Automated Chatbots: How Are They Progressing for Helpdesks?

One of the major developments over the last few years is the use of automated chatbots for help desks.

While these started out as relatively rudimentary applications, they have evolved to become more intelligent and capable, making them a greater asset to the help functions that use them. For example, help desks are finding that today’s chatbots are bringing them better cost and labor efficiencies.

It’s an interesting field, and we decided to take a look at how chatbots are progressing. What lies ahead for help desks?

Free download: Pros and cons of chatbots

The evolution of chatbots

One thing that surprises many people about chatbots is that they have their origins in a paper published by Alan Turing, the English mathematician and computer scientist. His 1950 paper, Computer Machinery and Intelligence, asked the question: can machines think? To this day, the “Turing Test” is used to analyze an intelligent program’s ability to mask itself as a human.

Since the 1950s, multiple chatbots were created that followed basic programming, but it wasn’t until 2001 that we saw the predecessor of Siri and other smart bots with apps such as Smarterchild. Over the last decade, Siri, Alexa, Cortana and others have risen to prominence and continued to advance in terms of capabilities. The technology behind them is found in the chatbots that power help desks, customer service and other key enterprise functions.

Fun fact: Chatbots have actually been around since the 1950s

Chatbots in help desks and IT support

Earlier iterations of chatbots for helpdesks began with "robotic process automation" (RPA), which automated simple tasks such as password resets. This was useful in terms of taking care of tasks that don’t require human input, but wasn’t effective for automating entire processes.

Intelligent automation was the logical next step because it introduces natural language processing (NLP). This allows chatbots to “understand” more complex user requests, such as those in unstructured formats. An example of this is when you simply type a request. There are many different ways that people might make a request - we all word things differently when we speak. NLP is about understanding the context of the request and has developed rapidly over the last decade.

NLP involves training the chatbot over time so that it becomes more intelligent when it comes to understanding context. The chatbot can then intelligently automate processes based on its understanding of requests.

This is the point at which many IT help desks are now sitting. Challenges with these types of chatbots include that they require training with the appropriate skills over time, which makes them difficult to deploy at scale. This is because it’s still a very manual process and has to be repeated with chat logs over time.

There is a third level of automation that will advance service desk chatbots to the next level, although this isn’t yet widely in-use. With NLU and the vast amounts of existing enterprise data, chatbots can be created that will automatically improve their capabilities and allow for deployment at scale. For example, consider all the data your company has from service desk tickets to CRM data - this is ripe for AI to learn and develop.

Conversational AI challenges

Conversational AI (artificial intelligence) is the term given to chatbots that can automate communication and enable long-running conversations with customers either via text or voice. Automation and AI combines to provide a personalized service to the customer.

While conversational AI is definitely improving over time, there still remains multiple challenges with it. For example:

• How people type text information. Everyone does it differently. Some people use full and correct sentences while others use short-hand. Some might misspell words which could confuse the AI. On the other hand, a human operator can often look at a misspelled word and understand what the person was trying to say.
  
• Use of regional terms or slang words. Even among native speakers of a language there can be vast regional differences in the terms that are used. For example, someone in the UK heading out to the pub might say they’re “off for a pint,” which is commonly understood to mean a beer. Without context, an American might say “a pint of what?” If a chatbot is likely to encounter regional terms, it can be difficult to train it with precision.
  
• The many different words we use (often regionally specific) to mean the same thing, or to mean something specific to the area you are in. For example, consider the many different contexts of the word “football.” In the UK and Europe, this is commonly understood to mean soccer, in the US it refers to American football and in New Zealand and Australia “the footy” is rugby or rugby league. AI tends to be black and white, rather than seeing the shades of grey that make up the human experience.
  
• Human emotion. Humans express their emotions in many different ways - sometimes it’s hard even for other humans to understand! This is where chatbots can miss vital context. For example, consider this statement if it were typed into a chat: “Oh sure, charge me AGAIN!” A human reading that sentence can most likely pick up on sarcasm, especially with being able to place context with any other parts of the chat. A chatbot though, is most likely to take that as consent to charge the customer.
  
• Misunderstanding or just not “getting” user intent. Most people are savvy to the fact that they’re probably chatting with a bot these days, but that doesn’t stop them from wanting and expecting the “human” experience. They don’t want messages back like “Sorry, I don’t understand” or “Are any of these options helpful?”, with 4 irrelevant links listed. Humans tend to expect quick, efficient service, whether that’s from a human operator or a bot, so in this sense, a bot can’t generally replace humans entirely.

Abbas Faiq, CIO of PTC has talked about how chatbots are not yet mature for help desks and that you can’t yet simply give over most queries to bots. He described three key lessons that his company learned from implementing chatbots:

“Through this entire process, we’ve learned a few things. First is that the IT service management space for intelligent chatbots is not very mature, so don’t expect robust solutions. In fact, vendors will claim that their product will automate 80 percent of your interactions; take that with not just a grain — but a bucket — of salt. We still have several years before products in this space mature.

Second, unless you have robust internal engineering and development resources, it’s generally not worthwhile to build your own solution. And finally, with assisted learning, there’s a significant time commitment required of internal subject matter experts within IT. We didn’t realize how much time it would take us every week to perform manual reviews during the initial testing and the pilot to teach the bot new skills.”

Download our “pros and cons of chatbots” here

Final thoughts

The summary of all of this is that while chatbots have evolved and are now able to provide valuable supplementary service to help desks, there is still work to be done in creating mature bots for the service space. You still need your human agents!

Here at JitBit we’ve built a very simple automated chatbot with “rule chaining”. This means that a workflow looks like this:

1. When XXX happens
2. And YYY conditions are met
3. Do these [actions]:
1. Action 1: aaa
2. Action 2: bbb
3. Action 3: trigger another rule which will, optionally, wait for its own trigger.

Our “automation engine” is one of our most heavily used features and chatbot automation has really added to it. We’re always interested to hear your thoughts on features too! You can submit ideas to our ideas forum please do so here.