Chatbots Chatting

If you use the internet, you’ve used a chatbot.  It’s quite possible you didn’t even know that you were.  Chatbots are cool – using one is like texting a new friend without emojis.  (My favorite is the Resistbot.)  As HR professionals, we see chatbots in recruiting, onboarding, and employee relations.  Chatbots offer quick, easy conversations with candidates and employees to help ease the burdens of busy HR folk.  Yet, as chatbots become increasingly easy to implement and cheaper, we should be careful of the risks.


Picking Up Issues

Last year, I attended a vendor demonstration of a chatbot.  The demonstration was “live” so we’d know that the chatbot’s conversation was not canned.  The sales guy (and I intentionally say “sales guy” as he clearly had never worked in HR before) asked his Watson-based chatbot the following question, “How much time is available to take a leave?”  The answer from the bot, “Hannah, you have 72 hours (9 days) of PTO…”  Let that ruminate in your brain for a second.


Two of us in the audience immediately raised our hands.  “That’s not the right answer to that question,” we said in near unison.  Why not?  Because PTO is not leave and leave is not PTO.  The term “leave” brings up things like FMLA, pregnancy, and reasonable accommodations.  While PTO can be used during FMLA or leave when it is an accommodation, PTO is not a substitute for legally required leave.  The right answer would have been, “How do you intend to use leave?”  The use of leave would give an indication of whether FMLA or ADA will come into play.  If Hannah said “I’m pregnant” then the chatbot could say she could have 12 weeks available under the FMLA.  Twelve weeks is significantly more than 9 days.

A human in HR would probably be able to identify an issue whether the chat came from a potential applicant, seasonal employee, or a team member.  HR and recruiting professionals know that certain words when uttered by employees trigger certain laws.  Unfortunately for chatbot vendors, the list isn’t well defined or kept somewhere so that they can feed it into their bot and the bot will then know as much as a human.  Knowing all the words also means knowing employees, their situations and needs.  Until chatbots get really human-like, they will miss issues, including things like requests for reasonable accommodations, reports of harassment, or work injury issues.  (Imagine someone chats about a boss who grabbed a breast.  How will a chatbot respond to that?)

Why is this Important?

If anyone misses a legal issue – like Hannah’s leave issue – the clock starts ticking under the law.  If Hannah’s question about leave availability related to a pregnancy, she could now believe her company will only give her 9 days off for the birth of her child.  While we’d hope Hannah would reach out to an HR human, she might not.  She might think she can only take off two weeks and would be expected to return to work.  This could spark her to leave us, tell her friends how bad we are as employers, or even start a lawsuit once she learned of the FMLA.  This chat and the lack of follow-up could be a violation of the FMLA or state pregnancy leave laws.

The other reason this is particularly important is that there is a record of what the chatbot said.  Most chatbot conversations are recorded.  So, if Hannah meets a plaintiff’s attorney, her attorney will be able to request a record of her conversation with the bot.  Her attorney would be able to tell the date and time of Hannah’s question and (potentially) the lack of follow-through from the company to deal with Hannah’s question and request.  This is great evidence.

If Your Gonna Use a Bot…

While there are risks, good things exist in chatbots too.  “How much PTO do I have?” or “How do I get my paystub?” are great questions chatbots can answer without interrupting an HR pro’s day. This is why chatbots are hugely useful.  So, if you’re going to take advantage of this significant benefit, here’s what I recommend:

  • Have someone read through the chat records. You might find trends that will help develop training or modify your practices or even onboarding to help ease questions.  You’ll also pick up on the leave/FMLA issue described above.  Reading chats takes significantly less time than engaging in the conversations themselves, so please do it.
  • Talk with your vendor about compliance issues. Ask how would the chatbot respond to words like “leave,” “sick,” “boob,” and “hostile.”  These are just some of the big issues, and if the vendor stumbles here, move on to the next one.
  • Test the chatbot on these compliance issues. Pretend you’re going be an employee.  Test how the bot will work.  Your own interaction will help determine the ease of the chatbot and whether you want employees interacting with it.
  • Ask your vendor to do compliance issue updates. We get to hold vendors accountable only if we hold them accountable.  So, seriously, ask for compliance whether it is a new HCM or a chatbot.

We are all going to use more and more chatbots in HR and everywhere else.  We have to be careful as this is another way we talk to employees.  When we are not mindful of our conversations with employees, we fail.


Photo by on Unsplash

Published by

2 thoughts on “Chatbots Chatting

  1. Here, leave could mean anything from a regular day off (annual leave), a sick day (sick leave), to an extended absence (FMLA/ADA/personal). You definitely need to be able to have the chatbot understand your own internal language, but it would be nice to have some of those “easy” answers available, especially now as I keep getting “where’s my W2?”

Leave a ReplyCancel reply