M42 Intelligence Similarities and Classification (formerly Effie AI Ticket, Virtual Coach) is a licensable, additional feature available in Efecte cloud installations helping ESM users to find relevant information and work more efficiently utilising existing data.

Currently, M42 Intelligence Similarities and Classification has three features that can be configured by the administrator:

• Similar content provides ESM users with a list of items that M42 Intelligence Similarities and Classification considers to be relevant for the open data card.
• Suggestions enable ESM users to fill a configured field of a reference type with a value suggested by M42 Intelligence Similarities and Classification. This can be used to classify data cards.
• Copy is used to fill data cards faster by copying multiple values on a data card that is being edited from a Similar content item.

M42 Intelligence Similarities and Classification needs to be separately configured for those templates where it is needed. A maximum of 5 templates can be defined. M42 Intelligence Similarities and Classification will be accessible in the user interface only when working on these configured data cards if the user has enabled it in their user preferences.

The configuration of the functionalities is made on the ESM admin side. To access the settings, please go to Administration → Maintenance → System settings → Virtual Coach settings.

First, the M42 Intelligence Similarities and Classification needs to be enabled. There is a status light that indicates whether the software components that are needed for M42 Intelligence Similarities and Classification are working or not. If the light is red, all components are not working properly, and you should contact the Efecte Service Desk.

M42 Intelligence Similarities and Classification components

M42 Intelligence Similarities and Classification feature is built on top of the ESM user interface but relies on a separate AI component. The AI component has two major functions:

Suggestion model

When M42 Intelligence Similarities and Classification has been configured, it will apply the attribute parameters to build a statistical model that can be used for classifying tickets. If there is something wrong with the model the categorization engine has built for a given template, there will be an error shown in the configuration view. If the error does not disappear within a few hours of configuration, building the model has failed, and you should contact the Efecte Service Desk. The suggestion model is always rebuilt when a configuration is being saved, so new data cards are not automatically considered in the model. Rebuilding the model can take up to 30 minutes, depending on the amount of data in the system. 

Search index 

The search for similar content uses an index that the M42 Intelligence Similarities and Classification builds using the configured attributes. When the user has a data card open, M42 Intelligence Similarities and Classification will analyze the fields based on the attributes that have been defined and provide results it deems relevant. The search index is updated every time a data card is saved.

Component troubleshooting

The status of the Suggestion model and the Search index is shown in the configuration UI. In case there are errors, follow these steps:

1. Make sure you have at least 5 data cards in the system with values in the Search and Suggestion attributes.

2. In production systems with thousands of data cards per configured template, the Suggestion model can take up to 30 minutes to be built. The search index is always built in a few minutes. If the error persists after 30 minutes and the configuration has been made with attributes that hold enough data, there is something wrong with the AI component, and you should contact the Efecte Service Desk.

Analyzing the environment

M42 Intelligence Similarities and Classification is best utilized when it is used for templates that have thousands of data cards already in the system. Consider the following use cases before configuring:

What existing data might be useful when users work on a data card? 

Use Similar content configuration to allow M42 Intelligence Similarities and Classification to look up relevant items, such as historical incidents, by analyzing certain fields (e.g., Subject and Description) to help find incidents that might contain relevant information for solving the incident at hand (such as resolution or related incidents). This information can be useful when solving repeating issues or when onboarding new employees.

Where is the most time lost when choosing from multiple possible values? 

Suggestions can help users save time with each data card they are working on. They allow the user to select the most fitting values quickly without scrolling through possible values in, for example, Categories - instead, probable values for the attribute will be suggested in the top right corner, and they can be easily selected there by clicking. Additionally, the result set will be filtered by the selection upon clicking. 

First, look into the existing data and see what kind of a data set is available with the desired suggestion attribute. For example, if Category is used, are there categories that are 1) not relevant anymore, 2) overlapping, or 3) over- or underrepresented? Values that are most common are, in general, more likely to be suggested by M42 Intelligence Similarities and Classification, but the actual result will always depend on the values of the search attributes on the open data card as well as the whole data set and configuration.

What existing data could be reused with new data cards?

Use Copy to reuse values from a similar content item. This way, the users can fill necessary information more quickly as they can fill up to three fields using existing data with a single click.

