A: Hey mrsalsabor!

Great questions about managing multiple businesses with AgenticFlow.

1. Using for 2 Separate Businesses (Tiers):

- Best Practice for Separation: Ideally, for clean separation of data, agents, knowledge bases, and potentially billing/credit usage, you would purchase two separate AppSumo LTD codes (one for each business) and set up two distinct workspaces. This ensures complete data isolation.

- Workaround (Less Ideal): You could manage both businesses within a single workspace (using one LTD code), especially on higher tiers with more users/resources. However, all agents, workflows, knowledge, and connected accounts (MCPs) would reside in that single workspace, visible to all users within it. You'd need to rely heavily on naming conventions (e.g., "[Business A] Sales Bot", "[Business B] Support Workflow") to keep things organized. This doesn't offer true data isolation between the businesses if users from both access the same workspace.

2. How AgenticFlow "Learns" About a Business:

- Yes, absolutely! Agents "learn" and become knowledgeable about a specific business primarily through the Knowledge Base you provide.

- How it Works (RAG): You feed the agent information specific to that business. This can be done by:

Uploading Files: Soon (ETA May 12th for the revamped Library section), you'll be able to directly upload PDFs, DOCX, TXT files containing company info, product details, FAQs, SOPs, etc., into the agent's dedicated Knowledge Base. (Currently, you can use Table datasets).

Website Crawling: You can point the agent to the business's website, and it can crawl and index the text content to understand its offerings, about page, etc.

The agent then uses this Knowledge Base via Retrieval-Augmented Generation (RAG). When asked a question or given a task related to the business, it first searches its specific knowledge base for relevant information and then uses that context along with its underlying LLM to provide accurate, business-specific answers or complete tasks.

Prompts & Instructions: You also "teach" the agent by giving it clear instructions in its system prompt (defining its role, tone, specific rules) and within workflow prompts.

So, it doesn't "learn" by retraining the core AI model itself, but rather by accessing and utilizing the specific business information you provide in its knowledge base and instructions.

Hope this gives you a clearer picture for setting up your businesses!