Smart Assistants Making Progress…But Slowly

When Adele sings “Hello,” people are clearly listening—to the record-setting tune of over 3.3 million album purchases in a single week. When you greet your smart device with a verbal introduction, however, well, let’s just say the results aren’t quite as clear.

Though they were hailed as “The Next Big Thing” when they were first introduced, Smart Assistants, or Personal Assistants, such as Siri, Google Now, and Cortana, haven’t exactly torn up the charts. Yes, there was a big initial splash—especially after the initial release of Siri on the iPhone 4S in October 2011—but there has been an equivalent, if not even larger, backlash since then. In fact, Siri quickly devolved from something people marveled at to something people joked about.

To be fair, Apple has made significant improvements to Siri since then, and the introduction of both Google Now and Microsoft’s Cortana has raised the bar for the Smart Assistants category as a whole.

In order to get a better sense of where things stand now, I included a few questions on the recent TECHnalysis Research Consumer Device Usage study of 3,012 people across five countries (US, UK, Germany, Brazil and China) to find out who may be using these verbal assistance-based capabilities, what they’re using them for, and what they think about them overall.

The results show that while Smart Assistants are making progress, they still have a long way to go before they really become mainstream tools, particularly with older consumers. The chart below shows which age groups are using which Smart Assistants, and which use none at all.

WW Smart Assistant Usage

©2015, TECHnalysis Research LLC

While there are a number of very interesting points that can be gleaned from this chart, the first thing that stands out is that nearly half of all respondents across all age groups (Total) said they don’t use Smart Assistants at all, and another 9% don’t know what any of the Smart/Personal Assistants actually are. Amongst the 1,024 US respondents, the result was just over 50% saying they don’t use Smart Assistants and 7% who don’t know what they are. (Note that the totals add up to more than 100% because respondents could select all the Smart Assistants they use, with the average number they use being 1.2.) So, for both worldwide (WW) and the US, only 43% of device-owning consumers said they use Smart Assistants. Clearly, this suggests that more work needs to be done to make these voice-based capabilities more compelling (and, likely, more accurate) before a much wider audience will actually use them.

Breaking the results down by age group, only in the 18-24 and 25-34 age groups are there more people who use Smart Assistants than don’t (in the US results, only 25-34s). Also, there is actually a smaller percentage of 18-24-year old Millennials using Smart Assistants than 25-34-year old Gen X’ers (47% vs. 62% in the US and 52% vs. 57% worldwide). These somewhat surprising numbers suggest that younger device users are not necessarily the most proactive when it comes to using the latest features, or that voice-based control isn’t as interesting, or necessary, for younger users as it is for slightly older users (or, most likely, some combination of these two possibilities). Regardless, it reflects a subtle, but potentially important shift about the expectations that future generations of device owners may have.

Interestingly, the activities for which Smart Assistants were used (by the 43% who used them) was actually very similar across age groups, both worldwide and in the US. The chart below shows the top-level results for the full five-country sample of 1,286 Smart Assistant users.

WW Smart Assistant Activities

©2015, TECHnalysis Research LLC

The top activity by far was requests for information searches, followed by asking for directions. (One difference in the US results is that over 59% of US respondents asked for driving directions, versus just under 53% worldwide). Asking to do certain activities on the device, such as play music, launch apps, adjust settings, etc., was done by fewer than half of Smart Assistant users. Truly “smart” activities, such as using a Smart Assistant to make suggestions of other things, music, restaurants, etc., were only done by 1 in 5 Smart Assistant users.[pullquote]It is a bit disappointing to see how few people are using Smart Assistants. It’s also frustrating to see how little they’re being used for truly smart activities.”[/pullquote]

Of course, the other big data point from the first chart was the dominance of Google Now usage over other Smart Assistants, particularly among younger users. It turns out Google Now also edged out both Cortana and Siri on an overall satisfaction rating as well, but just barely, with an average of 3.8 (on a scale of 5), versus 3.7 for Cortana, and 3.6 for Siri. In the US, Google Now and Cortana were statistically tied at 3.9, while Siri had a 3.5.

The notion of a voice-based interface, and the concept of talking at your devices was more science fiction than science fact for a very long time, so the fact that we’re making any progress in this challenging area is unquestionably a good thing. Still, expectations continue to be very (and probably unrealistically) high for these technologies, so it is a bit disappointing to see how few people are using Smart Assistants. It’s also frustrating to see how little they’re being used for truly smart activities—the most common usages are arguably little more than simple typing replacement.

