Dialog Insight uses AI in certain features to support users in their marketing activities. AI can help with content creation, campaign preparation, contact segmentation, information retrieval, and certain types of analysis. AI does not make decisions autonomously on behalf of users. The generated results are suggestions that must be reviewed and validated before use.
This article explains how to add context for AI. It also presents the features concerned and describes the data they use, as well as, where applicable, the information transmitted to a third-party artificial intelligence service provider.
To learn more about the principles and commitments governing Dialog Insight’s use of AI, consult the refer to the Responsible Artificial Intelligence Principles, which is the reference document regarding responsible use of AI .
This article describes in greater detail the data processed for each AI feature.
Adding AI Context
The DiA assistant supports several artificial intelligence features in Dialog Insight, including campaign, message, and group generation. To improve results and avoid having to repeat the same information, you can create one or more DiA contexts and select the appropriate one when using a compatible feature.
→ Learn more about adding contexts
When you use an AI feature, the instructions contained in the selected DiA context are transmitted to the AI.
Data accessible to AI
AI can access the structure of the project’s fields, including their names, types, and descriptions. This allows AI to suggest results that take the available fields into account without accessing the values stored in contact records or relational tables.
For example, the group generation feature can determine that a Country field exists and suggest a condition based on that field. However, AI does not access the country associated with each contact. Dialog Insight’s Groups engine then applies the generated conditions to determine which contacts belong to the group.
By default, AI does not have access to the values contained in contact or relational table fields. However, certain features may require access to the values of specific fields in order to provide a result that takes project-specific data into account. In such cases, access must be explicitly authorized in the configuration of each relevant field by enabling the corresponding option.
This authorization allows only compatible AI features to use a field's value when necessary for their operation. Enabling this option does not, therefore, mean that all AI features automatically have access to the field.
| Warning! Only authorize access to data that is necessary for the feature to operate. Do not authorize access to sensitive, medical, or financial information, personal identifiers, access codes, or other confidential data. |
Access to values can improve the relevance of certain features when they need to take project-specific data into account. However, each feature uses data differently. Refer to the Features that use AI section to learn which information is used in each case.
Features that Use AI
To provide context to certain features, Dialog Insight transmits the necessary information to a third-party artificial intelligence service provider. This information is processed by a large language model (LLM) to generate the requested result.
The information transmitted varies by feature and is described in the tables below. By default, no values from contact records or relational tables are transmitted to the third-party provider. Access to these values, when handled by a feature, requires explicit authorization in the configuration of the relevant fields. When a DiA context is selected, the instructions it contains are also transmitted to the third-party AI provider.
Writing and Campaign Creation Assistance
AI can generate, correct, or translate content based on instructions provided by the user. For example, it can produce a first draft of a message based on the objective, target audience, desired tone, and elements to include.
| Feature | Data processing |
|---|---|
| Email Content Generation | The selected DiA context, the brief and structural information from the generated campaign, the communication types configured in the project, and the instructions provided during the conversation are transmitted to a third-party AI provider. No contact-related values are transmitted to a third party. |
| Email Subject Line Generation | The selected DiA context, the instructions entered, and, if the corresponding option is enabled, a summary of the email content are transmitted to a third-party AI provider. No contact-related values are transmitted to a third party. |
| SMS Content Generation (coming soon) | |
| RCS Content Generation (coming soon) | |
| Assisted Translation | The selected DiA context, the instructions entered, and the email content in the version to be translated are transmitted to a third-party AI provider. |
| Campaign Generation | The selected DiA context, the instructions entered during the conversation, and the structural and process information required to suggest a campaign—such as the available fields, relational tables, acquisition processes, and forms—are transmitted to a third-party AI provider. No contact-related values are transmitted to a third party. |
Natural Language Segmentation
Certain features allow users to describe a need in natural language. AI then interprets the request and suggests, for example, segmentation criteria or a scenario. For segmentation, the suggested criteria remain visible and editable. The user must verify the fields, operators, time periods, exclusions, and number of targeted contacts before using the group.
| Feature | Data processing |
|---|---|
| Group Generation | The selected DiA context and the information about the structure of the available fields, acquisition processes, and communications required to interpret the conversation and generate the group are transmitted to a third-party AI provider. No contact-related values are transmitted to a third party. |
Behavioral Analysis and Optimization
Certain features use contact interaction data to identify behavioral trends or improve the delivery of communications.
| Feature | Data processing |
|---|---|
| Interests | The third-party AI provider analyzes the content of the web pages linked in communications and the pages tracked using Journey to associate them with the configured interests. No contact-related values are transmitted to a third party. |
| Send Time Optimization for emails | No data is sent to a third party. Processing is performed locally on Dialog Insight’s servers. |
| Send Time Optimization for SMS messages | No data is sent to a third party. Processing is performed locally on Dialog Insight’s servers. |
Configuration Assistance
AI can also suggest matches between fields when importing contact or relational table data or configuring an integration. The AI-powered suggestion process is enabled by default. The suggested matches must be reviewed before proceeding with the operation.
The list of relevant fields and their structure is transmitted to a third-party AI provider. No values associated with these fields are transmitted to a third party.
Known Limitations
Features that use a language model generate results based on the instructions and information provided to them. Despite the mechanisms implemented to improve consistency, certain limitations must be taken into account:
- Variable results: the same request may produce slightly different responses from one use to another.
- Accuracy is not guaranteed: a result may appear plausible while being incomplete, inaccurate, or incorrect.
- Dependence on instructions: the quality of the result depends on the clarity and precision of the request.
- Dependence on available information: missing, ambiguous, incorrect, or inaccessible information may reduce the relevance of the result.
- Context interpretation: AI may misinterpret an instruction, a business rule, or the meaning of a field.
Dialog Insight uses structured instructions, predefined output formats, and application controls to reduce variability and errors. However, these mechanisms cannot guarantee the accuracy of every result.
Important: always review content, campaigns, segmentation criteria, and other results suggested by AI before using or activating them. Users remain responsible for validating the results generated by AI features.