g., 'A-Row: engineering', 'B-Row: marketing'). Calls Haiku 4.5 with prompt: 'User [Name] has meeting with [Team] at [Time]. Suggest desk near [Cluster]. Available desks: [List].' Returns suggestion: 'Book D-24 (near marketing team, quiet zone).' Displays in Lovable UI.
Optional: Monitor environmental sensors (CO2, temp, humidity)
IoT sensor webhook → Isolation Forest → SupabaseIf office has smart sensors (e.g., CO2 monitors in each zone), sensor data streams to Supabase. Isolation Forest runs hourly to detect anomalies. If anomaly detected: create `alert` record and notify facilities team via Slack webhook: 'Zone B CO2 exceeded 1,200 ppm.'
Dashboard: Occupancy forecast + utilization trends
Lovable UI + rechartsFacilities team sees: (1) occupancy forecast chart (next 7 days % occupancy), (2) historical utilization heatmap (time of day vs. occupancy), (3) most-popular desk clusters, (4) environmental sensor stats, (5) recent alerts.
Estimated cost per request
~$0.00002 per desk suggestion (Haiku 4.5: ~50 input + 20 output = $0.00002); ~$0.001 per occupancy forecast (Prophet training is weekly, amortized).
Cost calculator
Drag the sliders to model your actual usage. The numbers update in real time so you can stress-test economics before writing a single line of code.
This calculator models monthly COGS for a workplace-ops consultancy selling AI smart-office packages to mid-market / enterprise clients. Assumptions: each location has Skedda Premium ($249/mo); the consultant layers on occupancy forecasting + desk suggestions + optional sensor monitoring. Fixed costs cover Supabase + Fly.io; per-location costs are minimal (Haiku 4.5 is cheap).
Estimated monthly cost
$40.00
≈ $480 per year
Calculator notes
- This model assumes Skedda Premium ($249/mo) is paid by customer; consultant adds AI layer on top.
- Haiku 4.5 cost per suggestion is negligible (~$0.00002). Total monthly LLM cost across 5 locations: ~$0.10–$0.50.
- Prophet retraining is weekly, cost amortized in Fly.io $15/mo (included in fixed costs).
- Optional sensor integration (CO2 monitors, badge readers) adds ~$50–100/mo hardware cost per location (not included here).
- Supabase Pro supports ~20–50 locations before hitting query limits. At higher scale, upgrade to Team ($599/mo) or self-host PostgreSQL.
Build it yourself with vibe-coding tools
In 1 weekend with Lovable Pro, you'll have a working MVP that pulls Skedda booking data, runs a simple occupancy forecast (Prophet), and suggests desk locations. By Sunday night, you'll have a demo showing a facilities team how AI can optimize their 5-location portfolio.
Time to MVP
12–16 hours (1 weekend)
Total cost to MVP
$25 Lovable Pro + $60 API credits (Claude Haiku, Prophet compute)
You'll need
Starter prompt
Build a workplace-management AI tool with these features: 1. Authentication: Supabase Auth. Each customer (workplace operator) is a separate tenant. 2. Skedda integration: Form to enter Skedda API key. On submit, fetch all locations + past 30 days of bookings via Skedda API. Store in Supabase `bookings` table (location_id, date, num_desks_booked, num_rooms_booked). 3. Occupancy forecasting: After fetching bookings, trigger async job to train Prophet model on `bookings` data. Forecast next 7 days' occupancy (% of total desks booked). Store forecast in `occupancy_forecast` table. 4. Dashboard: Show occupancy forecast chart (next 7 days). Heatmap of 'time of day vs. occupancy' (e.g., 'Monday 9am is always 80% booked'). 5. Desk suggestion (optional for MVP): When user selects a location + date to book, show 'Suggested desks' dropdown. For now, just list available desks; later add LLM suggestions. 6. Styling: Professional (blues, grays), clean layout. Tailwind + shadcn/ui. Database schema: - customers (id, customer_name, skedda_api_key, created_at) - locations (id, customer_id, location_name, skedda_location_id, total_desks) - bookings (id, location_id, date, num_desks_booked, num_rooms_booked) - occupancy_forecast (id, location_id, forecast_date, predicted_occupancy_pct) Edge Functions: - POST /fetch-skedda-bookings: takes customer_id + skedda_api_key, calls Skedda API, stores bookings - POST /train-forecast: takes location_id, trains Prophet on past 30 days, generates 7-day forecast, stores in DB - POST /get-occupancy-forecast: returns forecast for a location
Paste this into Lovable
Follow-up prompts (run in order)
- 1
Add Claude Haiku 4.5 desk suggestions: when user selects a location + date, fetch their Outlook/Google Calendar for that day (list meetings + attendees). Call Haiku with prompt 'User has meeting with [Team] at [Time]. Suggest desk near their team's usual cluster.' Return 1–2 suggestions with reasoning.