There’s no question that truly Smart Assistants will be a critical part of future devices and services we use, but it’s also increasingly clear that we still have a long way to go before we reach that digital nirvana.

Published by

Bob O'Donnell

Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a technology consulting and market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on Twitter @bobodtech.

670 thoughts on “Smart Assistants Making Progress…But Slowly”

  1. The problem with smart assistants is that they aren’t smart enough. They’re still lacking in listening comprehension (getting the words I said right and understanding what I am asking for) and in general and contextual knowledge. Going through one or two frustrating episodes of repeating then rephrasing a question was enough to make me stop using it except in the rarest circumstances, usually when I’m driving.

    1. Yes, agreed that these are significantly harder to do than they first appear and hence the frustration that many people have with them. Contextual awareness is also a key aspect to making these things really work the way we want them too and that’s a very difficult challenge as well.

    2. I think we’ve got to set expectations and restrict the context. We’re probably a long way away from a computer being able to pass as my best friend; but having it competently handle requests regarding meetings and contacts, searches, OS and app commands seems quite attainable technically.
      I bought Dragon Dictate to do that on my PC 20 yrs ago though, and never really used it past the initial “let’s be amazed by something” rush. And it worked quite well.

  2. Thank you very much for this article. There has been huge hype surrounding these smart assistants, and lots of discussion around whether or not Apple’s stance to privacy will hinder development of these. However, despite looking hard on the Internet, I have never found any report on a) what percentage of people actually use them, b) what they are used for, and c) are customers satisfied with them.

    To my knowledge, all previous discussions have been based on anecdotes, and while entertaining, they are unlikely to be anywhere near the truth. I applaud you for being one of the first people to publish real information.

    I would however appreciate more granularity. The countries that you included are hugely diverse, with China for example not being receptive to Google, and Brazil being a relatively low income and hugely Android market. I am sure there are some interesting differences among the countries.

    I am also very curious as to what “look up information online” encompasses. I used Siri a lot to look up weather (which I haven’t done since I got the Apple Watch, because tapping a complication is more convenient), but I never appreciate it when Siri offers to look up something on Bing. I would never use Siri to look up nearby restaurants, because I always want to browse through the results on a Yelp-like app. I only use Siri to look up very simple information on topics that I ask frequently. It’s also difficult to drill down on queries with a voice interface, because you don’t get a quick sense of the drill down options. Therefore, my hypothesis is that “look up information online” is very different from what people do on PCs and even on their smartphones via touch keyboards.

    It’s also very interesting (although expected) that people do not think too much of the suggestions that these smart assistants make.

    As a general trend, my hypothesis is that these smart assistants will not work as something on which you could put ads on. Information searches are very simple, generic and specific. Recommendations are mostly a nuisance. Furthermore, given the very simple questions that I ask, I strongly doubt that the information that I might be giving away would be valuable for ad targeting (they’ll only know that I use a kitchen timer and check the weather in the morning). More data is needed to verify/reject this hypothesis.

    I’m sure that your clients are going to get the details first, but if possible, I would appreciate more information at a later date, and maybe follow up.

    1. Unrelated, but… What I find funny is the lengths to which “they” go to track us, when I’d be perfectly willing to tell them what I’m looking to buy: gifts for friends and family, a yuuuuuge tablet (I’ll call it Donald and give it a toupé), yuuuuge phones too. Yet every month I have to spend several hours looking for that kind of stuff, to then be inundated by ads after the fact.
      Google Now seems to be getting better a pinning me down, but it doesn’t see my email nor half of my searches which is quite a hobble. There used to be a Google service to save recurring searches, it seems to have disappeared. And there’s nowhere to register to get gift suggestions for a boy born in 2012, purchasable in Canada (a surprisingly limiting constraint), for xmas and his bday.

      1. They way that I think of it is this;

        Bad AI is hungry for data. It needs all the data it can get it’s hands on, and still more, because it cannot intelligently expand on what it already knows. It only “knows” things very superficially, and that is why it cannot expand. This is the AI as we know it.

        On the other hand, good AI, or any good intelligence for that matter, is not. Intelligent people hear one thing, and infer ten. Even babies.

        Google’s photo AI views thousands of photos of Gorillas and billions of photos of Humans. It still made a mistake. A child may see 10-100 photos of Gorillas and would always be accurate thereafter (of course, that child would have to have seen a black person before, and must know that you can’t make the decision on colour alone).

