Tidbits about Bots

The software world is all atwitter about the latest technology to head their way: bots. Touted by some as the next evolution of apps and others as the critical first step in enabling realistic human-to-machine interactions, bots have become the hot new topic du jour. Featured recently by Microsoft at their Build conference, predicted to be further discussed by Facebook at their F8 conference, and already in reasonably wide use by messaging platforms such as Twitter, WeChat, Slack and others, bots are seen as an exciting outgrowth of recent advancements in AI (artificial intelligence).

Conceptually, the idea of an automated electronic assistant that excels at performing specific tasks is certainly an appealing one. Bots that could take the place of those widely despised automated customer service systems we all have had to interact with on the telephone, for example, might be a godsend in comparison. However, most efforts toward more intelligent automated telephony support have been a dismal failure resulting in desperate cries for a human operator.

Some web-based support systems are a bit better but most of the ones that are actually useful are staffed by real people. Now, that’s great for you and me as customers, but expensive and often difficult to scale for companies who use them. So, there’s clearly a business incentive to drive the creation of automated bots that could perform across a variety of different communications mediums and platforms.

But I have to admit I get concerned about how far this could go. For one thing, the idea of certain kinds of bots proactively reaching out to me when I don’t initiate the conversation could get annoying (and time consuming) very quickly. The notion of bot spam is fairly disconcerting and yet, utterly predictable.[pullquote]The notion of bot spam is fairly disconcerting and yet, utterly predictable.”[/pullquote]

More important though, is the question of how many bots people can and would be able to interact with. Right now, much of the discussion around bots seems to suggest they focus on doing one specific task, such as creating a flight reservation or booking a good restaurant with an available opening.

To a point, that’s fine and good. Very quickly, however, it’s not difficult to imagine getting overwhelmed by bot requests and interactions, not all of which will probably even work the same way. Plus, unlike mobile apps, which you consciously have to choose to download and use, interactions with bots could very well be foisted upon us. The mere act of visiting a web site could immediately launch an interaction with a bot.

In some instances, the additional level of support and help which that bot interaction might enable could prove to be very useful but if we already know what we want to do, it could serve as interference and actually slow us down. Obviously, bot designers will need to take these kinds of permutations into consideration as they develop their bots. But as the technology first ramps up, and before some of these lessons are learned, I have a feeling there could be a lot of frustrating bot interactions.

Part of the problem is there will likely be different kinds of bots on different platforms. In theory, of course, any bot should be able to handle normal human interactions, rendering platform differences moot. But the realistic application of technology never really works this way. As a result, either the subtle differences in how bot-type services get deployed across different platforms or the kinds of lock-in strategies various platforms will likely leverage to drive higher usage for themselves are bound to get in the way of our early bot encounters.

As platforms standardize and leaders rise to the surface, these issues could fade away, particularly if companies can create a solution that makes using tens or even hundreds of individual bots feel like natural extensions to a single experience. In the interim however, I expect we’ll see some vigorous new platform battles that will keep our bot interactions from achieving the kind of awe-inspiring wizardry of which they are potentially capable.

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.

717 thoughts on “Tidbits about Bots”

  1. Real exchange from a T-Mobile Store I visited last week.

    I was at the mall killing time, browsing around. I walked into a T-Mobile store to look at accessories for my phone. T-Mobile is not my carrier.

    Salesman: Hi can I help you?
    Me: (Obviously browsing around) No thanks, just killing time. If I have questions, I’ll ask.
    Salesman: Who is your current carrier? (beginning to cross the line…)
    Me: The one I have in my pocket. (also beginning to cross the line…)
    Salesman: Never heard of that one before…
    Me: Yeah, it’s the one I switched to “from” T-Mobile.

    I hate being “sold to” with a passion. What you describe can become one of my dystopian nightmares.

    1. Great. Now I have a dystopian nightmare of encountering a klahanas merely loitering in a store. Isn’t there some kind of law?

      The paragraph above was intended to contain levity.

    2. After reading Jan Dawson’s piece on bots today, I kind of think that you are hitting a very important point that is not being talked about enough.
      For both Salespeople and Bots, before you can sell anything, you have to be at least minimally likeable. And this often means, you can’t just go talking about the stuff you’re selling all the time.
      Which might be related to why Apple gave Siri a character. Although far from perfect, Apple is obviously trying to make Siri into a bot that you would want as a friend. This might be a very important piece of the puzzle.

      1. It’s really about the unwelcome disruption. Yes, entering a store not only entitles, but demands the salesperson greet you and ask if they can help. When politely told no, anything else is now an invasion of personal space, especially when the pitch has nothing to do with why you entered the store to begin with.
        If I may add, entering a store is a much more overt activity (from a consciousness pov) than simply navigating to a web site. This potentially makes bots more intrusive. As a humorous aside, can I punch a bot in the nose?

