iPhone users are now used to “Hey Siri” to dial the phone, read a text, answer a burning question, read you a weather report or feed verbal you driving directions. Google’s Alexa is targeting in-home applications, allowing you to manage your home environment, appliances and, can even order your groceries for you. Can Watson help us deliver service excellence to commercial real estate, with natural language conversations between occupants, tenants, facilities teams and service providers? I set out to train Watson.
IBM’s Watson uses “experience” and data to discover patterns, and act on those patterns found in future data. Some people call it artificial intelligence, or cognitive recognition, but without a database of “memory”, Watson would be as dumb as the bricks & mortar building.
I plan to point Watson at the BuiltSpace’s 14,000+ digital buildings, so Watson can better understand how people build, use, operate and maintain these buildings, and Watson, through the BuiltSpace platform, can help building owners, and their tenants, to better utilize their building assets.
BuiltSpace offers Watson a very unique digital image of each physical building, presenting Watson with a rich data set which can include building address, structure type, size, usage, number of levels, spaces, assets and equipment within the spaces, service history of the equipment and spaces, inspection results for safety and compliance inspections, tenancies within spaces, suppliers to the buildings, energy consumption, operating cost history, and finally tasks or service requests identifying needed changes to each building. This information is structured in a manner that will allow an intelligent person or our Watson powered droid named “Bildi” to navigate through the building to locate people or things in the building.
So, my first order of business was to get Bildi, my new droid, to converse in chat. I created a workspace in IBM’s Bluemix environment and launched the Watson conversation service. I started to train Bildi, by getting Watson to respond to “Hey Bildi”. It was pretty easy. Watson defines potential actions (IBM calls them intents) that you may want to perform. My first intent is #HeyBildi, where I tell Watson that I want to start a conversation with my droid Bildi.
Next I created an intent #TooHot, where a building occupant may feel that the temperature is too warm, and they want Bildi to correct this, followed by an #Emergency intent that defined five or more ways that people may describe an emergency. The #Emergency intent could include “Call 911”, “I smell smoke”, “Fire”, “I don’t feel safe”, “There is an intruder” etc. Bildi now understands that these, or similar sentences could mean emergency. Then, I begin training Watson on how it should deal with an intent, by creating a sample dialog.
I then started to train Watson about things (entities) that may be found in buildings, introducing Watson to synonyms for each thing. Buildings are pretty complex, but we can classify entities and use our database to train Bildi about buildings. We can even submit photographs of equipment in buildings, and use Watson’s photo recognition to service to train it on what the equipment looks like.
It’s a work in progress, but it’s been a lot of fun so far. My plan is to train Bildi to understand individual buildings, and to converse in native language with occupants and visitors, then implement Bildi to provide conversational tenant service requests, replacing web-based service forms with “Hey Bildi, I’m too hot”. “Sorry to hear that Rick.”
If your real estate company would like to participate in this project, please connect with our team, via the contact form, or contact me directly at firstname.lastname@example.org.