        Current AI is super fast, but still super stupid. It’s trying to make up for its stupidity by speed and volume, and that’s why it needs your data. It sometimes works.

        1. An ad hoc inexhaustive list of questions a computer should be abled to positively address so as to accurately identify photos of gorillas, after having been previously “shown” a few pictures of gorillas. And to get better at it each time:

          1. What are the salient features and combination of features that are unique to a gorilla?

          2. The above is relatively easy, this one isn’t: What are the salient features that might make me think that what I am looking at is a gorilla when in fact it isn’t?

          3. Based on the pictures of gorillas I’ve seen, what would a gorilla look like if I viewed it from an angle different from anything I’ve seen in the pictures? What would it look like if it changed its bodily position?

          4. How can I verify whether or not I correctly identified a gorilla without relying on a human or some other external source telling me?

          1. Agreed, but I don’t think that current machine learning algorithms incorporate item 3. That is, they don’t actually construct a mental 3D model of what they are seeing is actually like in the real world.

            Although I don’t know the details, I’m pretty sure that items 1,2,4 are incorporated in the machine learning processes.

            Of course, the real issue is that we aren’t sufficiently sure how we humans recognise/distinguish things. Animal brains are amazing and truly magical machines that we don’t know enough about, let alone emulate with a digital computer.

            As I understand it, the hope is that the pure speed and processing power of digital computers, combined with massive training data sets will overcome the crudeness and ignorance of current algorithms. This works well in Chess for example, but then living organisms did not evolve through billions of years to play Chess.

            Artificial Intelligence can beat humans when the humans are doing very logical thinking (Chess, arithmetic, etc) and essentially bending themselves backward to do what computers do best. Artificial Intelligence cannot yet beat humans when computers are trying to do what humans do best.

          2. I’m curious how a computer would be able to do #4. I’ve always thought of a computer’s mind, if such a thing can even exist, as a (or the) perfectly solipsistic entity.

          3. I miss the part where you mentioned you can’t rely on a human.

            Can humans identify Gorillas if they aren’t taught? Surely babies can understand that gorillas aren’t the same as their parents, but they would probably apply the same to other human beings of a different race or skin colour, unless they had contact with such people.

            I think that #4 may not be applicable to humans either.

        2. I think I’d roll that back one step: they’re trying to have AIs deduct stuff that we’d be willing to simply declare. When you’ve got a hammer, everything looks like a nail. Let’s AI everything !
          Why not position ads as a shopping assistant, letting me enter recurring needs (the kids ! ^^), low-priority stuff (a 12-14″ dual-boot tablet with OK specs, took me a year to find that)… they’ve made advertising utterly one-sided: sellers decide what they show me (and they mostly seem to have doubts about my equipment both ownstairs and upstairs, and stuff I already bought or decided not to). Show me a shopping assistant, not ads !

  3. Lincoln Casino offers an array of Match Deposit Bonuses which actually begins ranging from the Welcome Package the casino offers. There are five deposits that you can make with the casino right in the beginning which lets you claim one such bonus. Adding to that, over time the casino keeps introducing such fresh bonuses to ensure that you have fun besides the Welcome Bonus. After a detailed review by our independent casino review team, Lincoln Casino has received a Good reputation rating. This means the casino is a good option but there are even better casino sites on the internet. Therefore, taking advantage of offers from this casino should generally not result in any problems, as long as players from your country are actually eligible for them. Trabzon’da ‘şampiyonluk’ biletleri karaborsaya düştü: En ucuzu 30 bin TL!
    https://arthurpckq666544.tkzblog.com/20710260/szybkie-kasyna
    Regarding bonuses and rewards, Cafe Casino likes to roll out the red carpet for its new and existing players alike. This new online casino offers a fantastic 250 percent up to $1,500 welcome bonus on its fiat currency side and an even better 350 percent up to $2,500 for crypto deposits.  A no-deposit bonus is one of the most sought-after promotions available from online casinos. Real money deposits are not required before extracting this type of offer. No deposit bonuses are available to new customers and often require a bonus code. Since you don’t need to deposit money into your account beforehand, no deposit bonuses are usually available for relatively small sums of money. Yes, it is possible to win real money at online casinos. Online casinos offer a wide range of games, such as slots, table games, and live dealer games, that you can play for real money. By making a deposit and placing bets, you can win real money prizes, including jackpots and other rewards.