The least intelligent object in a smart building may be the label on a controller. It is missing, misleading or written for the contractor who installed it two renovations ago. Somewhere above, a dashboard promises predictive maintenance. Down here, nobody is quite certain which box runs the west-side air handler. Switch Automation has spent more than a decade making a business out of that awkward distance.
The Denver-headquartered company sells an enterprise software platform for building owners and facilities teams. It connects building-management systems, HVAC controls, lighting, meters, sensors and work-order tools; translates their data into a common model; and shows operators what is wasting energy or drifting toward failure. From there, teams can benchmark a portfolio, prioritize work and, under defined guardrails, let software adjust the building itself.
The company says it now supports more than 6,500 buildings in at least 15 countries, including 2,100 financial-services sites, representing more than 200 million square feet. Those are company figures, but the shape of the business is easy to understand: Switch is not selling a clever thermostat. It is selling a way to operate many complicated properties as one system.
01The machine room before the machine mind
The origin is pleasingly accidental. Co-founders Deb Noller and John Darlington had built logistics and freight-tracking systems for mining companies. Later, an American automation product arrived in Australia and could not control Australian lighting. Darlington worked on the integration. Over coffee, he kept telling Noller that lighting could talk to security and other building systems. Eventually, on what she later called “the right day,” she listened.
Switch Automation was founded in Australia in 2012 and expanded into the United States the next year. Today Noller is chief executive and Darlington is chief technology officer. Their bet was that building owners would want the same thing logistics companies wanted: a consistent operational view across a distributed, untidy physical network.
“The first business that John and I created was a labor-based model.”Deb Noller, on the lesson that pushed Switch toward scalable software
That sentence explains the company better than a showroom full of sensors. A pure services firm can inspect equipment, reconcile point names and recommend repairs, one building at a time. It is useful work and a rough business to scale. Switch tried to encode the repeatable parts: connectors for many manufacturers, a semantic model that gives raw data context, reusable analytics, portfolio dashboards and automated workflows. Its engineers still help with onboarding and optimization. The software is the multiplier, not a denial that buildings need people.
02What failed first
In enterprise proptech, the first failure is often earlier than the demo. The building network may be unreachable. A gateway may drop data. A device may speak a protocol nobody expected. Equipment points may be badly named, or the owner may not have credentials. Noller has described how Switch evolved its approach after finding that some “smart” projects actually needed basic connectivity upgrades first.
That changed the sequence. Instead of marching directly into analytics, Switch developed Dx3, short for Digital Device Discovery. The tool scans networks, systems and devices, then reports on connectivity, network integrity, latency, potential security issues and readiness for real-time data collection. It is the preflight check for a building that would otherwise attempt takeoff with three unlabeled switches and a binder from 2004.
The move is strategically neat. Readiness is a smaller, legible purchase. It exposes the work that must happen before a broad transformation and puts buyer and vendor on the same side of the evidence. It also prevents an analytics platform from being blamed for data it never reliably received.
03The product is a portfolio habit
Switch’s buyers include commercial-property owners, banks, retailers, healthcare organizations, schools, hotels and government operators. The daily users are less abstract: facilities managers triaging alarms, energy teams comparing sites, property managers checking comfort complaints, engineers testing schedules and executives watching operating expense and carbon targets.
The platform combines a building data layer, digital twins, fault detection and diagnostics, energy management, alert analysis, benchmarking and automation. That places it above the local building-management system and below the executive scorecard. Traditional BMS software is commonly strongest inside one building or one manufacturer’s ecosystem. Switch’s claim is that an independent layer can accommodate old and new equipment from many vendors, then apply the same operating logic across a portfolio.
That is also where competition crowds in. Honeywell, Schneider Electric, Siemens and Johnson Controls sell broad building suites. Specialists including Clockworks Analytics, Facilio, BrainBox AI, KODE Labs, Metrikus and Radix IoT attack parts of the same stack. The familiar low-cost alternative remains the incumbent BMS, a spreadsheet, an energy consultant and whatever institutional memory has not yet retired.