- 2
Add environmental-sensor integration: allow customer to input CO2 monitor endpoints. Fetch readings hourly; if CO2 exceeds 1,000 ppm, create an alert and notify facilities team via Slack webhook.
- 3
Add utilization trends: show which desk clusters are most popular by time of day / day of week. Heatmap: rows = desk clusters, columns = time of day, color = % utilization.
- 4
Upgrade forecast from simple Prophet (linear trends) to NeuralProphet (non-linear). This requires GPU training; use a Fly.io GPU-enabled worker (costs more but handles complex patterns).
- 5
Add multi-location management: support a customer managing 5–30 locations. Forecast is per-location. Show a 'Locations' dropdown to switch between forecasts.
Expected output
A working Lovable app where a workplace operator can: (1) authenticate with their Skedda Premium account, (2) pull booking history, (3) see occupancy forecast (next 7 days), (4) view occupancy heatmap by time of day, (5) (optional) get AI desk suggestions when booking. No production-grade sensor integration or enterprise SLA monitoring required — proof-of-concept.
Known gotchas
- !Skedda API has rate limits (1K requests/day free tier). If you're pulling hourly, you'll exceed limits. Solution: pull once per day (nightly) or upgrade Skedda to higher rate limits.
- !Prophet requires 30+ data points to train. If a location has <30 days of booking history, forecasts will be very noisy. Work around by using a simple moving average as fallback until you have 30+ days.
- !Calendar integration (Outlook/Google) requires OAuth + user consent. For MVP, ask user to manually paste meeting list (e.g., 'I have a marketing meeting at 2pm with 3 people'). Automate later.
- !Desk clusters (e.g., 'A-Row: engineering', 'B-Row: marketing') must be manually defined per location. There's no standard way to extract cluster info from Skedda. Store as a JSON config in Supabase `location_clusters` table.
- !Environmental sensors (CO2 monitors) require hardware + API. Not all offices have sensors. Make sensor integration optional; show alert message 'No sensors configured' if data is missing.
- !Prophet / NeuralProphet require Python; Lovable's built-in compute doesn't support Python. Solution: run forecasting as a scheduled Fly.io worker (cron job), store results in Supabase, and Lovable reads from Supabase.
Compliance & risk reality check
Smart office platforms handle employee location data (who booked which desk, badge-entry records), meeting schedules, and sensor data (CO2, temperature). This implementation must protect employee privacy (GDPR for EU employees, CCPA for California), ensure badge/biometric compliance (BIPA for Illinois), and maintain equitable access to desks (no algorithmic bias in desk suggestions).
GDPR / CCPA for employee location tracking (desk bookings + badge data)
Desk booking records and badge-entry logs are location data under GDPR (Art. 4(11)) and CCPA (Cal. Civ. Code § 1798.100). Tracking which employee booked which desk at which time can infer work patterns, health status ('She books conference room every Tuesday at 3pm for therapy'), and social networks. Processing this data without explicit consent is a violation.
Mitigation: Add an explicit consent banner: 'Your desk-booking history and badge-entry records are used for occupancy forecasting and desk suggestions. We do not track individuals; we analyze aggregate patterns.' Provide a right to deletion: employees can request erasure of their booking + badge records. For EU employees, host data in EU Supabase regions. For CCPA, add 'Do Not Track' button allowing employees to opt out of algorithmic suggestions. Store consent + opt-out status in Supabase `employee_consent` table.
Biometric data (BIPA Illinois, CUBI Texas) for face-recognition badge entry
If the smart-office system uses facial recognition for badge entry (e.g., 'AI recognizes your face, unlocks the door'), biometric data is subject to Illinois BIPA (740 ILCS 14/) and Texas CUBI (Tex. Bus. & Com. Code § 503.001). Both require explicit consent before collecting biometric data; fines are $1K–$5K per violation.
Mitigation: If using facial recognition: (1) Display explicit consent form before enrollment: 'You are enrolling in facial recognition for door unlock. Your biometric data will be stored securely and deleted if you leave the company.' (2) Obtain written signature. (3) For Texas, provide rights information per CUBI § 503.003. (4) Store biometric data in encrypted, access-controlled database (separate from booking data). (5) Implement auto-deletion: 90 days after employee departure, delete biometric data. For MVP, skip facial recognition and use badge/RFID instead (not biometric-regulated).
Algorithmic fairness in desk suggestions (no discriminatory clustering)
If the smart-desk algorithm clusters employees by department or meeting frequency, and those clusters correlate with protected characteristics (race, gender, disability), the algorithm could perpetuate discrimination. For example: 'Engineering cluster skews male' → algorithm suggests desks near engineering → new female engineer is never suggested engineering desks → hiring team perceives low integration → bias compounds.
Mitigation: Document desk-suggestion logic transparently: 'Desks are suggested based on meeting attendees and collaboration frequency. We do NOT consider department, gender, race, or any protected characteristics.' Audit suggestion data monthly: run a fairness check (e.g., 'Do women get equally suggested engineering desks as men?'). If bias is detected, retrain model or adjust thresholds. Store fairness audit logs in Supabase `fairness_audits` table.
Environmental-sensor data (CO2, temperature, humidity monitoring)
If sensors detect patterns (e.g., 'Zone B always has high CO2 Friday afternoons'), this can infer which departments work in which zones and which days are highest-occupancy. Combined with employee location data, this creates a detailed movement map.
Mitigation: Keep sensor data aggregated (zone-level, not individual-level). Never link sensor zones to specific employee names. Disclose sensor monitoring: 'We monitor air quality and temperature in shared zones to optimize comfort. Data is aggregated and not linked to individuals.' Allow employees to opt out of shared-zone sensor displays (they can see aggregate stats but not zone-specific CO2).
Visitor + badge data retention and deletion policies
Offices often retain badge-entry logs indefinitely for security. Over time, these logs combined with desk bookings create a complete movement history. Retaining this data indefinitely violates privacy best practices (GDPR principle of storage limitation, Art. 5(1)(e)).
Mitigation: Implement a retention policy: 'Badge-entry logs and desk-booking records are retained for 12 months. After 12 months, data is anonymized (no employee names, only aggregate counts like 'Zone A had 150 badge scans on Fridays') or deleted.' Add a scheduled job (Fly.io worker) that runs monthly to anonymize / delete old records. Notify employees: 'Your booking history is retained for 12 months for occupancy analysis.'
Build vs buy: the real math
8–12 weeks (for AI sidecar, assuming Skedda/Robin API integration)
Custom build time
$25,000–$45,000 (AI sidecar + Skedda/Robin integration)
One-time investment
3–6 months (5+ locations at $500–1500/mo per location)
Breakeven vs buying
The winning play is NOT building an all-in-one competitor to Envoy/Robin (that would be $50K+ and 6+ months). Instead: recommend Skedda Premium ($249/mo per location) for booking, then layer a DIY Lovable AI sidecar ($500+/mo per location) that adds occupancy forecasting + desk suggestions. Customer pays Skedda directly (~$249/mo), consultant charges for AI sidecar (~$500–1500/mo), total cost to customer is ~$750–1750/mo per location with ~85–95% gross margin for consultant. At 5 locations × $900/mo average = $4,500/mo revenue with <$500 COGS. DIY breaks even after 3–4 months. If you validate demand with 5+ locations, then invest in custom build ($25K–45K) to differentiate (proprietary NeuralProphet models, sensor integration, visitor forecasting).
Skip the DIY — RapidDev builds the production version
A Lovable MVP gets you a demo. Production needs auth that doesn't leak data, AI calls that don't bankrupt you, observability when models drift, and code you can audit. That's what we ship.
Discovery call (free)
30 minWe map your exact AI Smart Office Management Tool use case: who uses it, target volume, AI model choice, integrations, compliance scope. You get a detailed scope document and fixed-price quote within 48 hours.
AI-accelerated build
8–12 weeks (for AI sidecar, assuming Skedda/Robin API integration)Our engineers use Claude Code, Lovable, and custom tooling to ship 3–5x faster than agencies. You see weekly progress in a staging environment — not a black box.
Launch + handoff
1 weekWe deploy to your infrastructure, transfer the GitHub repo, set up CI/CD and monitoring, and train your team. You own 100% of the source code, prompts, and model configurations.
