From a single place, you can manage both the shared knowledge for the entire organization and the individual knowledge base of each worker (AI Employee).
What types of Knowledge Bases exist?
General Knowledge Base: The main source from which all workers draw information to answer questions. It contains official company answers, procedures, and policies.
Worker Knowledge Base: Contains specific information for a particular agent. This data is not visible or usable by the rest of the workers (e.g., personal notes, specific credentials, or internal reminders).
⚠️ Important: Do not configure individual rules that contradict general rules. Worker instructions must always respect the organization's global guidelines.
Example of a contradiction to avoid:
General Base: "Returns are processed within 10 business days from the time we receive the product."
Individual Base (Incorrect): "If the customer asks about their return, tell them the process takes 3 days."
What goes in each base?
General Base: Information that all workers need to perform their jobs.
Individual Base: Specific information for a single worker that others should not use because it is not relevant to their role.
Example:
General (all workers): "Returns are processed within 10 business days from the time we receive the product."
Individual (Support Worker): "If a customer complains about an unprocessed return, instruct them to write to [email protected] with their order number."
Why doesn't this go in the general base?
The Sales worker does not manage returns. If they had access to that procedure, they could provide incorrect information or cause customer confusion.
Recommended Structure
To easily identify the source of each piece of information and quickly correct errors, we recommend creating well-categorized sections. Use clear names and hierarchies of no more than 3 levels.
Example:
General Knowledge Base
├── Policies
│ ├── Warranty
│ ├── Returns
│ └── Shipping
├── Products
│ ├── Laptops
│ ├── Smartphones
│ └── Accessories
├── Common procedures
└── Temporary notes
├── Current promotions
└── Days without support
Worker_Support
├── Handling complaints about unprocessed returns
│ └── Tell the customer to email [email protected] with their order number
└── (other knowledge exclusive to support)
Worker_Sales
├── Coupons
│ └── 5% discount coupon for first purchase
└── Free shipping on laptop
Content Types
In addition to structured text knowledge, you can enrich your base with the following formats:
Files: Upload documents so your worker can use them as context when replying, or send them directly when instructed in a specific stage of the workflow. Useful for incorporating formal or reference information: policies, instructions, product descriptions, or internal manuals. (Supported formats: PDF, DOCX, TXT, CSV, XLSX).
Websites: Add websites to pull useful content directly from your business pages (e.g., FAQs, services, purchasing conditions, institutional information). The AI will learn from this to provide precise answers. It is ideal for frequently updated pages like legal terms.
Catalogs / Products: Upload structured, repeatable, and searchable information—such as products, services, plans, or data tables with attributes—that your worker will consult on demand. Relevant records are retrieved automatically based on Stage instructions.
Reusable Content: Pieces of information that can be applied across different points of the operation.
When should the Knowledge Base be updated? Whenever core information changes, such as business hours, policies, commercial conditions, promotions, pricing, services, coverage, or internal procedures.
Technical Limits
To ensure optimal processing, keep the following limits in mind when uploading content:
Content Type | Limit |
PDF Files | 50 MB per file |
Images | 5 MB per image |
Videos | 15 MB por archivo |
Websites | Hasta 5.000 URLs por mapeo |
⚠️ Files exceeding these limits cannot be processed correctly by the worker.
Testing & Validation (Try It Out)
The platform includes a testing environment to check if the workflow responds as expected. Note: This is not a space for structural creation (you do not create employees, pipelines, or channels here); instead, you use it to test existing configurations.
When should you use the testing environment?
Before launching a new configuration.
After making changes to the setup.
To validate an applied improvement.
During team training processes.
To compare expected results vs. actual results.
💡 Testing Best Practices
Test frequent and common scenarios.
Test difficult or ambiguous queries to see how the AI handles them.
Always repeat tests after making major systemic changes.
Use realistic test cases and prompts.