Agent Intelligence from ServiceNow - neural networks in the service of technical support
According to a ServiceNow survey , 89% of IT managers use or implement machine learning technologies in their organizations. Of these, 87% say that automating business processes brings great benefits to businesses: it saves time, money and human resources.
The introduction of machine learning requires major changes in the work of the organization. However, according to the same survey, only 48% of respondents decided to make these changes. In addition to organizational changes, the use of machine learning requires the involvement of specialists — in particular, data analysts. And the demand for them, according to IBM research , far exceeds supply. According to forecasts, by 2020 the situation will become even worse. Let's tell you what the solution from ServiceNow can help in this situation.
/ Photos / Chris Isherwood / CC
')
One solution is offered by ServiceNow. The new product of Agent Intelligence is built into the platform itself, so it does not require additional implementation costs and is available without the involvement of a staff of data mining specialists. According to the developers, Agent Intelligence automates the classification, prioritization and assignment of tasks. This reduces incident resolution time, reduces errors and increases customer satisfaction.
Agent Intelligence uses machine learning to create user interaction models from specified templates. These models are based on saved data from ServiceNow customers. The data is used to predict the values of the ticket field. Pattern-based learning and text analysis in the ticket description allows the system to automatically determine the category, priority, and appeal destination group.
Agent Intelligence models can be enhanced — add other data: segments, customer severity and preferences, or frequent incident solutions. After debugging the template, you need to set the threshold values of the classifiers, test the model and start it up.
In the process, the model uses text in the ticket description field and other data to suggest the category, priority, and assignment group to the first-line technical support specialist.
Each ticket, the parameters of which the model determined correctly, is considered successful. Changing parameters manually is fixed as an error. These errors are used in the planned periodic retraining of the model. Re-learning intervals are set by the user.
Accelerate categorization and prioritization. The ServiceNow study showed that 40% of support service specialists are forced to classify the incident into one of 100 possible groups, and 25% of specialists into one of 300 groups. Finding the right group may take too much time from a person.
According to Accenture Consulting's report , technical support specialists spend 12% of their time to assign a category and priority to a request.
Agent Intelligence instantly analyzes and classifies queries. This helps to save people from manually sorting and redirecting requests. Professionals no longer need to scroll through several lists or search for which category or group to assign to the request.
Reduced incident resolution time and number of errors. Automatic routing draws the attention of a specialist to the tasks and incidents that should be addressed first. This increases the overall level of customer satisfaction.
Agent Intelligence classifies and routes requests faster and with fewer errors. This increases the productivity of support staff.
Toolbar and Reporting. The accuracy of categorization and routing can be monitored in real time using the toolbar. The system provides feedback and monitors ongoing activities to continuously improve the quality of service.
The launch of the pilot version of the program showed that users can expect to save 8% of their working time or 26 thousand hours per year due to improved classification, prioritization and appointment of incidents.
Experts estimate that Agent Intelligence saves $ 100,000 in 3 years of use and pays off in 10 months.
The first ServiceNow Agent Intelligence machine learning solution will be available with the next version of the Kingston platform.
Read more about the product here .
Reviews of users of the pilot version are looking here .
The introduction of machine learning requires major changes in the work of the organization. However, according to the same survey, only 48% of respondents decided to make these changes. In addition to organizational changes, the use of machine learning requires the involvement of specialists — in particular, data analysts. And the demand for them, according to IBM research , far exceeds supply. According to forecasts, by 2020 the situation will become even worse. Let's tell you what the solution from ServiceNow can help in this situation.
/ Photos / Chris Isherwood / CC
')
How to solve a problem
One solution is offered by ServiceNow. The new product of Agent Intelligence is built into the platform itself, so it does not require additional implementation costs and is available without the involvement of a staff of data mining specialists. According to the developers, Agent Intelligence automates the classification, prioritization and assignment of tasks. This reduces incident resolution time, reduces errors and increases customer satisfaction.
How it works
Agent Intelligence uses machine learning to create user interaction models from specified templates. These models are based on saved data from ServiceNow customers. The data is used to predict the values of the ticket field. Pattern-based learning and text analysis in the ticket description allows the system to automatically determine the category, priority, and appeal destination group.
Agent Intelligence models can be enhanced — add other data: segments, customer severity and preferences, or frequent incident solutions. After debugging the template, you need to set the threshold values of the classifiers, test the model and start it up.
In the process, the model uses text in the ticket description field and other data to suggest the category, priority, and assignment group to the first-line technical support specialist.
Each ticket, the parameters of which the model determined correctly, is considered successful. Changing parameters manually is fixed as an error. These errors are used in the planned periodic retraining of the model. Re-learning intervals are set by the user.
Agent Intelligence Features
Accelerate categorization and prioritization. The ServiceNow study showed that 40% of support service specialists are forced to classify the incident into one of 100 possible groups, and 25% of specialists into one of 300 groups. Finding the right group may take too much time from a person.
According to Accenture Consulting's report , technical support specialists spend 12% of their time to assign a category and priority to a request.
Agent Intelligence instantly analyzes and classifies queries. This helps to save people from manually sorting and redirecting requests. Professionals no longer need to scroll through several lists or search for which category or group to assign to the request.
Reduced incident resolution time and number of errors. Automatic routing draws the attention of a specialist to the tasks and incidents that should be addressed first. This increases the overall level of customer satisfaction.
Agent Intelligence classifies and routes requests faster and with fewer errors. This increases the productivity of support staff.
Toolbar and Reporting. The accuracy of categorization and routing can be monitored in real time using the toolbar. The system provides feedback and monitors ongoing activities to continuously improve the quality of service.
What is the benefit
The launch of the pilot version of the program showed that users can expect to save 8% of their working time or 26 thousand hours per year due to improved classification, prioritization and appointment of incidents.
Experts estimate that Agent Intelligence saves $ 100,000 in 3 years of use and pays off in 10 months.
When released
The first ServiceNow Agent Intelligence machine learning solution will be available with the next version of the Kingston platform.
Read more about the product here .
Reviews of users of the pilot version are looking here .
A couple of related links
Source: https://habr.com/ru/post/340682/
All Articles