No Jitter is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Tracking the Evolution of Genesys' Kate

  • Meet Kate opening slide
    “Meet Kate. Kate is unique. She is not your average girl. She is genuine in everything she does. Her loyalty is phenomenal and you never have to doubt her. Kate’s kindness inspires others to do good things. Kate is not weak though. She is a strong independent girl who can get herself through anything she sets her mind to. She is hilarious! Sometimes she has a dry sense of humor and other times she just knows how to make people laugh. :)
    Kate has been told by many that her advice changes the way they look at life. Kate has a very bubbly personality that makes anyone comfortable talking with her. She is able to learn quickly and apply. Kate has many good personality traits but she is not self-conceited. As a matter of fact, she has some downers too. Sometimes Kate can be a little too sarcastic. That can offend some people, but she apologizes. She also has a tendency to procrastinate from time to time. But she still manages to get everything done.”
    This is the backstory for Kate, the intelligent virtual assistant Genesys CMO Merijn te Booij first described in May 2017. The Ask Kate intelligent virtual assistant deployed to support conference attendees using the mobile application at last week’s Genesys Xperience19 event is fulfilling this initial promise. But Kate’s path has been a journey, just as it is for customers when they seek to purchase a product or solve a problem.
  • May 2017: Kate Introduced
    May 2017: Kate Introduced
    With its acquisition of Interactive Intelligence finalized in December 2016, Genesys brought together customers, partners, consultants, analysts, and media of both companies for its first combined event, CX17, in May 2017. As I wrote at the time, Genesys publicly announced Kate, described as customer service-specific artificial intelligence (AI), at that event.
    From a solution perspective, Genesys described Kate as combining AI, bots, machine learning, and micro-applications so companies could deliver personalized, proactive, and predictive experiences while running smart businesses. The combined power of Kate and live staff to solve customer problems is what Genesys was, and is, calling "blended AI."
  • 2018: Kate Becomes a Platform
    2018: Kate Becomes a Platform
    While Genesys discussed Kate the next June at CX18, she wasn’t the “star of the show” as she was in 2017. As discussed in the No Jitter review of top news at the event, Predictive Routing, stemming from the February 2018 acquisition of Altocloud, took center stage (in the post, Kate got reduced to a mention in the last sentence).
    During the event, I posted the tweet, shown above, referencing Peter Graf, Genesys’ chief product officer, who at the time of CX18, had held the title for just a few months. During his keynote remarks at CX18, he positioned Kate as a blended AI platform.
    As shown here, the event featured a breakout session titled, “Kate, Chatbots and Conversational AI.” It included a Kate demo. The fact that the demo used a fictious brand left attendees with the impression that Kate wasn’t, at the time, generally available.
  • Xperience 19: Ask Kate
    Xperience19: Ask Kate!
    I was excited when I first opened the mobile application for the Genesys Xperience19 event and saw “Ask Kate” as one of the options. It was Saturday, June 8, two days before the event was formally scheduled to begin. But that didn’t stop me.
    The industry analyst meeting being held in conjunction with Xperience19 was due to begin on Monday. As the event would be taking place at the Gaylord Rockies, a new hotel in Denver I hadn’t yet visited, I decided to ask Kate for some directions.
    I asked Kate the location of meeting room Maple 3. Kate responded that she couldn’t help me with that and that her “humans” were all en route to Denver. I should check back later. When I asked another question about the hotel, Kate responded that I should ask her a question about Xperience19. Hmmm. I expected pretty basic information about the venue to be available.
    I then asked what time registration would be open on Sunday, June 9. Kate responded with the available times for Monday, Tuesday, and Wednesday. Better, but not perfect. I would have preferred the response to include the fact that registration didn’t open until Monday, so there was no registration on Sunday.
    I shared my thoughts with the Kate marketing team, suggesting that information about the hotel be added to Kate’s bag of tricks.
  • The  Ask Kate Team
    The Ask Kate Team
    Since Kate’s introduction in May 2017, Christopher Connolly and Cliff Bell have been a marketing/product tag team for her. Connolly did the first (standing-room-only) sessions at CX17, and he and Bell jointly presented the Kate session at CX18. Dan Stoops is a more recent addition, and one of the technical leads on the Kate team.
    I directed my note about my early experience with Ask Kate to Connolly. He quickly replied saying, in part, “As the week starts, there will be a team that is helping train her in real-time, so I’m hoping you see an improvement.” He went on to say, “If you have time, we’d be happy to step you through how Kate works. We’re using a number of (Natural Language Processing) NLP engines in concert. Once her humans are watching the interactions, we should start to build the corpus.”
  • Ask Kate Performance
    Ask Kate Performance in the Xperience19 Mobile App
    Late the morning of Wednesday, June 12, three to four days into the live deployment of Ask Kate in the conference mobile application, I sat down with Connolly, Bell, and Stoops. The first thing they showed me was a live application of Kate’s performance, a screenshot of which is shown above.
    The dashboard has a number of elements. The statistics shown are as of noon on that Wednesday.
    • Intent Accuracy -- In chatbot terms, an intent is the user's intention. For example, if a user types “show me yesterday's financial news,” the user's intent is to retrieve a list of financial headlines. Ask Kate achieved 82% accuracy in determining user intent.
