Romancing the Rules


you've got mail3

Just a little more than 16 years since we found love in ‘You’ve Got Mail’ (Warner Brothers, 1998) mobile telecom is trying to bump us off.  In this sequel, plucky heroine demo’s that were all up to the challenge of  tangling with ML  algorithms.

Can mastering machine-learning rules for email, messaging, social media feeds and other collaboration tools make us better occupants and operators of  intelligent buildings?

I didn’t know that using traditional email was so over. But, the following tweet from Ben Evans, capsulizing his reaction to Google Inbox got me thinking:

“Lots of people are trying to ‘solve’ email… Maybe the solution will be that everyone who uses email will die off.”  Ben Evans, Analyst, Andreessen Horowitz

Ouch! That’s harsh. I’ve remained fairly satisfied with email for collaboration, but I check it primarily from an office desktop. When you are a mobile worker that mostly accesses email via phone or tablet, you have no tolerance for long dumps of unfiltered email and the related risks of missing or losing important notifications. You’ve been waiting for something new. Maybe it is the new app launched this week that lets you ‘Yo’ as Jessie Pinkman?

Here is a summary of the evolution of how people are digitally getting each other’s attention:

  • email = ‘dumb’ file manager
  • email + machine learning for auto-cleansing = Google Inbox
  • chat and messaging are more direct and effective for communicating via mobile: SMS, iMessage, WhatsApp, SnapChat, Facebook Messenger, BBM (formerly known as BlackBerry Messenger)
  • single-tap, zero-character messaging = YO  The YO app can be combined with the online service IFTTT (“If This Then That”) to turn on and off lights, or perform some other simple activate/deactivate function in home automation.
  • Team chat app’s =  next frontier in collaboration. Services like merge all your communications from email, various messaging services and push-notifications from integrated vertical app’s into an integrated, searchable feed

Facilities managers and building operations teams have likely already migrated away from traditional email. Just as an app like Google Inbox is set to eclipse email, so is machine-learning technology transforming the business of managing buildings – and the business of collaborating about the management of buildings.  Again I have to thank Ben Evans for stating what I’m trying to say in a simpler way. At GE’s Mind + Machines event in October he made the point that the tech tools we use to conduct business, eventually change the business we’re conducting.

At GE’s Mind + Machines event in October Ben Evans made the simple point that the tech tools we use to conduct business, eventually change the business we’re conducting.


My current angle of interest in learning/collaboration platforms comes from my work creating content for commercial building energy efficiency solutions and services firms. With the current perpetual commissioning model, their charter extends to transforming the client building operations staff into a self-learning organization. This work has me asking  “How can content for notifications, learning and collaboration about buildings be organized and shared in a way that provides the best experience for all the target users — commercial building operations and facilities management staff, as well as occupants?”

It’s worth noting that Google Inbox is not receiving stellar reviews by all its early users. Intelligent algorithms can be pretty dumb when they first start trying to predict human preferences and behaviors. But, they get better with adjustment. This raises more challenges and questions, such as “How can we get out ahead of the algorithms and better train them to do our bidding?” Then, considering the tech tool feedback loop phenomenon, “Can the learning we engage in to define better machine-learning rules for our mail, messaging, social media feeds and other collaboration tools help us improve our skill at operating and occupying intelligent buildings?”

A paper just released by Professor Julia Day, formerly of Washington State University and now with Kansas State University, finds that “More than one-third of new commercial building space includes energy-saving features, but without training or an operator’s manual many occupants are in the dark about how to use them.” Based on another university study, Carnegie Mellon University (CMU) Professor of Architecture Vivian Loftness recently said “automation has led to more complexity, leaving the occupants of all these offices disempowered and uncomfortable.”  Both sources agree that over-dependence on automation is a problem and that getting more people-engagement in machine-learning is the way forward. As with email, there is this preliminary sense of “All this automation is making us work too hard,” along with the promise that if people work through the machine-learning, we can get this right.

With Comfy, Building Robotics is among the first to give occupants a simple, convenient app that ‘learns‘ their thermal preferences, connects to heating and cooling systems, and dynamically adjusts setpoints. In effect, it empowers the people in a space to help tune the rules that define HVAC automation control sequences.  Likewise, the CMU team is using their research as a launch point for a mobile app aimed at helping occupants drive machine learning. This app builds upon OSIsoft’s PI database system to capture Building Automation System (BAS) data and compile the HVAC, lighting, and other streams into a single source. In this case, Microsoft’s Azure ML is used for simple, visual programming of the rules to run against the gathered data. Early users on the CMU campus are able to adjust lighting and temperature from their smartphones.

John Petze of Sky Foundry explains the process of working through the machine-learning for building operations: “Data analytics is taking knowledge of your best operators and translating it into machine-learning rules. Rules are software that identifies patterns in data—equipment faults, deviations from expected performance, actual results versus goals or benchmarks. The result is like an ever-present expert watching all systems 24X7 for sensors that have failed, systems that have been overridden—all the things that degrade performance.”

When it comes to the ‘train-the-trainers’ task of engaging building operators and occupants in inventing and refining machine-learning rules, Matt Schwartz, commissioning expert with Altura Associates had this to say : “By introducing data analytics, Altura has been able to shift our clients to proactive building management.  They don’t have to learn programming, but the entire team does need to reprogram their thinking – the facility engineers, property managers, asset managers and ownership. It does take changing human processes to make this shift. Altura places great emphasis on the people aspect because it’s the critical factor for long-term success. So we engage building operators throughout the process. We give them ongoing training in the analytics. We create O&M-specific training documentation that references the analytics. We’ve rewritten the procedures for responding to hot and cold maintenance calls to include analytics steps. This is technology to make people better, not to replace them.”

So what will a whole-building collaboration hub for such ongoing O&M content soon look like?  Team chat platforms like the Slack app that integrate popular general-business services (Google  Drive, Dropbox, etc.) with messaging and notifications are a start.  Here’s another prototype: ‘Building in One’ from France – coming to North America soon.  Note this hub targets the information needs of occupants as well as operators and owners. It was envisioned by by a French Real Estate industry consortium with development by a Google Cloud Apps developer partner. building-in-one

Other integrations might be separate apps for HVAC, lighting, access security, etc. Even dynamic glass windows are controlled via intelligent algorithms designed to ‘learn’  user preferences, so there will be an app for training those machine-learning rules too. Then the hub will need to gather the notification streams of building operational analytics software like Sky Foundry’s SkySpark® and OSIsoft’s PI.  Likewise, maintenance-call and job-tracking services within enterprise workplace management suites will need to be integrated. Slack boasts that all content integrated into the platform becomes “instantly searchable, and available wherever you go.”  That’s what you want, even extending to archival content. Self-learning would be enhanced if there were apps for sites like and where seekers could find the wealth of information they’ve published on building operations and facilities management over the years.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s