A year with our recruiting chatbot

This is a review of our experience using a chatbot (built by Job Pal) in the recruiting process at hub:raum (the startup incubator of Deutsche Telekom AG) over the last ten months and our honest reflections on our expectations, the lessons we learned and the next steps we might want to take with our chatbot.

These are thoughts on things hub:raum has actually been doing since late-2016, when there were only a few articles around the impact of chatbot technology on HR, and especially on recruiting. We didn’t know how a chatbot would be perceived by our candidates, or how much it would take over from standard human interaction — in this case, in the recruiting process.

But that’s the thing with new technology: unless you try it, you’ll never know.

In theory, there are obvious advantages to companies implementing a chatbot, especially in high-volume hiring. During hiring conversations, many of the same questions are answered with many of the same answers, ad infinitum. Yet, beause it is humans who will use the chatbot, this is where the theory becomes a little unpredictable.

Over the last few months, we’ve had several hypotheses and some interesting observations:

#1 Jobseekers engage with hubbot mostly outside of working hours.

Knowing that talent is a scarce resource, and that most talented candidates will look passively for their next career step even when they already have a job, I would have assumed that the great convenience of chatting with a chatbot after work or on the weekends will show in the statistics.

But only 62% of interactions with hubbot happen when I’m off work, or on the weekends. To be clear: interactions are measured continuously, and it may be that candidates started conversations with hubbot while browsing job offers at work, but then continued them in the evenings or over weekends.

Yet, what this highlights is that 62% of talents are interacting with our employer brand while our workforce and recruiters are not working. According to an interesting (but not surprising) study from Indeed, the candidates you want most are more likely to interact with your employer brand during non-business hours. A chatbot is one good way to utilize this information.

I also believe that with the omnipresence of chatbots for all kinds of interactions with companies, be it for customer service or recruitment, users will adapt, and learn to embrace and use a real, 24/7 service cycle.

Which leads us to…

#2 Mobile vs. desktop chatbot usage.

I would have similarly expected to see jobseekers using hubbot more on their mobile phones than on their browsers. After all, we’ve built it on the Facebook Messenger platform, which is a mobile app on most smartphones.

But when we pulled the statistics, they showed that only 36% of candidates used hubbot on their mobile phones. That tells us that the engagement with a chatbot may be perceived as analagous to that with emails — another natural forum for professional communication.

But it also tells me that job searches don’t start on phones yet — or at least that ours don’t. If we had lured job seekers with mobile ads to a chatbot to build their application, then these figures would surely be different. Since we didn’t, it would appear that our candidates look for jobs via their desktop and then, most likely, continue the conversation on their mobile phones.

Not an unreasonable assumption: one of the main advantages of using a chatbot is its ability to re-engage candidates with messages that appear as notifications directly on the lock screens of their phones. But Facebook does not show detailed statistics on whether and how people switch platforms over the course of a conversation. This would be an interesting avenue for further investigation, should Facebook decide to make this data available.

And, as we see that our candidates are not “off grid” when they’re looking for jobs, it would be interesting to track trends in chatbot-interaction for jobs such as those in hospitality, with a high churn, quick application processes and mostly mobile searches and contact with employers.

#3 Candidates are asking questions I wasn’t aware they would have.

Although I have been working in HR for 17 years now, I was amazed to see the sheer number of questions candidates asked hubbot, as well as the vast array of topics they covered.

During the chat, candidates can ask any additional questions regarding our jobs or about working with us. If we haven’t provided hubbot with information sufficient for answering some of these questions, they come straight to my inbox. Then, I feed the answers to those questions to hubbot, who learns to answer similar questions independently in the future.

We began to do this only a few weeks ago, and already we’ve received a lot of questions of this kind — most of which are salary-related.

Here is a percentage of these questions by topic:

  • Salary: 27%
  • Job specifics: 24%
  • Hub:raum and our portfolio companies: 20%
  • Recruiting process: 11%
  • Work culture and office environment: 9%
  • Chatbot related: 7%
  • General questions: 2%

And it’s nice to get this information from a chatbot. Everyone has fun doing this with Siri or what have you. Which brings us to my fourth point:

#4 People love extracting information from algorithms.

It seems as though candidates understand the advantage of not speaking to humans: skipping the BS but also gathering information you couldn’t have asked for in your first conversation with a recruiter, for example, regarding your salary.

It is also obviously clear to potential candidates that they are interacting with a machine and not with a human. This is something we were concerned about going in, and the reason we named our chatbot “hubbot,” to avoid its being mistaken for a human interaction.

Another concern for us at the beginning was the possibility that some candidates would be repelled by using a private channel — namely, Facebook messenger — for a professional purpose. We can see from the questions that have been asked, though, that people are aware that we see neither their personal Facebook profile nor any personal information; not even their names.

Until a candidate actually applies, the whole conversation, and every question asked, remains anonymous. Even when they apply, I don’t see their conversation history.

This particular benefit is one I didn’t expect but which I completely love about our chatbot. The power of information extraction is entirely in the hands of our candidates, whereas in a traditional recruitment process, I as a company would have been in charge of what information I provide to a candidate, and what I don’t. With hubbot, candidates can ask me literally anything, and I must answer any and all questions to maintain (and even improve) candidates’ experience with hubbot.

By the way, my highlight is: “Where do I get the most money?”

The time I spent answering these questions from candidates also got me thinking:

#5 One must never underestimate the personality of a chatbot.

Feeding hubbot with answers and information it can learn from and use in every future conversation, I began to intuitively assume the tone we used when we built hubbot: a bit witty, mostly casual, friendly and polite. Because that’s how we are at hub:raum! Because we want hubbot to reflect this in its conversations with all our potential team members.

This is a very powerful tool for any talent acquisition expert: a chatbot that transports your company’s tone consistently, in every conversation with a candidate.

This is the benefit I least expected in using a chatbot.

Which brings me to the pressing point:

#6 Fellow HR people still are unsure about chatbots.

Since we started using hubbot, I have talked endlessly about it at various occasions, including panels, webinars and conferences.

Most of the time, the audience is made up of general HR and recruiting professionals interested in hearing about a chatbot implementation, and what we’ve learned from it. What I observed is that most industry professionals are hardly aware of the advantages of a chatbot, and still consider it a “fun feature,” making their company stand out.

This is not exactly surprising, since “HR Tech” is fairly new, but it also tells us that we need to continue educating people in HR on the positive impact of technology on their work.

For the better, tech will let us spend more time on human interactions. When algorithms take over repetitive and inefficient tasks, we can get back to what we signed up for: human relations.

Because:

#7 You can be human and still use a chatbot to the advantage of you and your candidates.

One of the most common areas of concern addressed to me when I extol the virtues of our chatbot: that it is not human and therefore not what HR should be about. But our statistics show that sometimes, candidates prefer machines over humans. Instant, 24/7 information extraction, to-the-point interactions and anonymity are all only things a chatbot can give you: the ability to feel out your experience and expectations and take things at your speed.

Next steps and work in progress:

Over the next months, we will continue feeding information to and training hubbot to allow it to

  • re-engage more with candidates from previous conversations not only based on newly vacant and relevant job positions, but also based on other content, career related events and other things
  • get feedback from candidates on the chatbot experience

If you have any additional ideas on what we could build into the chatbot, please let us know in the comments below!

Anna Ott is the HR expert at hub:raum, the corporate incubator of Deutsche Telekom. Luc Dudler is the founder and CEO of JobPal.