Data requirements

An important aspect of configuring M42 Intelligence Similarities and Classification is the type and quantity of data that is available. The results provided by the M42 Intelligence Similarities and Classification can only be as good as the data set allows it to be. Here are a few guidelines for the data that should be considered before configuring:

Use M42 Intelligence Similarities and Classification for:

• Environments that have a lot of data (at least 5000 data cards of each template where M42 Intelligence Similarities and Classification is desired to be used)
• Items with human-generated content and natural language
• Environments that use some sort of categorization or classification
• Data cards that have correct selections 
• Analysing items that typically have at least five words
• Analysing issues that are repeating in nature

Don't use M42 Intelligence Similarities and Classification for:

• Machine-generated items where the majority of content is generated automatically, such as automatic alerts
• Issues that are not recurring
• Analysing attributes that typically contain less than five words
• Data cards where attributes might have wrong values

M42 Intelligence Similarities and Classification will be available for use only for those templates that have been properly configured in the Virtual Coach settings.

Configuration

M42 Intelligence Similarities and Classification is configured with following drop-down selections, allowing only supported options to be selected. Mandatory selections have been marked as “Required” in the user interface, making the configuration straight-forward. M42 Intelligence Similarities and Classification configuration is done mainly via drop-down selections, which only allow supported selections for the configuration.

Template

First you need to define, which data cards M42 Intelligence Similarities and Classification will be used with. The data cards of the chosen template are analyzed, and M42 Intelligence Similarities and Classification will show content of these data cards when it is deemed relevant for the open data card. Results from a configured template can be shown for data cards of other templates as well. You can choose the template based on desired use case:

1. Show results of data cards from the selected template while working on data cards of that template
2. Show results of data cards from the selected template while working on data cards of other configured templates.

You can for example, configure "Ticket" and "Problem" data cards, so that each can be shown in the results, when there are data cards from either of those templates with similar content.

Search attributes

The search attributes define, which fields of the data cards in this template will be analyzed for determining the similarity. Two attributes are marked as required, and they should be string or text type attributes. Static and multi-value attributes are not supported. Optionally, you can also set an additional reference type attribute to be analyzed and taken into consideration of the search. The AI will read the text attributes and take only what it considers meaningful content into account. The M42 Intelligence Similarities and Classification is not an exact search engine by nature. Instead, it builds a search index from which it will search for the similar content when opened.

Suggestion attribute

The suggestion attribute is used as the attribute for the suggestion feature, which can be used for classifying tickets, such as Category. M42 Intelligence Similarities and Classification builds a statistical model using the Suggestion attribute along with the Search attributes to provide ranked suggestions for the value of this classifying attribute. The suggestion attribute needs to be a reference-type attribute. Multi-value attributes are not supported. Additionally, the suggestion attribute result is used as a filtering attribute for similar content for the end user when the user selects a result in the M42 Intelligence Similarities and Classification view.

Copy attributes

M42 Intelligence Similarities and Classification allows the end user to reuse existing content from similar content data cards of the same template by copying. It is possible to choose up to three different attributes to be used for copying values to the open data card. Currently, supported attribute types are text, string, static string, reference, and external reference. Rich-text content can also be copied and previewed in a separate preview window, before copying.

Data set size

Use the Data set size to provide M42 Intelligence with a relative size of the available data set. M42 Intelligence uses a ranking algorithm to determine, which results are the most similar to the open data card. The similarity score shown to the end user is always a relative number derived from the available data. Therefore, M42 Intelligence Similarities and Classification uses a scaling factor, which is used to scale the results of Similar content between 0 and 100, depending on the available data. The scaling factor can be adjusted with the Data set size setting, which is set to 40 by default. This will affect how the score for the similar content is calculated. Adjust the data set size, if you feel that results tend to be always too low or too high. Before altering the Data set size, consider the following best practices:

• The scaling factor value is set to 40 by default. This is suitable for templates with around 10,000 - 60,000 data cards.
• A value of 4 is recommended for templates with 100-200 data cards.
• A value of 10 is recommended for templates with around 1000 data cards.
• Values over 40 are suitable for bigger production environments.
• Only values between 0 and 100 are allowed.

Adjusting the data set size does not affect which results are shown, but only the relevancy score of the data cards shown in Similar content.

Depending on the system, the startup and learning phase will take somewhere between 10-30 minutes. This time is needed for the M42 Intelligence to analyse the data and create the service.

After this the users will be able to enable the M42 Intelligence Similarities and Classification for themselves. To test, please refresh the screen and go to your user profile.

See the user guide for details on how to use M42 Intelligence Similarities and Classification.