What you get
Timeline
8–12 weeks (for AI sidecar, assuming Skedda/Robin API integration)
Investment
$25,000–$45,000 (AI sidecar + Skedda/Robin integration)
vs SaaS
ROI in 3–6 months (5+ locations at $500–1500/mo per location)
30-min call. Fixed-price quote within 48 hours. No commitment.
Frequently asked questions
How much does it cost to build a smart-office platform?
A DIY Lovable MVP costs $25 Lovable Pro + ~$60 API credits (1 weekend). A custom AI sidecar (Skedda/Robin integration) costs $25K–$45K (8–12 weeks). Monthly infra: $30–60 (Supabase, Fly.io worker). If reselling at $500–$1500/mo per location, COGS is <$100/mo per location, yielding 85–95% gross margin. Breakeven on custom build: 3–6 months at 5+ locations.
Should I build a full booking platform or just an AI layer?
Build the AI layer only. Envoy, Robin, and Skedda already have beautiful booking UX. Competing with them would take 6+ months and $50K+. Instead, recommend Skedda Premium ($249/mo per location) for booking, add your AI sidecar ($500+/mo) for occupancy forecasting + desk suggestions. Customer gets best-of-breed UX + your proprietary AI.
Can RapidDev build this for my company?
Yes. RapidDev can build an AI sidecar integrating Skedda/Robin APIs. Typical scope: $25K–$45K over 8–12 weeks, covering Prophet/NeuralProphet occupancy forecasting, Claude desk suggestions, environmental-sensor integration, and multitenancy for 5–30 locations. Email seopartner@rapidevelopers.com for a free 30-min consultation.
How do I handle GDPR compliance for employee location tracking?
Add explicit consent: 'Your desk-booking records are used for occupancy analysis; we analyze aggregate patterns, not individuals.' Provide right to deletion: employees can request erasure of their booking history. For EU employees, host data in EU Supabase regions. Store consent in Supabase; let employees opt out of AI desk suggestions.
Should I use Prophet or NeuralProphet for occupancy forecasting?
Start with Prophet (simple, fast, free). Prophet requires ~30 days of data and handles seasonality well (weekends vs. weekdays, holidays). Once you have 3+ locations with complex patterns (e.g., 'Friday occupancy fluctuates'), upgrade to NeuralProphet (learns non-linear trends). Cost difference: Prophet $0/mo compute vs. NeuralProphet ~$50/mo GPU worker.
How do I integrate with Skedda's booking API?
Skedda offers API access in Premium tier ($249+/mo per location). Get API key from Skedda dashboard. In your Lovable app, store the API key in Supabase (encrypted). Daily scheduled job calls Skedda API to fetch bookings: `GET /api/spaces/{space_id}/bookings?date_from=2024-01-01&date_to=2024-12-31`. Parse response and store in Supabase `bookings` table. Train forecasts from this table.
Can I add facial-recognition badge entry?
Yes, but beware of biometric regulations. Illinois BIPA and Texas CUBI require explicit written consent before collecting biometric data. Fines are $1K–$5K per violation. If adding facial recognition: (1) Display consent form, (2) Get written signature, (3) Store biometric data in encrypted, separate table, (4) Auto-delete after employee departure. For MVP, skip facial recognition and use RFID badge (not regulated as biometric).
How many locations can one system support?
Supabase Pro ($25/mo) supports ~20–50 locations before hitting query limits. Each location's forecast is trained separately (one Prophet model per location). At 50+ locations, upgrade to Supabase Team ($599/mo) or self-host PostgreSQL. Most workplace consultants start with Pro and upgrade once they have 15+ paying clients.
What's the cost to the workplace operator (my customer)?
Skedda Premium: $249+/mo per location (customer pays Skedda directly). Your AI sidecar: $500–$1500/mo per location. Total: ~$750–$1750/mo per location. At 5 locations, customer's bill is ~$3750–$8750/mo. You keep ~95% of the AI layer revenue as gross margin.
Can I use environmental sensors without a booking system?
Yes. If you just want to monitor CO2, temperature, humidity for facilities management (no desk booking), skip Skedda integration. Just pull sensor data hourly, run anomaly detection (Isolation Forest), and alert facilities team if thresholds exceeded. This is a lightweight MVP that can be bolted onto any existing booking system.
Want the production version?
- Delivered in 8–12 weeks (for AI sidecar, assuming Skedda/Robin API integration)
- You own 100% of the code
- AI cost monitoring built in
30-min call. No commitment.