One of the biggest problems in smart buildings is the difficulty of getting good data from the places that matter, without turning every deployment into a building project. That is what makes PointGrab© new CogniPoint 2 Flex™ so interesting. Announced in February 2026, PointGrab describes it as a battery-powered edge AI occupancy sensor, and the industry’s first enterprise Thread sensing solution, built to bring large-scale, real-time occupancy intelligence into commercial real estate without the cabling, contractors and disruption that usually slow these projects down.
That matters because the modern office no longer behaves like a static asset. Hybrid working, fluctuating desk demand, changing collaboration patterns and stronger environmental, social, and governance (ESG) targets all mean that workplace operators need a much more dynamic picture of how space is being used. PointGrab’s broader platform is built around exactly this idea: a single sensing data system that can support operational efficiency, employee experience and sustainability goals. The company says its sensors can provide people counting, motion detection and area-based occupancy insight, giving building operators data that can drive room booking, digital signage, heating, ventilation, and air conditioning (HVAC) control, cleaning workflows and longer-term space planning.
Traditional smart-building infrastructure tends to assume that sensors will be installed during construction or major refurbishment, when ceilings are open and cabling can be planned in. That works in theory. In practice, it means many organizations are forced to wait for a future fit-out cycle before they can improve visibility into their workspace. PointGrab’s argument with CogniPoint 2 Flex is that workplace intelligence should move at the speed of the business, not at the speed of concrete. Its battery-powered design, magnetic or adhesive mounting, and self-healing Thread network are all aimed at removing the friction that has kept many existing offices from becoming truly data-driven. PointGrab says enterprises can deploy hundreds of sensors in hours rather than weeks, with a three-year battery operations guarantee and no need for wired infrastructure.
This is exactly the sort of application in which edge AI has an advantage over more conventional approaches. An occupancy sensor is only useful if it is accurate, responsive and simple to scale. It also must work within strict privacy expectations. PointGrab says CogniPoint 2 Flex performs all spatial inference on the device itself, with no files stored or transmitted and no personal identifiable information ever collected. That is a far better fit for workplace deployments than pushing raw visual data through the network and sorting it out somewhere else. In a world of rising scrutiny around privacy, compliance and employee trust, on-device inference is not just a technical choice. It is part of the product value.
And this is where PointGrab states that CogniPoint 2 Flex runs its latest-generation AI models directly on the device, enabled by the low-power Ensemble® MCU family from Alif Semiconductor®. To be useful in a modern workplace, the system must support accurate counting, distinguish between areas of interest, operate continuously on a tight energy budget, and remain small enough and affordable enough for wide deployment. Alif’s Ensemble® family is designed precisely for accurate counting, distinguish between areas of interest, operate continuously on a tight energy budget, and remains small enough and affordable enough for wide deployment of edge AI workloads, combining Arm-based processing with integrated Arm Ethos-U55® neural processing support in a low-power architecture.
This is an important distinction. In edge AI, what matters is whether architecture makes inference practical in the end product. Alif’s materials highlight that the real gains come from combining a CPU for embedded control with a dedicated NPU for AI functions, so that inferencing can be handled efficiently without forcing the rest of the system into a high-power operating mode. That system-level balance is what allows a product such as CogniPoint 2 Flex to do more than just gather coarse utilization data. It enables a purpose-built sensor to run useful AI models at the edge while preserving the long battery life and low-friction deployment model that the application demands.
What makes the PointGrab design especially compelling is that the product is not aimed at a narrow proof-of-concept. It is intended as a foundation for a broader workplace IoT layer. PointGrab positions Thread as a secure, energy-efficient and future-ready network for additional workplace devices, and presents CogniPoint 2 Flex as a way to establish an interoperable sensing infrastructure in both legacy buildings and new developments. That is a strong example of what happens when edge AI is paired with practical systems thinking: the result is not just a smarter sensor, but a more flexible path to digital transformation.
For Alif, the significance of CogniPoint 2 Flex shows that edge AI is not only about consumer devices, wearables or headline-grabbing demos. It also has a major role to play in the built environment, where the challenge is often to make intelligence invisible, dependable and easy to deploy at scale. In that sense, PointGrab’s new occupancy sensor takes a real-world system bottleneck, applies on-device intelligence where it matters most, and turns advanced AI capability into something practical enough to disappear into the background.