Wednesday, July 12, 2017

Checking Out Chatbots

I recently spent a day with the Oracle Applications UX team in a Conversational UI for Enterprise SaaS workshop.  Let’s be clear…in the context of this workshop, “Conversational UI” is a spiffy term meaning chatbots for enterprise applications.  Amazing workshop put on by the UX team.  Made even better by including a mix of attendees ranging from neophytes like me up to experienced experts sharing their tips and tricks.

I learned quite a bit about what Oracle is currently doing with chatbots, how to design a chatbot and how to build a chatbot.  I was inspired to the point of staying up all night to build a chatbot from scratch.  Hey, I’m on the road with little to do at night other than stare at hotel room walls, so what are ya gonna do if you don’t geek out?  Some takeaways from the workshop and my own research on chatbots:
  • Messaging apps are growing at insane rates.  For example, consider Facebook Messenger.  It’s used by over 1 billion people every month and is outpacing the growth of Facebook itself.
  • Getting things done with a bot is much faster than working through a website or mobile app.  While websites and mobile apps have to be loaded and navigated, bots load instantly…and people will consistently choose the path that loads the quickest.
  • Bots win on the ease of use front as well.  No navigation needed with a bot…just start the conversation. And…this is key…language is the interface people understand best, and it’s the interface used by a bot.
  • We’re at the very beginning of developing and applying bots.  But think of the potential:  would you rather navigate Amazon’s website looking for red posthole plugs, or simply ask for red posthole plugs and have them shown to you?  There are millions of simply tasks better performed by bots than through a website or mobile app…which could mean that bots will greatly reduce our current reliance on websites and mobile apps.  You can get more of an idea here 
  • Learning to code a chatbot is easy.  Lots of choices for coding languages, along with drag-n-drop IDEs.  I think it took me about 15 minutes to pick it up.
  • While the coding is easy, the underlying logic is not.  Lots of variables in regards to different terms that mean the same thing, providing easy exits to users who find the bot frustrating or unusable for their particular purpose.  Even very simple tasks require some substantial brain power for laying out the logic involved in completing a task through a conversational interface.  We’re talking about some complex decision trees.
  • To keep your underlying logic simple and to keep your users on track, build chatbots for very basic and focused tasks:  open a service request, buy a pair of shoes, add a dependent to an employee’s benefits plan…that kind of thing.
I came away from the workshop with a pretty substantial list of use cases for chatbots in the enterprise.  I’ll just share one example here.  

In the Oracle HCM Cloud Center of Excellence team, we have a group of experts who help strategic customers with technical issues.  We take in requests for their help through the COE Request Applications, which is a service request database created in Oracle APEX.  I’m thinking that a chatbot for creating requests would be a better user experience than the current data entry form that requestors must complete for each request.  So I spent part of yesterday’s workshop creating a low fidelity wireframe for such a chatbot, which I’m sharing here (note the phrase “low fidelity” - no taking potshots at the appearance of a wireframe I put together in about 20 minutes):




The upshot is that I walked away from the workshop pretty excited about chatbots.  And I’m hoping that my takeaways might whet your appetite.  If you want to know more, you can look here.

So I'm admittedly a neophyte beginning my exploration of chatbots.  What about you?  Any experiences?  Feedback?  Thoughts?  Show some comment love and share with the rest of us.

No comments: