Smarta works with your real business numbers. So you should know exactly what happens with that data. Here it is, step by step.
Smarta only accesses the data areas that sell & pick has set up for Smarta in your brand, and the dashboards you can open yourself. Data from other brands is technically excluded.
When you ask a question, the AI model processes:
your question and the earlier conversation
the name of your brand and the instructions sell & pick has stored for Smarta
the description of the data fields (e.g. "Net revenue", "Branch")
the results of the queries Smarta runs for your answer, e.g. revenue per branch
for dashboards: name, description, filters and tile results, and the name of the person who created a custom dashboard
Your name and your email address are not passed to the model as part of the instructions. One exception: if you request a test email, the Smarta platform confirms your address as the recipient.
Important: Smarta sees what is in the shared data. If your data contains personal information, e.g. names of employees in the HR analysis, it can show up in an answer. Smarta is not meant for evaluating the performance of individual employees (see Terms of use).
Smarta runs on Google Cloud Vertex AI. By default, your data is processed in the EU, within our DPA. As long as Smarta is used inside the sell & pick environment, the data is not passed on to AI providers or used for training.
Setting | Meaning |
|---|---|
EU (default) | Processing stays in data centers in the EU. |
New Smarta instances always start with EU.
In exceptional cases, a GLOBAL region can also be ordered for higher availability. Since this violates the DPA, it needs a written declaration of consent. Contact the CSM team for this.
According to its cloud terms, Google does not use the data processed through Vertex AI to train its models.
So that you can continue conversations later, sell & pick stores:
your questions and Smarta's answers
the queries and their results, including the charts in the chat
Smarta's reasoning
your feedback (thumbs and comment)
the token usage per answer
This data is kept in the sell & pick infrastructure in the EU. With Unterhaltung archivieren you hide a conversation from your history. If a conversation should be deleted completely, contact [email protected].
The sell & pick team can view conversations to analyze errors, provide support and improve Smarta. In our monitoring systems we only log technical values like duration, model and token count, not the content of your questions or answers.
Smarta can act for you, but only within clear limits:
Smarta can | Smarta cannot |
|---|---|
query data and show charts | change data in your source systems |
create and edit your own dashboards | delete or share dashboards |
read and filter sell & pick dashboards | change sell & pick dashboards |
create, change and pause reports sent to you | delete reports or send them to other people |
send you a test email | send emails to other addresses |
Smarta only changes something when you ask it to. Text found in your data is treated as data and never as an instruction.
Besides Smarta, you can also query your sell & pick data with your own AI assistant, e.g. Claude, ChatGPT or Gemini Enterprise (see Using sell & pick data in Claude, ChatGPT or Gemini). For data processing, this is a basic difference:
Smarta | Own assistant via MCP | |
|---|---|---|
Who processes the data? | sell & pick via Google Cloud | your AI provider, e.g. Anthropic, OpenAI or Google |
Where? | in the EU by default | depending on provider and contract, also outside the EU |
Under what terms? | contract between you and sell & pick | your contract with the AI provider |
Who stores the conversations? | sell & pick | your AI provider |
As soon as you connect your own assistant, the queried data leaves the sell & pick platform. From that point, sell & pick has no influence on where and how it is processed, stored or used to train models. This processing is not covered by our data processing agreement.
The decision and the responsibility for it are yours. Before you use MCP, you should check:
what contract you have with your AI provider, e.g. a business or enterprise plan with a data processing agreement
whether the provider uses your input for training and whether you can turn that off
what data your employees may query there, especially personal data such as HR data
who in your company may set up the connector
A connector can be removed at any time in the assistant's settings. After that, the assistant no longer has access to your sell & pick data.
Smarta only reads your shared data and dashboards. Processing runs through Google Cloud, in the EU by default. Your conversations stay stored with sell & pick, and Smarta sends nothing to other people. If you connect your own AI assistant via MCP, your AI provider processes the data, and you are responsible for that yourselves. If you have questions about data protection, you can reach us at [email protected].
