<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=979905748791482&amp;ev=PageView&amp;noscript=1">
Future of ChatbotsNovember 17, 2017Written by Alex Debecker

Quasi Case Study: Chatbots Powering HR Transformation

hr transformation chatbot

Our most avid readers will know this: HR is paving the way for chatbot adoption.

Our pipeline is full of massive companies trying to improve their HR processes. The industry in general is experiencing the same thing. We see articles from other companies like ours reporting HR builds.

How are chatbot driving transformation in HR?

I'd like to take the opportunity of this article to share an actual build we are in the midst of deploying. This quasi case study (more on that below) will give you real-life insight into the work we do for large businesses. It should also give you some inspiration for what you could achieve in your own business.


The company

As you know, I cannot share much about the clients we work with. This particular client may actually be an exception. They have asked us to publicise the work we are doing for them as much as we can.

Unfortunately, this 'identity embargo' only lifts after we deploy the final build. So, I will have to remain silent until then.

For now, consider the company to be a very large and well-known financial service company with over 150,000 employees.

Let's call the company Walled.

walled fake company logoEven made them a little logo. Because, why not?

This name/logo is completely made up. Any resemblance to a real business is pure coincidence.


Could a chatbot help?

Walled contacted us about HR. Servicing their growing workforce had been a challenge for a long while. They had learned to accept that reaching their SLAs or KPIs may never be possible.

Now, with chatbots poking their heads around the corner, they saw an opportunity. And pounced.

  • What if chatbots could help them reduce some of their workload?
  • What if chatbots could help them answer easy questions on the spot?
  • What if their SLAs finally became achievable thanks to new technology?

They contacted us not about delivering a piece of software. They contacted us about transforming their HR department and service.


HR transformation, one meeting at a time

During our first meeting with them, it became clear we could help them. Walled showed us their HR processes, their SLAs and KPIs, and the margins by which they miss them every month.

When we dug a little deeper, we found a series of low-hanging fruit. If we could take over on these, we would already give them 40% of their time back.

Together, we devised a three-stage plan.


Stage one: proof of concept

In our experience, stage one of any chatbot project is always a proof of concept. In fact, we push clients to do so.

Chatbots are new technology. Even if everyone is excited about them, they still need to prove their worth to some people on your team. A proof of concept allows us to put everyone on the same page.

We go away, develop something that will show everyone the power of chatbots and show it to them. Done.

chatbot proof of concept

For Walled, the proof of concept is simple. We need to show our chatbot can tackle the questions most frequently asked by their workforce. The lowest of the low-hanging fruit.

If we can do that and amaze the board, we're on to the next phase of development. Of course, we did a great job at this and got to move to stage two.

Note: Out of all the projects we have worked on at ubisend, we have never ever bombed at a proof of concept. Instead, we have found excitement rises beyond our dreams when we put people in front of a working prototype.


Stage two: supervised testing

Stage one was about building the chatbot's function well enough to impress everyone. Easy peasy.

Stage two is about testing the chatbot in real-life situations. We are, of course, early on and thus should supervise the process. Walled is a highly respected financial institution after all, we need to make sure we do things properly.

During stage two, we invite a series of beta testers (aka employees) to use the chatbot. Over the course of two to three months, 50-odd employees will interact with the chatbot and report back to us.

It's important to note that these 'beta testers' are not pretending to use the bot (aka reviewing the chatbot). That's done in production and in-house, way before anyone 'real' plays with it. They are actually using the bot in real-life situation, asking HR-related questions.

Stage two allows us to learn from interactions. This not only trains the chatbot's AI, it also gives us insight into how people are using the bot. Do the interactions make sense? Does the conversational UX flow properly? Did the users get what they came for? If not, did they get pointed in the right direction?


Stage three: system integrations

Like most of our clients, Walled need us to integrate with their systems. They have employee databases, HR portals, and more.

All this information is highly valuable. It will train the bot and make it more efficient. It will help us give personalised answer to each chatbot user, instead of general advice.

Stage three is all about integrating the bot with these custom systems. As you can imagine, this is a significant part of the project. This is where the real HR transformation happens. No longer will the employees have to log into 12 different platforms. No longer will they have to email the right HR inbox to get their answer. No longer will they have to wait 3 to 5 business days for an email back.

This is where the magic happens.


Transforming HR in 3 (not so) easy steps

Just like that, we have disrupted the entire HR department of a massive company. These three stages can take a long while to get through -- and it doesn't stop there.

Typically, we would then deliver a working product and release access to the entire workforce. The next steps are iterations of that product to make it better, and better, and better over time.

A chatbot, much like software, is never 'done'. We work closely with the team to continue to deliver amazingness.