  2. It was ever so clever of you to avoid mentioning the utter fiasco of Microsoft’s “teen social” bot that started spewing racist tweets after being allowed to interact with real people outside the lab. Because that wouldn’t go well with the marketing spin you’re selling in this article.

    Thinking bots can replace human beings for customer service is pure moonshine. We already have replacements for human beings for customer service — they’re called interactive websites and automated phone trees. If your needs are simple, they work well. If they aren’t, then they work horribly. 20 years of diligent research into finding ways to minimize expenditures on customer service agents has failed to make the phone trees or the interactive web sites work better.

    The whole “we still can’t even begin to figure out how to make a computer that can exercise judgement” issue aside, holding bots up as the new holy grail in the quest to reduce customer service staffing costs overlooks two huge issues

    First, if you need customer service, you probably are frustrated and upset and as social animals, we all desperately want to be able to talk to a fellow human being in such situations. Replacing the human on the other end of the phone with a bot merely increases the frustration and pain and makes the problem worse, not better.

    Second, the companies trying to minimize customer service costs aren’t actually interested in solving customer problems. Rather, they are seeking to prevent customers from doing anything that might cost them money. This is the dark underside of why the phone trees and interactive websites are still so woefully inadequate after decades of refinement. Phone and cable companies don’t want to permit you to downgrade or cancel your service on the website. Companies don’t want to make it easy for you to obtain warranty service when you buy a lemon from them, because that will cost them money. The very concept of customer service involves the company spending money on customers rather than receiving money from customers. And that is anathema to the moronic accountancy based management that is in control of most large companies. Thus, the entire customer service engine is tainted with bad faith decisions made by corporate accounting. The experience is bad and frustrating because it’s designed to be bad.

    Bot makers are basically seeking to provide improved customer service, but the companies they delusionally envison as their clients are seeking to degrade customer service. It’s easy to train humans to provide bad customer service — you give them scripts, you deny them permission to do certain things, and you’re done. It’s easy to make a website that cannot do certain things that would cost the company money. But if you’re going to make an AI-ish expert system bot that’s supposed to solve customer problems, then it’s not going to be something you can sell to the major corporations that could benefit from it, because they don’t want to solve their customer’s problems. And if you design a bot to provide customer dis-service, then you’re not going to have happy employees at your bot making tech startup firm, unless you only hire sociopaths.

    1. “…the companies trying to minimize customer service costs aren’t actually interested in solving customer problems.”

      Nailed it.

      A bot will only be as useful as the amount of information it has about me and my specific needs.

      I have a Life Mission to give as little information about myself to untrusted/unknown third parties as possible.

      Therefore, by definition, every bot is an intrusion into my privacy and a major source of annoyance.

    2. Even before that, I struggle to find a mildly useful AI, let alone a reactive/proactive bot.
      – MS’s Bob was a disaster.
      – Amazon’s music/books recommendations… are they even still doing those ? I really tried to commit way back when, but dropped it after the 10th round of recos for stuff I already had or didn’t want.
      – Google Now’s “you should read” does have a non-zero hit rate, maybe 5-10% is interesting-and-not-a-dupe, that’s similar to what I get from my handpicked RSS feeds.
      – The one smart-ish feature I use is Amazon’s “customers who bought that also bought”, handy to find a sleeve for a tablet, but mostly because Amazon’s regular, declarative UI is a mess.

      I think bots are lacking both the smarts and the data to be effective, for now. It’s nice they’re trying, and maybe in a few years we users will get something out of it, but for now…

  3. I think the issue is that our tolerance of bad bots is fairly low.

    If I want something done, the bot better get it right and do it quicker than I can, and achieve that all the time. The French saying is “nobody acknowledges when the soup is good” (but everyone complains when it isn’t). I think I’ll get irredeemably annoyed at a bot that’s 90% flawless 10% idiotic. That’s for on-demand bots.

    For interrupting bots, the bar is probably similar, depending on how intrusive the interruption is: it’s easy to ignore/dismiss an occasional voice alert, less easy for a pop-up that steals input focus and visibility from my current app. If they make me take my phone out, or even flick my wrist, I’ll get mad. If I just have to throw a glance at my glasses or lend 10s of an ear, probably less bad. I get annoyed at reminders I set myself, when I remember on my own and make the alert is useless.

    I’m thinking a nice way to make the bot’s smarts and actions available would be the way I (we ?) handle news (RSS) and messages (IM, email): put it in a dedicated app, or at least in a secondary window, for me to look at when I want. No interruption, not even the effort to formulate what I want: if the bot is smart enough to do it, it’s probably smart enough to anticipate the request. Then again, I’m alert-intolerant, maybe I’m in a minority. IRQ vs polling, heh ?

  4. “The mere act of visiting a web site could immediately launch an interaction with a bot.”

    I’m already seeing that, a lot. I’ll go to a website like a hardware store and soon a “Can I help?” message pops up. I usually click on “No” and that’s the end of it. It’s at the bottom of the list of things that bother me.