Over the next few days we will be publishing a series of short articles written by our VP of Products, Peter Schneider. This is Part 3 of the series, we will be adding links to the bottom of the page as the other parts are released.
Self-Service Powered by AI
For a IT Service Desk Managers, the preferred incidents are those that can be solved even before they are sent to the service desk. Already today, issues can be solved by self-service through providing known solutions in frequently answered question sections. Powerful search capabilities in the knowledge base of the self-service portal reduce already today the incident volume. When user type the subject of the incident, a live search feature can suggest already today known solutions to similar topics. Machine-learning will help to make these suggestions even more accurate by utilizing natural language processing capabilities, scoring, and learning mechanism which go significantly beyond indexed keyword search used commonly today.
Artificial Intelligence may also be used in the future to assist self-service portal users in completing typical transactions in another form of Robotic Process Automation. When it comes down to approving service requests, artificial intelligence might review pending approvals based on historic behavior. The “self-service assistant” might suggest to approve or reject pending approvals based on past decisions reducing the workload to review the details of approvals themselves. Renewal of access rights, software licenses, or other things provided “as a Service” might be automated using artificial intelligence to assist the self-service portal user.
Natural Language Processing for Conversations
Through natural language processing ITSM tools will be able to “pretend” to be a virtual service agent in a conversation. Due to the omni-channel nature of ITSM tools communicating to users through various channels such as email and chat, natural language processing will need to be flexible.
The following use cases are likely to provide a satisfactory user experience in the first implementations due to the predictable flow of the conversation:
- service requests based on a structured service catalogue
- status inquiry on existing incidents, change and service requests
- providing known error information
It is unlikely that chat or email bots will provide a superior user experience for incident management over self-service portals or chat with human agents in the near future. Chat or email bots will serve initially only for incident reporting, but not for true issue resolution purposes due to the lack of analytical intelligence in machine learning algorithm. Hence, the benefits of virtual service agents will be similar to the benefits of a self-service portal.
Bots are able to hold a meaningful conversation as representatives of the IT Service Desk after few months of training. For some highly structured tasks, especially when dealing with a well-designed service catalogues, bots might be able to cover up to 70% of routine tasks similar to the performance of bots in use today in B2C environments.
Bots can be trained to deal with routine requests when they are well integrated to the ITSM tool. Information stored in the ITSM tool such as CMDB data, open issues, and knowledge articles are crucial to provide a similar user experience by bots compared to human agents, who have access to all information in the ITSM tool.
Today’s challenges of bots are based on limitations of natural language processing algorithm that can perform basic tasks such as recognizing the intent of a user even if s/he uses synonyms, plural instead of singular, or words instead of digits. However, dealing with multiple languages at the same time, serving customers for different enterprise services such as IT and HR at the same time, and identifying related issues and their solutions with high confidence will remain a challenge for years to come.
Besides limitations in natural language processing, user acceptance of virtual service agents will dictate the speed of growth of bot-based conversations. Recognizing semantics in the conversations and when it is time to hand over from virtual service agent to a human agent will play an important role in the acceptance of virtual service agents.
We will also be hosting a webinar on Tuesday the 27th of November, where Peter will be discussing the topic of AI with one of our key partners, ultimate.ai. You can find out more about the webinar and register from the link below.