This article was originally published on CFO.com

At least 40% of online transactions could be handled via a bot by 2020.

Last week was the annual “Tech Week” here in Chicago, highlighting a number of themes from the Midwest startup scene (we’re not Silicon Valley of course — not even up to Silicon Alley as yet, but there’s a surprisingly vibrant tech community). There was a lot of internet of things activity, as well as some interesting developments in payments and in commerce platforms in general. But the topic that caught my attention was the breadth of activity around assisted and automated services — around the proliferation of “bots.”

You’ve probably already encountered a “bot” in one form or another, in chat services for online help, for example. Possibly even via the artificial intelligence-enabled voice response systems that are proliferating in call centers and replacing one flavor of customer service frustration (idiotic interactive voice response menus) with another (idiotic voice recognition Q&A sequences). If you use Siri or Cortana or Alexa, you’re using a form of automation too. It you own a “smart” thermostat, that’s also a kind of bot. There’s been a lot of progress in these areas over the past few years, and we’re seeing an accelerating rate of improvement, even if much work remains to be done to lower the frustration level for the average consumer.

One of the companies I visited with last week shared with me that they expected “at least 40%” of online transactions to be handled via a bot by 2020. What exactly that means depends on the kind of transaction.

In some cases, it will be a simple pre-determined action that is user initiated (Amazon’s buttons for automated reordering of specific items, for example) or automated when a predetermined condition is met (Samsung’s internet enabled refrigerator). In other instances, it will be a capability for automated routing of a call to an appropriate human agent, while relevant information is assembled in the background so that the agent knows all about the contact before a connection is made. In others, it might be an entirely automated response to a common (frequently encountered) inquiry or request (book a dinner reservation or a flight for example), with the system autonomously improving its ability to respond by analyzing the issues it isn’t able to respond to automatically and learning from the actions of the human agents that provide satisfactory solutions.

There are a lot of technology, sociology, and productivity factors driving this. Automation for routine tasks has been around for a long time, and the application of analytic techniques to vast accumulated collections of events, actions, and outcomes spreads the range of what’s “routine” ever wider. Sociologically, we have an entire generation (almost two) that’s habituated to not dealing with each other face to face for an ever wide range of normal everyday activities (texting, anyone).

Today’s bots generally aren’t very flexible, are often limited in what they can do, and can make quite egregious mistakes when their “learning” abilities misfire, but then human agents aren’t all that close to perfect either. There will probably always be situations, especially related to errors and exceptions, that only human agents can handle, but as these become less and less routine, they will require much more highly skilled people to find workable solutions. As I have noted in previous columns, we’re going to have to figure out how to train these highly skilled people when there are far fewer opportunities to acquire skills through progressive experience.

We’re also going to have to do some work on the sociology of human-bot interactions. As a friend of mine remarked recently, not only are bots becoming more human in their behavior (and thus harder to identify as bots), but human agents are becoming more robotic, trapped in scripted dialogues that leave little room for initiative. If you’re going to make people behave that way, you might as well use software, which doesn’t get tired, angry, frustrated, or need breaks and vacations — and which can be updated as needs change and capabilities improve.

Like it or not, the bots are coming. We can make good and bad use of them, and will probably do both for some while yet. But they’re definitely coming. Better get ready.

About the Author
John Parkinson

John Parkinson is an Affiliate Partner at Waterstone. John brings extensive experience to the topics of technology strategy, architecture and execution having served in both senior operating and advisory roles.