The Smart Home Assistant
There’s been a lot of talk recently about advancements in the smart home arena, especially about new ways to control smart home devices. I’ve frequently heard Amazon’s Echo referred to as a smart home device and, just this week, web service IFTTT announced new partnerships which are intended to allow smart home devices to connect in an automated fashion to other devices and services. However, what we’re still missing when it comes to the smart home is a true smart home assistant – a counterpart, if you will, to the smart assistants that come baked into every modern smartphone operating system. This post dives into what that means in practice and who might be best positioned to deliver on this vision.
Before we go any further, I need to credit Matt Eyring, Chief Strategy and Innovation Officer at smart home company Vivint, for coming up with the term “smart home assistant”. The thoughts below are a combination of his initial idea and some of the additional thinking that I’ve done. (If you’re not familiar with Vivint, rather than selling devices on a standalone retail basis, they sell subscriptions to a smart home service which typically includes security and a variety of other features, built around a combination of their own and third party devices.)
The smartphone analogy
The analogy here is to the smartphone assistants we’ve all become familiar with over the last few years, whether you have an Android device that comes with the Google assistant, an iPhone and Siri, or a Windows device with Cortana. All of our smartphones now come with smart assistants which can respond to questions and proactively provide information we need. The proactive elements of these assistants are heavily dependent on their ability to gather and synthesize information from the device and its owner over time, learning patterns and behaviors, and combining the various things they know about us to provide just the right bit of information at the right time. My favorite recent example is one provided by my iPhone, which now tells me how long it will tell me to get to my weekly soccer games, even though I’ve never explicitly told it where I play soccer. It’s simply combined the calendar item “Soccer” with the place I usually go at those times to draw a smart conclusion about where I play soccer.
Our devices are able to do these things because they have access to several sources of information:
- Information we’ve explicitly fed them – who we are, who our spouse is, etc.
- Information from other data sets on the phone – our calendars, contacts, email, and so on
- Location data – information about where we go and when
- Sensors on the phone – accelerometer, barometer, compass, and others
What makes these smart assistants smart is they combine all this information to create a view of who we are, what we do, and therefore, what we’re likely to do and want to know in future. If they do this well, they allow us to rely on them more and free up our time and attention for other things.
The smart home assistant
Let’s now apply this concept to the smart home, how it works today, and how it might work in future. One of the best words to describe the smart home market and indeed, the typical smart home today, is “fragmented”. There are dozens of device makers and solution providers competing for attention and dollars and many of them have their own ecosystems that don’t talk to each other. And yet, few of these ecosystems truly combine devices from all the key categories in an integrated fashion. Enter Amazon’s Echo, IFTTT, and a variety of other solutions which are attempting to bridge some of these gaps.
However, these solutions are inherently dumb. I’m not using that term in a derogatory sense, but in the sense these solutions literally know nothing about the devices they’re controlling or the context in which they are operating. They simply take commands (or automated actions) and execute them. There may be some intelligence in the individual smart home devices behind those front ends but there’s nothing pulling together all that data and building a real-time picture of conditions in the home off the back of it.
A truly smart home assistant would do for the smart home what smart assistants have done for our phones and pull all this information together and make use of it to make our lives better. We’re still waiting for that in the smart home domain. What does this mean in practice? Well, there would be several necessary elements:
- Sensors – just as on our smartphones, a truly smart home would need an array of sensors to measure both atmospheric conditions and human presence across the different spaces in the home
- Data gathering – the smart home assistant needs to gather and keep data from these sensors and from user actions on an ongoing basis
- Machine learning and artificial intelligence – with all this data in hand, the assistant would need to leverage artificial intelligence and machine learning to draw conclusions from the data, learning patterns of behavior both on an individual basis and across a base of users
- Action – the smart home assistant would then learn to take actions on the basis of the data and the analysis, in ways that make users’ lives easier
The Nest thermostat currently does some of this on a narrow basis as it relates to temperature control in the home, based on individual thermostats and their locations. But there’s no smart home ecosystem today that does this on a broader scale across multiple devices in different device categories and throughout the home.
