How can chatbots improve service desk resolution times?

Chatbots can be ideal flag bearers for self-service initiatives. In my experience, I have used it to significantly reduce service resolutions times while leading the initiative for my team. Tasks such as furnishing "How to articles", FAQs, troubleshooting steps and subsequently requesting human assistance are pertinents ways of deriving value out of chatbots. In my experience, data from monitoring success and failure of questions with NPS scores were used to edit utterances files which in turn integrate with NLP and AI to respond to queries. The Chatbots can be made accessible via frequently visited webpages, and mobile apps for B2C services to provide self-service ability to end customer - without the need to open a tickets and wait.

In the past, most chatbots were scripted conversations that followed a pre-defined path. This limited the potential of chatbots to whatever the builder of the chatbot intended. Now, with the rise of Generative AI, a new generation of chatbots is being born that allow for much more flexibility, which makes chatbot conversations feel much more natural. This in turn increases the value business can get out of chatbots.

In my experience, many customers indicate that they actually prefer to use a chatbot if that chatbot can solve their problem. They indicate it's faster - no waiting times, it's easier - by using buttons, because the chatbot already has some data about you, and it's more anonymous - you don't have to be nice to a chatbot.

Information on simple and repetitive tasks such restarting a management service, documentation on network connectivity, how to articles etc can be seamlessly provided which obviates the need to open a support ticket. This enables service desk agents to focus on more value driven issues vs spending time repetitive tickets. With faste and more convenient access to information and solutions this will also give a boost to NPS scores as well.

My advice for companies getting started with chatbots is always: don't try to build the perfect chatbot straight away. Start by analyzing the questions your customers are asking, and see if there's any recurring questions that are easy to automate. Once your bot can answer those easy questions, publish it, and send all other questions directly to your service desk. This first go live, even if it only handles a few percentages of all incoming requests, will give you a lot of incredibly valuable hands-on experience which will decrease the typical risks of rolling out chatbots.

Chatbots need to be supervised by a human. People are often using chatbots because they can't find something or they are having an issue. There is nothing worse than having a complex problem that a chat bot cannot understand. Technology is great for solutions, but people who have limited technical ability need empathy and support from other humans.

One of the challenges is to derive the effectiveness of Chatbot if your team supports complex tasks such as configuration, troubleshooting networking issues such as packets traces, and maintaining uptime of critical systems. Though relevant information is available as a self-service, end customers would prefer a human involvement with empathy in case of time sensitive tasks or outages. Sifting through self-service guides furnished by Chatbots might not effective in such scenarios.

In my experience, following are the ways I have gone about implementing Chatbot. 1. Analyze the monthly tickets to identify areas where customers are frequently opening cases to service teams. 2. Delineate the categories or nature of requests. 2. Identify the repetitive tasks and nature of resolutions provided. Based on garnered insights, implement and roll out the availability of the resources such as Documentation links, FAQ's, blog articles etc. Have periodic scrum meetings with Chatbot developers and stakeholders to iterate on the feedback and implement improvements. (Note: getting a buy in from leadership and stakeholders to treat this as a marathon instead of a sprint can do wonders to success of the initiative).

Data preparation and embedding is step #1 to establishing a chatbot for the service desk. While training was mentioned to create the experiential output desired by the end user, without context the information is generic at best, worthless at worst. To overcome this, preparing data (knowledge base, historical tickets, conversations) and embedding this information for the AI using a vector index will empower the chatbot with the context necessary to deliver thoughtful, relevant, and appropriate content based on real interactions and knowledge. Failure to deliver on this portion will result in an ineffective implementation of AI.

Chatbots can improve service desk resolution times by providing instant responses to common queries, guiding users through issue troubleshooting, and freeing up human agents to focus on more complex problems.