Microsoft – Increasing productivity through Smart Buildings
The use of Machine Learning and Internet of Things to increase employee productivity
Problem Statement
In May 2018, Satya Nadella (Microsoft CEO) made a slew of announcements about its IoT business segment at its annual Build Developer Conference [1]. Three years ago, an Innovation team at Microsoft started researching on how cutting-edge technologies can be used to improve employee productivity. Organizations are grappling with the challenge of making employees engage in productive and impactful work and optimizing their physical workspace costs. How can an employee find relevant resources within a company operating in 121 countries [2]? Where does one find a space for an immediate meeting? Can the company optimize the use of its HVAC (heating, ventilation and air conditioning) systems? The team created a solution – Smart Building, an innovative IoT (Internet of Things) and machine learning based platform to address these issues.
Product Development
The solution aimed at creating a one-stop chat bot for all information an employee might need, that utilizes machine learning to improve its responses over time, helping employees save time and increase productivity. As a start, knowledge bases were added into the system to provide answers to frequently asked questions. To help answer questions further, more data sources were added (websites, articles, internal publications), which required the use of machine learning to filter all sources and provide relevant responses. In the short term, users’ feedback on the bot’s responses was sent back into the system to refine the weights assigned to various sources to make sure future hit rate (% of correct answers) increases.
To address the office space utilization problem, motion and temperature sensors were installed in various locations and data collected from those was collated together and heatmaps were generated. These heatmaps (motion and temperature), when overlaid over one another would provide insights into usage patterns and areas where HVAC systems could be shut down, leading to reduced energy costs for organizations.
Based on research conducted internally, one of the biggest challenges identified was that meeting spaces were booked but not utilized ~50% of the time [3]. To bridge this gap of supply and demand of meeting spaces, the platform provided real-time information on room utilization across buildings. The unique selling point was providing an estimated probability of the room remaining unoccupied for the period searched, based on historical patterns. This provided a scalable platform that could be modified to predict other sensory information as well, for example, finding a quiet spot for individual focus time.
In the long run, as more data is collected and algorithms are refined further, the system could be used to drive capacity and resource planning decisions with very limited or no human involvement. As the bot responsiveness is improved and larger datasets are added, the need for helpdesks would be eliminated, leading to reduction is workforce and thus, savings for the company. Also, questions where low quality matches/no matches were found can be reported back to authors to improve content.
The Way Forward
Moving forward, the company can make greater strides to improve the applicability of the product. As the technology industry moves to more collaborative workspaces, usage patterns can be viewed to evaluate how constructively these spaces are being utilized and how to best optimize them. Based on confidence of predictions, employees can be proactively prompted to free up spaces. Artificial Intelligence can be used to determine the number of occupants, leading to a more calculated control over HVAC systems [4]. In the long term, the chat bot can be developed to automate customized training and curriculum using request data provided by employees. This will specifically be helpful to new employees for ramp up and reduce HR effort required for the same.
The internet of things (IoT) and artificial intelligence (AI) have great potential to help building owners, operators, and occupants manage and dwell in buildings with greater efficiency—saving costs and energy, and organizing space in a way that best fits a company’s culture and goals. These benefits apply to a variety of spaces. For instance, hospitals, stadiums, and factories can all be smarter, and the same applies to the electrical grids that connect them and the cities that contain them [5]. The company is already partnering across industries and in the public sector, utilizing the concept to reduce operating costs in the city of Jaipur, India by implementing a smart lighting program [6].
Looking at the initial success of the solution to improve productivity using machine learning, there still remain questions on its feasibility. Is the solution scalable across countries? How do you deal with language barriers in chat bot applications for global companies?
(758 words)
[1] Nicholas Shields and Rayna Hollander, “Microsoft’s IoT announcements at Build — Nokia adds AI, ML, analytics to IoT business — Google Assistant now supports 5,000 smart home devices”, Business Insider, May 8, 2018, https://intelligence.businessinsider.com/post/nokia-looks-to-add-ai-ml-analytics-to-iot-business-2018-5, accessed November 2018.
[2] Microsoft Corporation, “Facts About Microsoft”, https://news.microsoft.com/facts-about-microsoft/#OperationCenters, accessed November, 2018
[3] Interview with Microsoft employee on November 9, 2018
[4] CBInsights, “In Smart Buildings, Energy Efficiency And Comfort Go Hand-In-Hand”, November 7, 2018, https://app-cbinsights-com.prd2.ezproxy-prod.hbs.edu/research/smart-buildings-energy-efficient-tech-comfort-expert-intelligence/, accessed November 2018.
[5] Bert Van Hoof, “Smart Buildings, built on Azure IoT”, Microsoft Blog, June 5, 2018, https://blogs.microsoft.com/iot/2018/06/05/smart-buildings-are-built-on-azure-iot/, accessed November 2018.
[6] Peter Newman, “Google may launch affordable screened smart speaker | Microsoft ramping up IoT presence in India | Google Home Mini overtakes Amazon Echo Dot in Q2 shipments”, Business Insider, September 20, 2018, https://intelligence.businessinsider.com/post/what-googles-after-with-screened-smart-speaker-microsoft–ramping-up-iot-presence-in-india-google-home-mini-overtakes-amazon-echo-dot-in-q2-shipments, accessed November 2018.
[7] Tom Krazit, “New Microsoft Azure tech could let smart building operators model the links between people and objects at work”, June 5, 2018, https://www.geekwire.com/2018/new-microsoft-azure-tech-let-smart-building-operators-model-links-people-objects-work/, accessed November 2018.
[8] Sascha Corti, “Welcome to the Microsoft Switzerland Co-Working Space and Smart Building”, August 30, 2017, https://blogs.msdn.microsoft.com/microsoft_developer_switzerland_news/2017/08/30/welcome-to-the-microsoft-switzerland-co-working-space-and-smart-building/, accessed November 2018.
Featured image: Frank Konkel, “The announcement could improve the company’s chances for the Defense Department’s JEDI competition.”, May 16, 2018, https://www.nextgov.com/it-modernization/2018/05/microsoft-announces-ic-deal-dramatically-improves-its-jedi-chances/148256/, accessed November 2018.
You raise a very valid point on the standardization of the IoT devices across the globe. The data is currently available in different format across devices and this may create issues in collating the data seamlessly. I belive that we need a common language that could be applied to the IoT world (such as Windows OS) to talk to each other irrespective of the region and context they operate in. This will help it to be consistent across borders.