There’s also another aspect to a smart home assistant worth mentioning. Today’s voice and touch interfaces tend to be built around instantaneous actions – that is, I tell Alexa or Siri I’m going to bed or leaving the house and, within a second or two, they turn off the lights and lock the doors. But in a real-world scenario, I might not want those actions to happen immediately – I might need time to go upstairs to my bedroom, to get the car out of the garage, or to walk out the front door. The first time I do this, the assistant might list possible actions to take and ask for confirmation. Once I confirm, it might then wait a while longer to allow me to either get to bed or leave the house before beginning the series of actions. It might start a bedtime routine by locking the doors and windows and turning off outside lights, then move on to turning off inside lights and arming the security system once it’s detected I’m safely in bed. Again, this is only really possible with sensors throughout the house. Once I’ve established my regular patterns in this way, the smart home assistant may simply begin taking the actions in a certain order with appropriate pauses in between.
Who’s best positioned to build a smart home assistant?
All of this raises the question of who is best positioned to build such a smart home assistant? We’ve already mentioned Amazon and its Echo device, which many people seem to think has the potential to become a serious player here. But as we’ve also discussed, the Echo is inherently blind when it comes to the devices in the home – it has no intelligence as such to direct a complex set of interrelated actions. But it could be a front end for a smart home assistant that lives elsewhere. Who, then, might build the smart assistant itself? Here are some possibilities:
- Amazon – I’ve already said the Echo doesn’t fill the role of smart home assistant because it’s not intelligent enough. There’s nothing to stop Amazon trying to build such an assistant as well (or eventually as a component of Echo/Alexa) but it lacks both the ecosystem and ubiquitous front ends to really make this work, so I’m skeptical
- Apple – Apple has no smart home devices of its own, but it does have an ecosystem working off the back of its HomeKit technology, as well as a front end in the form of both Siri and (from iOS 10 onwards) the Home app in iOS. It already has scenes and other necessary building blocks to create some of what I’ve talked about here but the biggest challenge may be Apple’s antipathy to building data profiles on its users. A smart home assistant would require capturing and storing significant amounts of data to be really useful. Apple doesn’t seem to want to do this kind of thing, so it would limit its ability to build a really smart assistant for the home. It would also need to explicitly support a range of sensors as part of HomeKit – today, HomeKit is mostly about controlling devices around the home rather than capturing data from them
- Google – Google is arguably in a pretty strong position, with both its ownership of Nest and the various Google assistant technologies and endpoints. With the Home device coming in the next few months as well, it will potentially check yet another box on the to-do list. But the big challenge facing Google is it’s had to keep its Nest and Google operations fairly separate in order to assuage data privacy concerns. Nest has also failed to build a hub for itself, preferring instead to allow for a series of automated integrations with individual Nest devices, so there’s no obvious central device or brain for a smart home assistant to work with. Google definitely has the AI chops to do the algorithmic work here well, but it would likely run into difficulties trying to capture enough data and make it usable
- Samsung – Samsung has of course made investments in the smart home with its SmartThings acquisition and with a massive installed base of smartphone users, it has a variety of endpoints with which to control smart home devices, too. But its smart home strategy so far has been pretty disjointed, with little to show for the integration between Samsung smartphones and tablets on the one hand and SmartThings devices on the other. Given Samsung’s weakness in software and services relative to major players like Google and Apple, it seems something of a long shot
- Smart home solution providers – I’ll return to Vivint here briefly, as one of a number of smart home solution providers also including Comcast, AT&T’s Digital Life, Alarm.com and its partners, and others, which could potentially build a smart home assistant. So far, none of them have built this capability, though I know Vivint is working on it. This set of players is likely in a strong position if they decide to make an investment because they’re already offering joined-up services. They would still need to roll out a variety of sensors throughout the home and would need to invest heavily in the kind of data analysis tools necessary to make use of the information coming out of those sensors. Google and others are much stronger in AI but these companies may be more able to capture and make use of data from existing broad smart home deployments
At this point, it’s fairly clear the smart home needs to evolve significantly if it’s to broaden its appeal, as I’ve written previously. But it’s less clear which of these various players will do what’s necessary to make the big leap forward and bring about meaningful improvements in how all this stuff works together.