Knowledge base
The knowledge base lets you give your AI Agents access to your own documents so they can answer questions using information specific to your brand, products, and policies rather than relying on general AI knowledge.
Overview
When you upload documents to the knowledge base, UltraCart processes them and makes the content searchable. During a conversation, the agent automatically searches your knowledge base before answering questions about your policies, products, or procedures. This ensures responses are grounded in your actual content rather than generic assumptions.
The knowledge base helps your agents:
Give accurate, policy-compliant answers
Maintain consistent brand voice across all interactions
Reduce manual support load for repetitive questions
Key concepts
Knowledge base document -- any file you upload (PDF, text) containing information about your brand, products, or policies
Global knowledge -- the knowledge base is shared across all your AI Agents, though each agent's instructions control how strictly it relies on it
Prerequisites
An UltraCart account with access to AI Agent configuration
At least one AI Agent set up for your storefront
Documents you want the agent to reference (policies, guides, FAQs)
Uploading documents
Navigate to CRM > Workforce > AI Agents and select the agent you want to configure.
Scroll to the Knowledge Base section below the SMS instructions.
Select Upload Knowledge Base Documents and choose one or more files from your computer.
Save to confirm the upload.
After upload, UltraCart automatically processes your documents -- extracting text, analyzing images, and preparing the content for search. No additional configuration is required.
Large documents may take longer to process. They don't need to be perfectly structured, but clear headings and logical organization improve results.
How the agent uses the knowledge base
When a customer asks a question involving your policies, products, or procedures, the agent:
Identifies that the question may be answered by your documents
Searches the knowledge base for relevant content
Uses the matching content as the primary basis for its response
Generates a conversational answer grounded in your documents
This search happens automatically before the agent commits to an answer, even if the underlying AI model has general knowledge of the topic. The goal is to ensure responses follow your specific rules and policies.
If the knowledge base contains a clear answer, the agent references it directly. If the content is partially relevant, the agent blends it with general knowledge and may note where it's supplementing. If no relevant content is found, the agent can ask the customer for clarification or escalate to a live agent.
Pairing with chat instructions
You can control how strictly the agent relies on the knowledge base through your chat instructions. Add directives such as:
"Always consult the knowledge base first for questions about policies, shipping, returns, warranty, or product care."
"If the knowledge base does not clearly answer the question, say you don't know and suggest contacting support."
"Follow the policies described in the knowledge base exactly. Do not make exceptions or guesses."
These instructions help ensure the agent treats your knowledge base as the source of truth and avoids inventing policies.
Testing and refining
After uploading documents and setting instructions, test the agent with questions that should be answered from your knowledge base:
"How long do I have to return an item?"
"What is your international shipping policy?"
"What warranty do you offer on electronics?"
Verify that answers match the content in your documents and use the correct conditions, timeframes, and exceptions. If answers are off, try:
Reorganizing or clarifying your documents
Strengthening your chat instructions
Breaking large, mixed-topic documents into separate, focused files
Designing documents for better results
You can improve answer quality by structuring your documents well:
Use clear headings for each main topic
Group related content into logical sections (e.g., "International returns," "Damaged items," "Warranty claims")
Avoid large, unstructured blocks of text covering many unrelated topics
What to upload
Strong candidates for the knowledge base include:
Shipping and returns policies -- rules for domestic and international orders, timeframes, conditions, exceptions
Brand overview -- brand story, values, mission, FAQs about your company
Product care guides -- washing, storage, and usage instructions by product category
Sizing and fit guides -- general sizing advice, measurement instructions, size charts
Warranty information -- coverage details, claim procedures, valid and invalid claim examples
Setup and installation guides -- manuals, how-to steps, troubleshooting checklists
Support playbooks -- internal guidelines for issue handling and escalation rules
Wholesale and B2B terms -- minimum orders, lead times, payment terms
What not to upload
Avoid uploading content that is too granular or changes frequently. These are poor fits for the knowledge base:
Large SKU-level catalogs -- massive product exports with per-item specs. The agent may retrieve details for the wrong product.
Frequently changing data -- current prices, live inventory, daily promotions. The knowledge base can't stay in sync with rapid changes.
Hyper-specific one-off notes -- details that apply to a single item or a past promotion. These are likely to be retrieved out of context.
Keep the knowledge base focused on global, reusable rules and explanations. For per-product details, enable the Grant access to storefront and item data capability instead.
Troubleshooting
The agent gives answers that don't match my policies. Confirm your policies are uploaded and clearly written. Strengthen your chat instructions to prioritize the knowledge base (e.g., "Always use the knowledge base for policy questions and never invent policies").
The agent makes up details that aren't in my documents. Add an instruction like "If the knowledge base does not contain the answer, say you do not know and suggest contacting support." Also ensure your documents cover the common questions customers ask.
The agent ignores the knowledge base and gives generic answers. Verify the knowledge base is enabled and attached to the correct agent. Check that your instructions don't tell the agent to rely on general knowledge. Test with very specific questions that can only be answered from your documents.
Customers get wrong product-specific information. If you've uploaded large SKU lists or per-item details, remove or reduce them. Use category-level information instead (e.g., "All items in this category have a 1-year warranty").
File upload fails or is stuck processing. Ensure the file is a supported format (PDF or text). Try splitting very large documents into smaller files. If the problem persists, contact UltraCart support.