Discussing Conversational AI with IBM Think Leaders
Aug 15, 2019
Our Wade & Wendy CEO, Drew Austin, was recently interviewed on the excellent IBM thinkLeaders podcast. The podcast hosts — IBM Watson’s Emily Winchurch and Laura Donaldson — chatted to Drew about the rich and fascinating world of conversational AI. It was a great opportunity for Drew to discuss our AI recruiting assistant, Wendy, and how she relates to the broader landscape of this cutting-edge technology.
Conversational AI, building trust, & making conversations human w Clément Delangue & Drew Austin
Emily and Laura are familiar with Watson Assistant, IBM’s AI conversational assistant — so they were able to host a very engaged discussion. Enlivening the conversation even further was the fact that it wasn’t only Drew being interviewed. Also present was Clément Delangue, co-founder and CEO of Hugging Face. Hugging Face is a fun and light-hearted chatbot app, described by TechCrunch as “an artificial BFF.”
Having Drew and Clément in the room together made for a fascinating contrast. Why? Because their two companies represent the full range of what conversational AI can offer. As Emily said, with Hugging Face you have the tech being applied to “interactions that are just fun,” but with Wade & Wendy it is helping “get the job done faster.” Talking Dog is aimed at teenagers, during their leisure time. Wendy is aimed at professionals, during their workday.
Yet despite the difference in approach, the similarities in their philosophies are fascinating. Drew and Clément both agreed about the importance of natural language. Hugging Face are “obsessed by natural language,” said Clément; obsessed with understanding and generating it. Similarly, Drew explained how at W&W we are striving to truly “understand the language and lexicon of recruiting, and then build a conversational platform that enables us to have real recruiting-oriented conversations.” It’s only with real natural language that an AI can help with recruitment, Drew said: “Everyone has their own career story, everyone’s situation is very personal. You need to be able to truly speak to that.”
In both instances, the goal is humanization. Despite scary Robocop-esque predictions, said Delangue, real experience with AI can show people that it is working for them, not against them. The whole point of Talking Dog, he said, is that it is entertaining, rather than transactional (like Siri or Watson). It is meant to bring joy to people:
“Human beings are social. Most people have pets, which are another form of non-human intelligence; and you enjoy playing and interacting with them every day, with no utility goal, and with a real emotional connection. It makes sense to think of human-AI interaction in the same way; every human can be willing to form a connection with an artificial intelligence.”
Drew echoed these sentiments when discussing Wendy. He explained how an AI carrying out early, top-of-funnel conversations actively promotes more human interactions overall:
“When you introduce an automated conversation into the process, it can actually make the following in-person conversation that much better. Because now there’s more context, more information; you’ve gotten out of the way the repetitive and mundane question-and-answer. And you can now bring all this to a more human conversation, and dive deeper.”
“The whole vision,” Drew continued, “is that humans and machines can collaborate together to provide better experiences for everybody. We say: allow Wendy to do machine-oriented, repetitive tasks that you can easily scale. Then humans can focus on relationship building, and getting to know people. When you have that automation up front, you can put that time into a human relationship.”
One of the hosts, Emily, recalled how her experience of job-hunting has included a lot of non-responses, which she found very discouraging. This is exactly what we want to put an end to with W&W, Drew explained.
Another key area where Drew and Clément found themselves in agreement was that of trust. Clément and Drew agreed that it is imperative to be transparent, and always let people know they are talking to an AI. Never have your bot pretend that it’s a human.
“We’re about to have a conversation about something very deeply intimate and personal: your career, your life, your interests, your goals,” said Drew. “To start that on the wrong foot and pretend that you’re a human is just not the right way to build the foundations of trust, and I think trust is essential to how we have interaction and conversation.”
Clément also explained how at Talking Dog they have “a whole set of features” that ensures the Talking Dog AI is treated in a light-hearted way. They encourage people not to take the app too seriously; not to think of the app as something that will replace their therapist, or human interaction.
Emily and Laura closed the podcast by asking: What’s next? How do you plan to make your iterations of conversational AI even better?
“Domain understanding,” said Drew. “We need to make sure we continually expand Wendy’s knowledge of different professional domains, and encourage a focussed growth of her knowledge of certain types of work.
Also, introducing empathy. When someone says they’ve been laid off, Wendy can say ‘I’m sorry to hear that’, rather than just rushing onto the next question. That bit of empathy makes people understood. People want to humanize machines. People say ‘thank you’ to Wendy at the end of conversations. They don’t say that to an applicant tracking system.”