AI

The AI (Artificial Intelligence) actions allow you to analyze and generate data within processes using AI. For example, you can use AI to assign incoming emails to departments, extract data from documents, or generate text.
To meet our high data protection standards, the AI actions are hosted on linqi servers. These AI actions are executed in the background, so it may take a few seconds for them to complete.

info

To work with AI actions, you must enable the corresponding feature.

Content created or calculated using AI may contain errors or inaccuracies.

If you are running linqi on your own servers, you will need the corresponding AI servers. Please contact us for more details.

AI Classification

This action can be used to classify a text. This means, for example, you can assign an incoming email to a department (complaint, invoice, general inquiry). You can also use it to analyze a document to determine whether it is an invoice, a contract, or something similar.
The following configuration options are available:

• Text to classify:
  Specify the text that the AI should classify. Using placeholders, you can reference the text of an incoming email or the result of an AI text recognition.
• Context:
  The context helps the AI better understand the text and can also be used to provide more specific instructions. For example, you can tell the AI that it’s a customer email or specify that certain things should be ignored.
• Options:
  Add the possible options, i.e., the possible types the text can be classified as. If you want to determine whether an email is a complaint, an invoice, or a general inquiry, enter these three options.

This action provides the following placeholders:

• Selected option:
  This placeholder returns the option selected by the AI. Using the above example: complaint, invoice, or general inquiry.

AI Translation

This action allows you to translate a text into a target language. The AI automatically detects the language and only translates it if the original language differs from the target language.
This ensures that texts are always available in a language your employees understand, helping to reduce barriers when working with international team members.
The following configuration options are available:

• Text to translate:
  Define the text that should be translated. You can use placeholders for dynamic content.
• Context:
  The context helps the AI better understand the text and can be used to provide more specific instructions. This helps the AI use the right wording in the translation.
• Target language:
  Select the language into which the text should be translated.

This action provides the following placeholders:

• Translated text:
  This placeholder provides the translated text. If the text was already in the target language, the original text is returned.

AI Summarization

This action allows you to summarize text, for example, to forward only the most important points to a specific group of people. You can summarize documents or longer justifications provided by a person.
The following configuration options are available:

• Text to summarize:
  Define the text that should be summarized. This can be a form value such as a justification for an investment.
• Context:
  The context helps the AI better understand the text and can also be used to give more specific instructions. This helps ensure the AI includes the right points in the summary.
• Maximum word count:
  Specify the maximum number of words the AI should use. The AI may use fewer words if the text is short enough.

This action provides the following placeholders:

• Summary:
  This placeholder is replaced with the summarized text.

AI Data Extraction

This action helps you extract specific values from a text. For example, you can extract an invoice number, contract duration, revenue, or target system from a text without manual intervention.
The following configuration options are available:

• Text to extract from:
  Specify the text from which the data should be extracted. This could be the result of an AI text recognition of an invoice or contract.
• Context:
  The context helps the AI better understand the text and can also be used to provide more specific instructions. This can help the AI find the right data points. In the simplest case, you can tell the AI that it’s an invoice.
• Values to extract:
  Define which values should be extracted. Use Add value to add a new one. For each value, you must specify:
  • Name:
    The name describes which value should be extracted (e.g., invoice number or year of incorporation)
  • Data type:
    Select the type of data to be extracted. This helps the AI find the correct value and allows for better formatting later.

This action provides the following placeholders:

• Extracted values:
  Each defined value is available as a placeholder. Depending on the data type, different formatting options are available.

AI Prompt

This action allows you to send general prompts (tasks) to an AI. You can use this to generate text or make other decisions.
The following configuration options are available:

• Prompt:
  Enter the prompt, i.e., the task. This can be simple, such as Write a reply to the following email: placeholder for email text, or more complex for larger tasks.
• Context:
  Use the context to refine the task. For example, you can define the tone of the email or provide data that should be searched.

This action provides the following placeholders:

• Result:
  This placeholder provides the AI’s response. You can use the context to further define how this response should look.

AI Text Recognition (OCR)

This action allows you to extract text from documents. For example, you can extract text from scanned invoices or contracts and then pass it on to another AI action.
This action can extract text from PDF files, image files, Word documents, or text files.
The following configuration options are available:

• File:
  Use a placeholder to specify the file from which the text should be extracted. This could be a file loaded from SharePoint, an email attachment, or a file uploaded via a form.
  Note: You can only use one file at a time. If a placeholder provides multiple files, you must process them using a loop.

This action provides the following placeholders:

• Extracted text:
  This placeholder returns the text extracted from the file.
• Extracted text with position data:
  This placeholder returns the extracted text along with position data as JSON. This is especially recommended if you use this action as input for other AI actions (like AI Data Extraction), as the positional data provides the AI with more context about the document.
  This placeholder is only populated for PDF and image files.