Agent Intelligence (AI) is new module available from London and next versions of ServiceNow. It is used to predict and set the fields like assignment group, category and so on based on short description during the record creation. It is based on machine learning algorithms which are used to train the predictive models from the previous records or data available in the instance.
We can create templates as per the requirement to get the solutions which are available under Solution Definitions. In this article we are going to elaborate the Incident Management process.
Incident Management is based on Incident table and its records. Basically, we have two solution definitions – one being prediction of assignment group and the other being category. So here the assignment group and category are the output fields and their predictions are based on input field short description. We need to specify the training frequency so that the solution works for that period and updating this frequency will add new records to the available data.
This process also supports the domain separation. As per the given conditions it fetches the records which are useful to train the predictive models. The input and output fields should never be empty in order to get the results of Agent Intelligence.
ServiceNow also provides various processing languages such as Dutch, French, German, Japanese and Spanish. If needed, we can see these predictions in the above languages along with English.
Training predictive models
After activating the model and fetching records we can start the training if it is a default template. If it is a new solution, we can submit the created template and then start the process of training. It is to be noted that only active templates can be trained. There should be a minimum 30,000 records to get effective solutions, if it is more than the minimum records it takes the most recent records.
In the process of training we can see them under ML Solutions with different training states, such as training is cancelled, error while training solutions and solution complete. Training process takes time according to the number of records i.e., available data used to train the new prediction. Thus, the training state and progress can be viewed under ML Solutions. It will show solution complete state only if the number of records is sufficient with the given conditions and the input, output fields aren’t empty in the available records.
If we want to retrain the predictive model by updating the conditions, we can go for an update and retrain these results in new version of the solution.
The results of Agent Intelligence are based on these solutions which runs using the Business Rules provided by default in ServiceNow.
This is the best way to check the accuracy of expected results. This can be done using REST API Explorer by calling in the Agent Intelligence API. We need to give correct solution template name and add a query parameter. We can test and check the results – if it is the same as expected, we need to create a test incident and check the result for accuracy.
Improvisation for effective solutions
If there are unused categories and assignment groups, they need to be removed using filters. Go to Solution Statistics and check your solution with the version of the solution you want to use. We can thus check precision as well as coverage of the solutions.
Moving to Production
Make sure production (prod), testing (test) and development (dev) instances are cloned so that the solution works in all the instances equally. If we want to move this to any of the instance, we need to capture the solution to update set by using add solutions to current updates set in the related links and move it to test and prod.
Agent Intelligence can also be implemented in Customer Service Management and HR Service Delivery. Agent Intelligence helps in reducing the resolution time of tasks and error rates while assigning groups and categories. It also reduces manual work of agents so that they can work on other difficult issues that results in higher customer satisfaction and increasing resolution rate.
It is not only useful for the Users and Service desk team but also helps the Developers in reducing the time to write script which results in increased performance of the tool. This is highly useful if the flow of incoming requests is high.
Contact us for further details