    • Answered Questions -- Self-explanatory, the number shown equates to roughly one question per registered attendee (approximately 2,500). It’s likely, however, that the number reflects questions from repeat users.
    • Escalation Offers -- Refers to the number of times that Kate couldn’t directly resolve a query and offered to transfer the interaction to live agents, referred to as “Kate’s humans.”
    • Recent Intents -- This is a graphical depiction of the categories to which intents belong. At the time of this screenshot, 25% of the inquiries to date were categorized as frequently asked questions (FAQs). You could scroll over each of the segments to get a category name and percentage. FAQ is the orange segment of the circle. The purple segment represented “Where” intents (21%), and all the different things that could be where -- for example, where sessions, meetings, breakfast, and receptions are taking place, Stoops explained. The gray segment represents the “When” intents (13%), and green shows “Agent” intents (10%), for attendees who specifically wanted to talk to a human assistant.
    • Recent Topics -- This word cloud shows oft-used words and phrases. “Shuttle” shows high use, referring to questions about the shuttle than ran between the Gaylord and overflow conference hotels. “Smoothie Bar” refers to the refreshment station operated by Genesys partner Altivon, with questions about where and when that was available.
  • Four-minuted update for Kate
    A Four-Minute Update
    On the first day of the conference, Connolly sent me a message that said, “Ask Kate for a coffee.” I never got around to it so when I sat down with the team, Bell reiterated the suggestion. He said I wasn’t limited to coffee but could ask for anything available at the coffee shop or hotel store, Marketplace. As shown on the previous slide as a recent question, I asked, “Kate, may I have a Pellegrino?”
    As shown at top left of the left-hand screenshot of my iPhone, I asked this of Kate at 10:34 a.m. The response from Kate shows that she didn’t understand the question and would require human intervention to help.
    After working to update Kate to understand “Pellegrino,” I got a different response to the second request I made four minutes later. As you can see in the right-hand screen capture, Kate now told me, “I can help you get a Pellegrino.” Note that Kate still said she had to get one of her humans -- virtual assistants can’t pay for goods, nor can they deliver them.
  • Ask Kate intents & entities
    Ask Kate Intents 
    So how did Bell and Stoops train Kate to understand Pellegrino? To explain it to me, they started by opening a tab in Google Dialogflow called Intents. Those with careful eyes will have noted that on the Ask Kate screenshots on the previous slide, the phrase “Built on Google Cloud Platform,” appears at the bottom. Dialogflow (formerly, Speaktoit) is a Google-owned developer of human-computer interaction technologies based on natural language conversations (see No Jitter’s ongoing “Decoding Dialogflow” series, from Brent Kelly, of KelCor).
    User intents are given a name, and you can see several -- agent, directions, and items, for example -- here. Kate was given pre-training, which is the developer team’s best estimate of what Kate would need to know, Stoops said. Training continues once the system goes live, and as he said, “nothing compares to live.”
    Kate hadn’t understood “Pellegrino” since no one else had made an intent request for that item, Bell said. Notice that the above page is one of 13, so at the time we looked at it, Kate was running about 240 intents.
    Ask Kate Entities
    An entity modifies an intent. The initial list for the items intent included coffee, coke, and water… but not Pellegrino. Bell added that, and, at my suggestion, also added a synonym, sparkling water. The new entities were added to the platform and Bell asked me to try again. Four minutes later, Kate understood Pellegrino, and one was on the way.
  • Kyley Eagleson: One of Kate's Humans
    Kyley Eagleson: One of Kate’s Humans
    A short time later, Kyley Eagleson -- one of Kate’s humans -- showed up in the conference room where we were having the meeting. She never asked where I was, so I assume the application picked up my geo-location from my mobile phone. I also assume I gave them permission to do that at some point.
    In addition to her conference duties being Kate’s human assistant, one of Eagleson’s responsibilities is to write Ask Kate responses. She was the fourth member of the Ask Kate team.
    One of the reasons I’ve dedicated this slide to Eagleson is I think she embodies the spirit of Kate first described by Genesys CMO te Booij in 2017, shown on slide 1. “Kate is unique. She is not your average girl. She is genuine in everything she does. Her loyalty is phenomenal and you never have to doubt her. Kate’s kindness inspires others to do good things. Kate is not weak though. She is a strong independent girl who can get herself through anything she sets her mind to. She is hilarious! Sometimes she has a dry sense of humor and other times she just knows how to make people laugh. :)”
    When I asked Eagleson if I could take her photo with the bottled water, she immediately went into “Vanna White” mode. Though she and the team are busy monitoring thousands of requests to help with the training of Kate and running around with drinks from time to time, her unbounded enthusiasm for what she’s doing came across.
  • Kate at Xperience 20
    Can’t Wait to Ask Kate Next Year!
    Genesys announced that Xperience 20 will be held next May at the same hotel, the Gaylord Rockies. With what she learned last week, I’m sure that Kate and the team will hit the ground running. I’m excited to see what may be new for Ask Kate next year… maybe a voicebot equivalent?

From concept to platform to generally available intelligent virtual assistant