Provide a supported source
Use the supported upload or entry controls for business information. Directly uploaded CSV data is distinct from files accessed through Google APIs.
What is a WhatsApp AI knowledge base?
A knowledge base supplies information that an assistant can consult when answering a question. In Bujji AI, local retrieval indexes and searches supported business information on your device. Retrieved fields can then become part of the context used to generate a reply.
Enterprise Brain is a retrieval workflow, not a promise to train a new foundation model on your files. Retrieval and generation have different data-handling paths: searching locally does not mean a later cloud-generated response is fully offline.
From business information to reply context
Search on the device
Local embedding and search models index the knowledge information and retrieve relevant records. This step is retrieval, rather than foundation-model training.
Use selected context
A supported AI workflow can include selected retrieved fields in a cloud prompt. The response still depends on source quality, retrieval relevance and the generating model.
Local retrieval and cloud generation
The privacy policy describes local indexing and retrieval as on-device processing. It also explains that selected knowledge context can be sent for cloud AI generation. These statements apply to different stages of the workflow and should be understood together.
The general question-answering assistant has a separate embedding and local-cache path described in the policy. Do not assume that every feature called “AI” uses the same processing path as Enterprise Brain.
Understand the data path before enabling replies
Identify the source information, which fields may be retrieved, and what selected context the reply workflow may send to an external AI service. Provider processing depends on the applicable service terms and configuration. The privacy policy gives the fuller feature-by-feature explanation.
Google API data is not a knowledge source
Bujji AI’s current policy states that the AI knowledge base and business catalog are built from directly uploaded files or information you type. They cannot be connected to Google Sheets or another Google data source for the knowledge pipeline.
Optional contact or spreadsheet import features have a different purpose. Data obtained through Google APIs is not passed to AI models or providers. A file that looks like a spreadsheet does not make these two paths interchangeable.
For contact organization, read the WhatsApp CRM guide. For how context fits into responding, read the AI reply guide.
Prepare knowledge that is useful to search
- Start with a supported business file format and a small, controlled example.
- Keep each record understandable: use clear product names, identifiers and descriptions rather than ambiguous fragments.
- Remove information you do not need the assistant to consult, especially sensitive or unnecessary personal information.
- Test questions that should match your source and questions for which your source has no answer.
- Review the released product’s update and deletion controls before depending on a long-lived knowledge collection.
Illustrative workflow · fictional scenario
A small product catalog
A directly uploaded catalog can provide product identifiers and descriptions for retrieval. A question about a listed product may find relevant information. A question about an absent product needs separate handling; uploading a catalog does not guarantee an accurate answer to every question.
This is an explanation of the retrieval approach, not a customer result or a measured accuracy claim.
Use human review for uncertain or important replies. Check the current plan and supported formats before relying on a business knowledge collection.
Questions about Enterprise Brain
Does Enterprise Brain train a new AI model?
No. The described knowledge workflow indexes and retrieves information. Retrieval-assisted generation is different from training a foundational AI model on your business files.
Does my knowledge stay entirely on the device?
Local indexing and search run on your device, but selected retrieved context may be included in a cloud prompt to generate a reply. Read the policy and the controls for the workflow you enable.
Can I connect Google Sheets as an AI knowledge source?
The current policy says the knowledge base and business catalog cannot use Google Sheets or other Google API sources. Directly uploaded files and typed information are the stated knowledge sources.
Can I rely on every generated answer?
No. Source quality, retrieval relevance and model behavior affect the result. Test known and missing-answer questions, and review uncertain or consequential replies.
Where can I see the product before launch?
The homepage includes an Enterprise Brain preview. The public Chrome Web Store launch is coming soon; this guide does not provide an unavailable installation link.