Why Your Smart Building IoT Should Skip WiFi

By Robin Chien | IPMVP-Certified | 15+ Years Managing $1.2B in Commercial Real Estate

The most counterintuitive lesson from 10,000+ building deployments has nothing to do with algorithms, machine learning, or control theory. It is about connectivity.

WiFi does not work in the places where your building's most critical equipment lives.

๐Ÿค–

Run this analysis on your building

Our AI agents use the same methodology. First query free โ€” no credit card. ย  ย Try the Agent โ†’

The Problem Nobody Talks About in Sales Demos

Mechanical rooms are typically in basements. They are surrounded by concrete walls, steel beams, metal ductwork, and heavy equipment. WiFi signals attenuate through every one of those materials. A signal that works perfectly in a vendor's demo room on the third floor drops to zero bars in the boiler room two floors below grade.

I have watched this scenario play out multiple times across my deployments. A vendor installs beautiful IoT sensors and controllers, connects them to the building's WiFi during commissioning, and everything works. Two weeks later, the building's IT team changes a network configuration, or the access point in the utility corridor loses power, or the concrete-and-metal environment simply degrades the signal below usable thresholds. Data gaps appear. The AI system loses its eyes and ears. Optimization stops.

The root cause is not the AI technology. It is the assumption that building WiFi infrastructure --- designed for human occupants on occupied floors --- will reliably serve machine-to-machine communication in the harshest RF environments in the building.

What 10,000 Buildings Taught Runwise

Runwise solved this problem by eliminating WiFi from the equation entirely.

Their sensors use embedded Verizon SIM cards on a patent-protected wireless mesh network that operates independently of building WiFi. Each sensor connects directly to the cellular network. No access points to maintain. No building IT coordination required. No dependency on infrastructure you do not control.

The specifics matter:

The result: installation time drops to less than one day. No construction required. The total cost is approximately 20x cheaper than traditional BMS wiring.

That last number is worth repeating. Twenty times cheaper. Because you are not pulling cable, cutting drywall, or coordinating with building electricians. You are sticking battery-powered sensors on walls and walking out.

When WiFi Works Fine

I am not arguing that WiFi is universally wrong for building IoT. There are scenarios where it works well:

Modern commercial buildings with strong infrastructure everywhere. If your building was constructed or renovated in the last 10 years and has enterprise-grade WiFi with access points in utility spaces, mechanical rooms, and vertical shafts --- WiFi-based IoT can work. The keyword is "everywhere," including the spaces humans do not normally occupy.

Above-grade installations only. Sensors monitoring occupied spaces on typical floors --- air quality monitors, occupancy sensors, thermostat-level devices --- can rely on the same WiFi infrastructure that serves occupants. The signal environment is fundamentally different from a below-grade mechanical room.

Buildings with dedicated IoT network segments. Some forward-thinking building operators have deployed separate SSIDs and VLANs specifically for IoT devices, with access points positioned for machine coverage rather than human coverage. This works but adds $15,000-$40,000 and 30-60 days to your deployment timeline.

When Cellular Is Mandatory

For the following scenarios, cellular connectivity is not a preference --- it is a requirement:

Older multifamily buildings. Pre-war and mid-century apartment buildings in New York, Boston, Chicago, and similar markets have thick concrete and masonry construction that makes WiFi propagation extremely difficult. These buildings represent the majority of the addressable market for energy optimization.

Basement and sub-grade mechanical rooms. This is the most common failure point. If your boiler or chiller plant is below grade, test WiFi signal strength at the exact equipment location before committing to any WiFi-dependent vendor. Bring a phone. Check the signal. If it is weak, it will be unreliable for IoT.

Retrofit deployments in occupied buildings. Running new WiFi infrastructure through occupied residential or commercial spaces requires permits, tenant notification, access scheduling, and construction coordination. Cellular eliminates all of this.

Multi-building portfolios with inconsistent IT infrastructure. If you manage 20 buildings with 20 different WiFi configurations maintained by 20 different contractors, cellular gives you a single, consistent connectivity layer across the entire portfolio.

The Practitioner's Rule

Before committing to any AI-HVAC vendor, do this:

Walk into the mechanical room. Pull out your phone. Check the cellular signal and the WiFi signal. If WiFi is weak or absent, your deployment plan just changed.

This is a 5-minute exercise that can save you 60 days of troubleshooting and $15,000+ in WiFi infrastructure upgrades you did not budget for. I have seen multiple pilots stall for months because the team assumed WiFi would work in spaces where it physically cannot.

The Broader Lesson: Horizontal Architecture

The connectivity question is actually a data architecture question in disguise.

Your building data --- temperature profiles, equipment performance, energy consumption, occupancy patterns --- should not depend on any single connectivity method, any single vendor, or any single platform. A horizontal data architecture means your sensors can talk to your AI system via cellular, WiFi, BACnet, or any other protocol, and your data remains yours regardless of which path it takes.

When evaluating vendors, ask:

  1. What happens to my data if the connectivity method fails? Is there local buffering? How much data loss occurs during an outage?
  2. Can I switch connectivity methods without replacing hardware? True horizontal architecture means the sensor is independent of the transport layer.
  3. Who owns the connectivity relationship? If the vendor manages the cellular SIM cards, what happens to your sensors if you switch vendors?

The vendors who answer these questions well are the ones building for the long term. The ones who cannot are building lock-in.

Assess Your Building First

Connectivity is one of 15 factors in the building readiness assessment I use before recommending any AI-HVAC pilot. Getting it wrong does not just delay the project --- it can kill internal confidence in the entire technology category.

Building assessment framework: The AI-HVAC Pilot Guidebook, Chapter 3 includes the full 15-point readiness checklist.

Building Assessment Checklist: Download the checklist [coming soon]

Robin Chien is an IPMVP-certified practitioner with 15+ years managing energy optimization across a $1.2B portfolio. He writes at ai-smart-buildings.com.