Company-published cases give the value proposition some welcome edges. Fifth Third reported average monthly energy savings approaching 25 percent. Austec reported a 12.6 percent reduction in energy costs at Hyperdome Shopping Centre. An Oxford Properties case described $160,000 in savings over three months. In 2026, Switch said a six-property deployment with Cushman & Wakefield covered more than one million square feet, cut energy use by over 10 percent and reduced annual costs by more than $290,000.
These are examples, not a universal yield. Savings depend on the baseline, utility rates, equipment condition, operating schedules and whether anyone fixes what the system finds. The useful version of the product is a loop: detect, prioritize, act, verify. A dashboard that produces alarms nobody owns is merely an expensive weather report.
04So, what does it cost?
Switch does not publish a current enterprise price list. Its terms describe SaaS subscriptions, consulting, support, hardware and custom work orders. They also list US$250 an hour for onsite support not covered by an agreement. An older Australian performance-optimization sheet offered unusually specific fixed pricing: AU$750 per month for a small building and AU$1,500 for a large one, before tax.
The subscription is only one line in a realistic budget. Buyers may need gateways, network access, controls contractors, data tagging, integrations, commissioning and staff time. That makes return-on-investment math site-specific. One investor described a Seattle property where Switch found savings above 20 percent and argued that a million dollars of annual net operating income could add roughly ten million dollars of property value at the assumed capitalization rate. The arithmetic is compelling; the assumptions deserve their own meeting.
05The playbook worth stealing
Switch’s most portable ideas do not require buying Switch. They require respecting the order in which physical technology becomes useful.
Sell the prerequisite
Measure connectivity, access and data quality before promising a transformation. A readiness score turns an argument into a work list.
Start with installed reality
Integrate equipment that still works. Rip-and-replace makes the project larger before it makes the outcome better.
Name things once
Create a common asset model so the second, tenth and hundredth building can reuse analytics instead of restarting the taxonomy.
Verify the fix
Do not count an alert as an outcome. Track whether a work order or control change removed the fault and sustained the saving.
There is a fifth idea for enterprise founders: align the product with the customer’s internal sequence. Facilities may own the machines, IT the network, sustainability the targets, finance the budget and procurement the contract. Smart-building sales drag because the buyer is a committee assembled from people who use different nouns. A discovery product gives that committee one shared artifact.
When the approach will not work
- A small standalone site has too little waste or complexity to justify integration.
- Proprietary controls expose no usable data and the owner will not fund an upgrade.
- No employee or contractor is accountable for alarms, repairs and setpoint changes.
- The portfolio lacks repeatable systems, so every deployment becomes a custom engineering project.
- Cybersecurity policy forbids required access and no approved edge architecture can bridge the gap.
- The business case relies on headline savings from a different climate, tariff or operating baseline.
06From seeing to doing
The company’s recent partnerships show where it wants to move. Neeve adds secure edge access for operational technology. Intel supplies an IoT appliance path. Metrikus contributes data aggregation and visualization. Microsoft Azure underpins cloud deployments. In 2026, Switch announced work with Lawrence Berkeley National Laboratory on the progression from analytics to autonomous operations.
In that account, semantically structured building data supports software that can spot inefficiency and write corrective actions back to controls within guardrails. The six-property deployment suggests a practical middle ground between manual operations and science-fiction autonomy. Software can adjust approved schedules and setpoints; operators can monitor the effect; the system can confirm whether the change held.
The fashionable label is AI-ready. The less fashionable ingredients are open protocols, stable connectivity, clean metadata, secure access and a clear person in charge. Switch Automation’s story is useful because it puts those ingredients in the correct order. Intelligence is not the opening act. It is what becomes possible after the building’s machines can finally introduce themselves.