AI Features and Data Processing

Plauti offers several AI-assisted features, and there are more to come. While these features can be very useful and save you quite a bit of time, of course you will also want to know how your data is handled before you start using them.

The AI-assisted features managed by Plauti section concerns the features where Plauti provides AI access. These are recognizable by the fact that you need to sign a Data Processing Agreement (DPA) before you can start using it.
For other AI-assisted features, such ask the ‘Ask AI’ actions in Plauti Manipulate, you yourself decide how AI is accessed. This also applies to the Plauti MCP. Read more about data handling in such AI features at the end of this article, at Other AI-assisted features.

AI-assisted features managed by Plauti

No AI actions without a signed DPA

First of all, no data is processed or recommendations made using AI at all, unless the DPA (Data Processing Agreement) is signed for your Org or Environment. The different AI-assisted features cannot be enabled without a signed DPA. In turn, signing the DPA ensures your data is processed securely and in line with privacy regulations.

Only relevant data is processed

The AI-assisted features use data at different levels. For example, the AI Scenario Builder solely looks at which fields you have in use, not what is recorded in those fields. The AI Merge Recommendation and AI Match Recommendation features on the other hand need to view the content of record fields, to be able to give recommendations: about which fields to keep upon merge, or which records are likely to be duplicates.

Plauti employs sub-processors for its AI-assisted features. However, data is sent for processing strictly at the request of Plauti, and only within the set boundaries; the AI provider does not have independent access to our systems.

Feature	Field Content Viewed by AI
AI Merge Recommendation	The content of all fields of the Object you are merging is viewed by AI.
AI Match Recommendation	The content of all fields of the Objects where you are checking for duplicates is viewed by AI.
AI Scenario Builder	No field content is viewed by AI. The AI Scenario Builder only views field metadata to check which fields are in use.

The key tenets for AI data processing

When it comes to processing this data, these are the points we want to make clear:

• The AI-assisted features only process data requested by a user with the appropriate permissions in Salesforce or Dynamics.
• Only data relevant to the AI feature is retrieved.
• All data is sent to the LLM via SSL encryption.
• Plauti does not store the data after processing, and it is never used to train AI models. Our AI sub-processor (Microsoft) may temporarily store prompts and outputs flagged as potentially abusive, only for abuse monitoring.

You are in charge

Furthermore, the suggestions made by the AI-assisted features are always visible for review, and require your confirmation before any changes are applied, unless explicitly stated otherwise. The AI Merge Recommendation, AI Match Recommendation and AI Scenario Builder features do not make any changes unless you confirm this first.

Who is handling your data?

In the DPA document, which you can download for review before signing, you'll find a list of the different sub-processors employed by Plauti. These concern data handlers for all Plauti's apps and features. If a sub-processor handles data for the AI features specifically, this is mentioned in its description there (DPA, Annex IV: List of sub-processors).

Other AI-assisted features

Next to the DPA-regulated AI features mentioned above, Plauti also offers a number of AI-assisted features where you select and manage your own LLM for handling the data. Currently those are the ‘Ask AI’, ‘Ask ChatGPT’, and ‘Ask ChatGPT Vision’ actions in Plauti Manipulate, and the Plauti MCP that can perform Deduplicate actions. More features might be added in the future.

AI actions in Plauti Manipulate

When you use the ‘Ask AI/ChatGPT’ actions, the prompts, field values, and documents you include in an action configuration are sent to the AI model of your choice, as configured in the Action Library.

Via these actions, that AI model then only has access to:

• The main prompt and follow-up prompts you define in the action,
• The fields and their content that you explicitly reference in the prompts (using Field Variables),
• The selected records that are being processed by the action, and any records you added to that selection via ‘Add Additional Data’,
• and any files you ask it to process via the action configuration.

The AI model (LLM) cannot query other records on its own, nor can it bypass your configured field selection for the selected records. However, whether record data is used for training purposes depends on the AI provider you are using.

Plauti MCP

Plauti MCP lets you use your own desktop AI app, such as Claude Desktop, Claude Code, Cursor or Codex, to work with Plauti Deduplicate. Plauti does not provide or manage the AI in this case. You choose the AI app and the AI provider, and their terms and data policies apply.

The AI app can read data from your Salesforce org, for example objects, fields, records, and your Plauti Deduplicate scenarios, jobs and results. It can only do what the Salesforce user you logged in with is allowed to do. When you review results, record values such as the fields of the records in a duplicate group pass through your AI app and its provider. The AI app asks you to confirm before it performs actions such as merging records, editing a scenario, or deleting a job.

Contrary to other AI features in Plauti, when using Plauti MCP, your AI app can read other data from your org as well, within the permissions of the Salesforce user. Whether record data is used for training purposes depends on the AI provider you are using.
If your organization has rules for AI usage and data processing, check them before you start using Plauti MCP with production or even sandbox data.

For information on where Plauti MCP jobs run and what is stored on your computer, see Location of Plauti Deduplicate Data Processors.