NOW Intelligence & Virtual Agent – Better Together

[darius_koohmare]

In our lives, there are many examples where the sum is greater than the parts. From automobiles to orchestras, there are countless scenarios where bringing together distinct elements results in a wonderful end result whose utility exceeds that of any individual component. NOW Intelligence and Virtual Agent are no different. So what are the parts that complement each other? Let’s walk through an example scenario to see how every part of a traditional ticket lifecycle, from creation to resolution, gets enhanced via NOW Intelligence & VA.

Part I: Virtual Agent with NLU

A user in your organization is experiencing a need – something is broken and they need it fixed! Fortunately for them, they don’t need to fumble around a desktop experience looking for getting help – there is a virtual agent available that can get them the help or information they need, from a single consolidated place.

The virtual agent can take advantage of NLU for two key employee experience improving use cases in particular. First, it can use intent identification to map the users input phrase (utterance) directly to the proper conversation topic. Second, using entity extraction, the bot can take context from what the user already told it so that it does not need to re-ask for information it already received.

Similarity within the virtual agent could even identify similar incidents opened by the same user, preventing duplicate tickets for the same issue.

The virtual agent will attempt to provide the user with information to self-serve, but in this case the user will indicate that the results are not helpful.

 

As a result, the virtual agent created the incident on behalf of the user, with information it collected during the conversation. All this occurred without requiring a call center resource logging the information or deciphering it from an unstructured email!

Part II: Predictive Intelligence: Classification 

So what happens after the incident gets created? Predictive Intelligence (classification) can be used to route the incident to the right assignment group, and enrich it with information like category, or potentially business service! While we are looking at the perspective of the virtual agent, keep in mind the classification runs on the incident regardless of how it was created, meaning it works for portal record producers or emails too! The predictive intelligence feature for classification was available as of Kingston, and was built using supervised machine learning. Without leaving ServiceNow data centers, data from your instance is sent to a training server in our cloud. That server provides a model back to the instance, which it can call through business rules to automate activities like setting assignment groups, categories, or business services. Incidents that employees create through the virtual agent can be enriched and automatically routed to the right fulfillment group through agent intelligence. Not only does this help speed up time to resolution, it also helps reduce the amount of information the user needs to enter when chatting with a virtual agent.

Part III: Predictive Intelligence: Similarity

With the incident routed to the proper assignment group, advanced work assignment will push it into a technicians inbox in Agent Workspace. This is where the next component shines in delivering the employee with faster service: Predictive Intelligence (similarity). In Madrid, the agent intelligence framework stepped into an unsupervised machine learning use case of clustering. Clustering enables a different set of use cases than the supervised classification, specifically around enabling users to see relevant content and groupings. The similarity feature will provide ‘agent assist’ capabilities that provide relevant resolution recommendations to the agent, such as similarly solved incidents and knowledge articles. In addition, it proactively identifies trends such as a major incident forming and recommends the agent to promote it.

 

For the virtual agent, it can also take advantage of similarity anytime a user tries to report an issue by comparing it to known problems or existing incidents the user has entered. Instead of allowing them to create duplicates, it can offer them to add comments instead!

Summary:

NLU's intent and entity extraction gets users into the right conversation topic, answering the minimal number of nodes (questions). Similarity and classification work to help identify existing issues, and route newly created tickets to the right categories, services, and teams. Once routed to teams, the fulfillers can utilize the similarity via recommended resolution through knowledge articles, catalog items, or even major incidents that they may not have been aware of. This means faster and self service driven incident and request management.

Bonus: Surveys & Performance Analytics

Before usage begins, you need to collect actual utterances from employees around how they would phrase specific intents. While data can be extracted from existing incidents, its common practice to explicitly ask users for examples as well. You can use the survey functionality to build a survey with text questions asking users to provide their own phrases for specific intents, which you can then assign to a subset of users.

After usage begins, remember that you can take advantage of reporting with performance analytics to pull insights into the virtual agents performance and top opportunities. Check out this blog on PA & VA for all the details!