“She fell into tech because of her love for creating and building,” EstherBot told me. “That’s even how I came to be. Because it’s 2016 and bots are everyyywhere!”
EstherBot is a creation of San Francisco product marketer Esther Crawford, who created the automated chatting tool to, essentially, market herself. With chatbots popping up to help users do everything from ccheck movie times to book airline tickets, Crawford decided to create a bot to answer questions about her career history, educational background, and hobbies.
“I wanted to use a bot to tell the story of how I got from having a master’s in international relations to being a product marketer for startups,” she says. “I’ve been fascinated by the emerging messaging and, specifically, bot space, and I felt like with my career and what I’ve done, I could just kind of merge those things together through the bot.”
The bot, which can communicate through Facebook Messenger or text messages, sprinkles its responses with all-caps keywords, indicating topics it’s capable of speaking about. The bot doesn’t understand full sentences or open-ended questions, only those keywords it provides. That makes talking to it more like navigating a dialogue menu in a video game than having a normal text message exchange. EstherBot is clearly a tool for exploring Crawford’s creativity and the information she chose to store in it, not an open-ended conversational bot like SmarterChild in the early 2000s or Microsoft’s ill-fated Tay.
The goal, she says, is to automatically answer the kind of questions about professional experience and cultural fit that recruiters often have—and that jobseekers can hear more than a few times during a job search—but that are too specific or idiosyncratic to put on a résumé or in a LinkedIn profile. Those could include basic qualifying questions that they’d now have to ask through email or on a phone call, like what cities a potential candidate would be willing to live in, or information about personal interests and side projects.
“One of her favorite pasttimes is dancing all night in a onesie,” EstherBot said, after mentioning that Crawford enjoys traveling for festivals. “She’s a lead organizer for her BURNING MAN camp and brought her first art installation last year, which was an interactive experience featuring digital motion art.”
Crawford says she’s not an engineer, but she was able to build the program with almost no coding using the Smooch bot-crafting platform, integrating support for Messenger and low-cost texting through Twilio. She’s released instructions, open source code, and sample scripts for the project, and says she’s had feedback from people who have created their own résumé bots.
The bot has also gotten a fair bit of publicity after being promoted on Product Hunt and in a Medium post Crawford wrote about its creation. And, she says, she’s already had a few potential employers reach out through EstherBot using a “connect me” option that appears during long conversations about potential job opportunities that sometimes sets up interviews with Crawford herself.
“I actually have had a couple of companies reach out, and I had a conversation with one this morning—they were really interested and excited about talking to me,” she says.
She’s also shared the bot with potential employers she was already in touch with, she says, partially as an example of her marketing skill and her ability to shepherd a product through its launch and engage an audience. Her current team is aware of the bot as well, she says, knowing that it may ultimately lead to her departure.
“They’re supportive of me building the bot and using it in whatever way makes sense for me personally, even if it is a catalyst for launching into my next thing,” she wrote in an email to Fast Company.
EstherBot is far from the only project harnessing technology to make the hiring process more efficient, says Brian Delle Donne, the president of Talent Tech Labs, a New York startup accelerator focused on technology for the job market. Other innovators in the field are building candidate-relations management tools to better organize information about potential hires, similar to customer-relations management tools used in sales, and building tools to help recruiters scour forums and bulletin boards for potential candidates, he says.
And others are working on bots like EstherBot that represent potential employers, answering job seekers’ questions before they send in applications or even reaching out to potential hires based on their online activities.
“There are bots that can be built in corporations and exemplify the persona of the corporation and work on the other side,” Delle Donne says. “Generally, it’s a move toward efficiency in reducing a lot of wasted time that goes on–candidates aimlessly sending out résumés or responding to jobs that are not a good fit.”
The challenge for both candidates and companies, though, is to create bots that are actually engaging to talk to and don’t feel like the text equivalent of a tech support phone menu, he says.
“Think of the inbound voice messaging—you call up Verizon about a problem, and you go through 15 screens of prerecorded messages, and even with voice recognition that acts like it’s engaging with you, you know that it’s not authentic,” he says.
Chatting bots can certainly be valuable if they make hiring conversations more efficient, argues Sharlyn Lauby, the president of human resources consultancy ITM Group, pointing at a survey from Glassdoor Economic Research showing the time it takes companies to hire candidates nearly doubling since 2010. But, she says, both candidates and recruiters will have to be patient when early bots aren’t perfect conversationalists.
“Both recruiters and candidates are going to have to be a bit forgiving until using a chatbot tool is perfected (or at least mainstream),” she wrote in an email to Fast Company. “I could see a chatbot being valuable for more objective responses or specific details. Then save the rest of the conversation for the actual in-person interview.”
Right now, EstherBot doesn’t pretend to be able to say anything apart from its prerecorded conversational cues, and by using text instead of speech, it avoids the problems of bad voice recognition and long, impossible-to-skip messages that can make phone trees such a nightmare. But once the novelty wears off, it’s clear that companies and job candidates alike will have to tread carefully to make their chatbots strike the right balance between being entertaining and efficiently informative–something Crawford readily acknowledges.
“Companies who gets serious about messaging will need to hire storytellers, comedians, and Hollywood types, because strong narratives will make or break these products,” she wrote on Medium.
Crawford says she’s already tweaked EstherBot to add responses to common questions she didn’t anticipate when she first launched the bot. Users naturally play with the limits of the bot’s conversational abilities, asking it questions about the weather or even sending it lewd propositions, or just try to talk to it with more sophisticated phrases than it can handle.
“Humans chat in such a particular way, and they have really high expectations when talking to a bot, even when they know it’s a bot,” she says. “I updated the script yesterday probably 50 times based on how I’d seen people interact with EstherBot, not the way I initially imagined they would.”
She also forwards the bot’s conversations to Slack, where she can jump in if a user is particularly perplexed or interested.
“If you get far enough down into the conversation, there’s actually an option to disconnect from the bot and connect directly to me,” she says. “That enables people who are more serious and interested in talking to me to stop talking to EstherBot and start talking directly to Esther.”
In the future, she imagines, job search platforms and services like LinkedIn might offer built-in support for more sophisticated bots, using natural language processing to converse more fluently with recruiters about candidates’ strengths. And in the meantime, Crawford says, EstherBot has already exchanged more than 24,000 messages since its launch last week.
“It’s kind of like a funny 1.0 version of what the future of recruiting looks like,” she says.