The Classify Emails smart service lets you integrate your AI skill into your process. This smart service takes emails as its input, uses custom artificial intelligence models to analyze those emails, and then returns a prediction of the email type. The smart service also returns data on its level of confidence, based on thresholds you can configure.
The smart service takes emails in the form of EML files as input.
When called, this smart service uses the published model within the AI skill you select. As you train the AI skill, you create models and evaluate their ability to meet your requirements. When you determine a model is ready to use, you publish it. To check which model is published and will be called by the smart service, see the AI Skill design object.
The user who executes the process model doesn't need to have access to the AI skill being called by the Classify Emails smart service. However, without those permissions, some smart service configuration options are masked.
To properly view the results, you need to have permission to the email being analyzed.
The Setup tab allows you to select an AI skill.
| Name | Data Type | Description | Required | Multiple |
|---|---|---|---|---|
| Emails | Document | List of emails to classify | Y | Y |
| Confidence Threshold | Number (Integer) | Optional parameter that can be set as a number between 0 and 100. Default is 0. Predictions with confidence scores above this threshold appear in the Above Threshold output, while predictions with confidence scores that fall below this threshold appear in the Below Threshold output. More on confidence scores.When left blank, all predictions appear in the Above Threshold output. |
N | N |
The smart service returns three lists of emails, organized based on the confidence threshold you set. Each list contains the email and the email type. The skill's published model also outputs a confidence value for each email type prediction it makes.
| Output | Data Type | Multiple | Description |
|---|---|---|---|
| Above Threshold | Classification Result | Y | Returns documentId, classification, and confidence classified above threshold value when set |
| Below Threshold | Classification Result | Y | Returns documentId, classification, and confidence below threshold value when set |
| Failed | Classification Result | Y | Returns documentId, error code, and error details of the emails that don't get classified |
Activity chaining is not supported.
EML files must include Date and From data in the header, and each piece of information in the header should be on its own line. Subject and body data are used for model training.
Keep in mind the following requirements when using this smart service:
If any of these guardrails are limiting, please reach out to your Appian contact.
At run-time, the smart service calls a machine learning model that is published in the AI skill configured in the Setup tab.
Classification by the model is asynchronous and may take some time to complete.
Error code/message captured in the "Failed" output may appear for any of the following reasons:
Document corrupted or could not be converted may appear if any EML file doesn't contain Date and From data in the header.
If you see the following error, it is because the Classify Emails smart service is referring to an AI Skill design object that no longer exists. Reconfigure the smart service to use an existing AI Skill.
There is a problem with the task "Classify Emails" in the process "process_model_name". The underlying AI Skill object has been deleted or is otherwise invalid. Please confirm the node's configuration and try again.
