This past week I started to “train” an instance of IBM’s Watson AI engine on the basics of building operations and maintenance. Watson is capable of learning, and carrying on natural language conversations with people, and promises to transform the management of real estate, but must first learn a lot about what happens in buildings before this value can be realized.
Like people, Watson will learn from the knowledge it’s exposed to. If my instance of Watson begins to learn about buildings from traditional work order documents, it will probably begin to understand the concept of a #fault, perhaps generated by a service request or the building automation system, and will set out to determine the best #fix, before issuing a work order instructing a technician to go an complete the actual #work. This doesn’t seem to be an improvement over current processes. By exposing Watson to a database of work orders, I could simply be training Watson how to copy people’s bad maintenance habits.
So, if using a CMMS database to train Watson doesn’t appear to be a good starting point, what does Watson need?
How about BMS or BAS? Building management or automation systems “know” more about how an individual building operates, but a BMS knows nothing about how people actually use each building, or the financial costs of building operations. If the objective of Watson is to improve the financial performance of building assets, then the BMS, on it’s own, can add little value to Watson’s initial education.
Consider this use case, a (future) natural language conversation about changing HVAC air filters, between Watson and a building services provider. It may go something like this:
Service Provider (Sue): “Watson, I have a technician on a site nearby the Main-Street Mall, tomorrow, with time available. Are there air filters that need to be replaced there?”
Watson: “Hello Sue. I understand that you will have a technician onsite tomorrow at the Main-Street Mall, located at 1240 Main Street. How many hours will the technician have available to complete filter changes?”
Sue: “Watson, the tech will be available for about 4 hours.”
Watson: “Ok, The 8 rooftop units on the west building, and the 4 make up air units above the food court are nearing maximum pressure drop with last replacement, at 121 days and 68 days. Estimated completion is 3.6 hours.”
Sue: “Thank you Watson. Please schedule the filter change, and notify mall security to provide access for my technician, and ensure that the filter pick list has been sent to the filter vendor. “
Watson: “Yes, Sue. Mall security will provide access to the rooftops on the west building, and central food court . Lifelines will be required, along with confined space procedures. The filter pick-list has been submitted to the vendor”.
The concepts covered in this brief exchange include #buildings, #building types, #location, #time, #building systems, #airfilters, #costs, #energy, #services, #logistics, #HVAC, #equipment, #wayfinding, #tasks, #security access, #supply chain, and #vendors.
It’s clear that to fulfill this use case, Watson needs a very deep general (operational and financial) understanding of buildings, so it can anticipate problems before they are reported as alerts or faults, direct facilities staff to take preventive actions, maximize business value for tenants, and operate buildings at maximum cost efficiency. Watson needs to understand that each building is of unique design, construction and use, that building systems differ with each building. Watson will need to develop concepts around systems failures so it can predict the financial costs of maintaining each building over time, and find ways to improve this efficiency. Watson will need the ability to navigate through each physical building, and explain to people where to find spaces and assets, so work can be scheduled efficiently.
Before Watson can adequately get this depth of knowledge, knowledge about buildings needs to come together into an integrated knowledge-base, structured in such a way that people and machines, like Watson, can understand. The BuiltSpace platform is quickly becoming that knowledge-base.
Today, I’m proceeding with Watson’s kindergarten level education. Tomorrow, BuiltSpace and Watson will be helping you drive operational efficiency across your entire real estate portfolio.
Rick: “Watson, where will you find rooftop units?”.
Watson: “Come on Rick! I may be in kindergarten, but I’m pretty sure I will find them on the #roof.”