Many in technology paint a rosy future for conversational user interfaces (CUI) — and rightly so.
There are many use cases where chatbots, as an example of CUIs, have the promise to make things easier for us, especially when it comes to consumer-to-business communication or conversational commerce. After all, who in 2016 wants to download an app for every new vendor they do business with?
At the end of the day, technology advancement is all about saving time or making things easier. And what can be easier than texting or messaging “where’s my order?” to a vendor instead of wading through a website, creating a new user profile on a mobile app, or, worse, having to make a call to get an answer?
Because of the multitude of vendors and developers rushing into chatbots, there is a clear hype/risk factor currently in play, as with all hyped technology. There is a risk chatbot performance and functionality will not meet the elevated expectations set by exaggerated claims and overly optimistic tech media.
If the gap between reality and promise becomes too great, developers set themselves up for failure — at least in the short to mid-term. As researcher Roy Amara taught us, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”
Closely related to chatbots is the technology of personal/virtual assistants, which hold the promise of narrowing that promise delta.
A couple months ago, Siri’s original cofounders stepped out of their labs with Viv and presented their vision of conversational commerce. They expect Viv to do far more than Siri by being able to look up movie times and compare ticket prices of the theaters closest to you, placing or canceling ticket orders, and recommending a place to grab some dinner after the show. Viv promises to conduct all of these transactions from a single point of contact, without the need for standalone, third-party apps. But this won’t be possible unless essentially all brands and businesses have open APIs for chatbots to integrate with and aggregate from.
So what do you need to have in place to build a personal assistant that can deliver on the promises? When it comes to the conversational aspects of chatbots for B2C or C2B communication, there are four capabilities to consider:
1.) Ability to transcribe the spoken instructions to written ‘digital’ ones
Most conversational UIs (CUI) have this solved with varying degrees of success. Accuracy has gotten much better, and what’s possible today works well enough.
2.) Ability to extract meaning, intent, and sentiment from the written representation
A lot of work needs to be done still for CUIs to consistently and successfully interpret the subtleties of natural language, mainly around accuracy, domain knowledge, and supporting more languages before we get this right.
3.) Ability to convert intent into actions so you can ‘get stuff done’
This is where the rubber meets the road, where real life converges with the actionable requests made possible by the digitization of our lives. Building out every domain needed for a complex scenario takes a lot of time, and there are hundreds and thousands of such domains. Say you wanted to organize a barbeque. Finding and ordering coleslaw and smoked meats is the easy part. Coordinating invitees and finding a date that works for all in addition to taking into consideration weather forecasts and executing communications takes a little more intelligence — intelligence that needs to be scripted manually, by people.
The team at x.ai said they needed 2 years and 60 people just to code the relatively restricted domain of appointment scheduling via email. Unfortunately, there really is no “public domain” of A.I. for mankind, so if you want to build an appointment scheduling solution today, you will have to start somewhat from scratch. In a nutshell: there is no one A.I.
4.) Ability on the receiving end to receive communication about the desired actions
This is the piece most people ignore, and yet it is where most of the work needs to be done. It’s key to really meeting market and consumer expectations for personal assistants. Aside from platform economy companies such as Uber and Airbnb, most companies do not make their APIs available today. For chatbots to really reach their potential, all of our service providers — restaurants, banks, hospitals, pharmacists, airlines, florists, lawyers, etc. — need to provide APIs for chatbots to tap into.
In the internet’s infancy, it took many years for small businesses to understand the need to have a web presence, so the tepid acceptance of digital brand engagement is hardly new to chatbots. It will no doubt take years for companies to understand how bots can address the new engagement preferences of today’s consumer. Once companies see how easy it can be for their customers to just pose questions through a message and get an authorized answer right away versus searching a page on a website that may or may not have the answer, they will not want to miss out. But until public APIs are widely available for bots to integrate with, Viv and any other assistant will not be able to live up to the standards creators are setting and consumers are expecting. In other words: no APIs, no chatbots.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.