A help desk, in layman terms, is simply a central hub (nowadays accessible online by default) where customers can deal with service-related issues. Depending on the nature of the service, options may range from basic tips to technical instructions.
As such, it is easy to see how help desk chatbots may be employed to efficiently automate almost all help desk functions available for businesses today. Not all, of course. But it is enough that potential customers can still come out pretty satisfied with the resolved issue no matter how technical or advanced it becomes.
So if chatbots can be that efficient for almost all help desk functions, how do we utilize them efficiently? Short answer: method of information delivery. This is the main theme that we shall focus on for this overview guide.
Main Types of Help Desk Chatbots
Long story short, help desk chatbots are essentially virtual assistants. The design philosophy might differ depending on the business or organization using it, but the method of information delivery through interactive natural language is all the same. For our exclusive definition, there will be at least four major types available:
(DISCLAIMER: these “types” are mainly to be considered only as feature formats. Nothing is restricting your chatbot workflow from incorporating elements of each following type.)
1. First-line Query “FAQ” Bot
- What it is: One of the most commonly noted chatbot features. Simply put, this is the type of chatbot that is designed around answering direct queries. If a question is directly related to a frequently asked question, the user is immediately redirected to the related answer.
- What it does: If the question is exactly one of the FAQs listed, it would simply provide a direct answer as a response. If there might be a few related queries, the workflow could introduce a clarification step. This is most simply done by just listing any of the possibly related FAQs on a list for the user to choose from.
- General expectations: While there is no sophisticated workflow design needed for each FAQ response, the chatbot does need to recognize the questions properly. Setting up the correct AI option(s) will be a necessary step for this. Also, additional interactions within the same query type would be a nice plus (for engagement), but ultimately not necessary if you just want the chatbot to resolve such simple queries directly.
2. Targeted Query “Q&A” Bot
- What it is: In the case of questions clearly outside of any related FAQ, the targeted query chatbot can engage and handle the issue, preferably all the way to its resolution. This is arguably a more fundamental feature of help desk chatbots as virtual assistants. The user provides the query on-demand, then the chatbot attempts to solve the issue per related case.
- What it does: The chatbot analyzes the sentence through a net of pre-qualifying requests, before the workflow decides which set of information would be most relevant to the supposed answer. Do take note that the responses do not always have to be dynamic. Static responses still fall under the targeted query category, so long as it is a question not asked too often.
- General expectations: Usually, the first query is the only one type-written by the user. The next set of qualifying requests is then built as multiple-choice questions. If the question gets complex/specific enough, the chatbot should know when to quit the analysis immediately and ask for human assistance.
3. Instructional Request Bot
- What it is: This is where questions that have a set number of step-by-step answers are categorized. Engagement is far more important here than the previous two, since the step-by-step process requires constant presence. It doesn’t have to be fancy, or humanly convincing. Just a confirmation that each possible additional question may still be entertained separately.
- What it does: It should preferably start with a standard targeted query. If resolving the issue requires executing more than a few steps, the next sequence will run like an FAQ response. Each step of the way may fork into several other paths, depending on how the instruction was executed, or depending on its result.
- General expectations: Empathy and quirkiness go a long way in maintaining the user’s attention span through the whole problem-solving process. Relatively “stupid” confirmation questions (e.g. “did you make sure to plug the modem back in?”) are okay to include, for the most part. But it is somewhat recommended not to stack them together consecutively.
And of course, each presented mode will (should) always include some form of activity or data collection from the user. This is both for basic performance analytics, and to simply survey the general activity of the chatbot itself.
Setting Up Your Help Desk Chatbot
“Set up the help desk in the ideal way that you might want or need to use one.” This…may sound very cliché and frustratingly generic. But that is basically the core idea, to visualize yourself using the chatbot’s customer service capabilities. Barring specific technical tips in designing workflow and optimizing interactions, a few general guidelines to help you start in the spirit of such basic concept are:
- Choose a virtual assistant “mode” based on the nature of your business. In other words, decide the type of support desk you would use the most. For example, do you want to build a self-service knowledge base? Or do you want to stack customer issues as they are reported? It can be both, of course. But you may want to focus initially on one as the chatbot is first deployed.
- Determine the space between your chatbots and human agents. How long before human intervention would be ideally required? Would your help desk even need human agents? What if a process transfer is required pronto? Average it out, and see how many replies it would take before any fallback procedure becomes strictly required.
- Clarify objectives and priorities to reduce chatbot interaction duration/amount. Like an actual human customer agent interaction, the more precisely the conversation ends (with a resolution), the better. True, most of the responsibility falls under the user (who asks the query). But, it is still the chatbot’s job to set their priorities straight and know what they really want, by using the function of directed questions.
- Constantly Tweak Your Knowledge Base (don’t just expand it). Adding more data to your (self-searchable) knowledge base is one thing. But tweaking your chatbot to access such info more smartly is another. Sounds too vague and esoteric of a tip? Try segregating qualifying question categories in simpler terms. For example, instead of “PC peripherals,” try using “PC input devices” and “PC output devices” instead.
- Analytics isn’t just for trends; it is a “trimmer” tool as well! Analytics typically help us determine recurring trends, and help us learn additional things that can aid your help desk chatbot. That being said, in the exact same vein, it is also possible to strategically eliminate unwanted/inefficient workflow portions, so that only the most useful (and therefore less potentially confusing) procedures remain.
Important Keys to Note
As more and more businesses default to using chatbots for just about anything, the meta of delivering fast, always available information to any customer service experience would more and more become the norm. But to finish things off, you also need to consider these additional things in mind:
What communication channel do your clients/customers mostly use? Needless to say, people typically favor channels that they regularly use to contact their friends and family. So while some simply default to the fairly popular and universal Facebook, it wouldn’t hurt to see if a certain channel would be used often enough to consider expanding towards it. (LINE, for example, is significantly more popular for East Asian users)
Another thing to consider is how your chatbot handles context. We already briefly hinted at this with our introduction of the “targeted query” chatbot. But generally, you need to have workflow editing options that can efficiently make use of AI-based natural language processing (NLP) features. Don’t go overboard, though. If a self-service knowledge base and FAQ are enough, then it’s better to trim everything down for more efficient management and troubleshooting.
And lastly, it helps to remember that all design principles for help desk chatbots discussed here, should work just as well when used internally (for back-office staff support) as externally.
