1. Understand the AI Knowledge Hub
A business AI assistant needs more than a general language model to answer questions about your products, services, policies, opening hours, pricing or business processes. Your Knowledge Hub is where you provide the business-specific information that can help the assistant respond with context.
Think of it as a digital reference library for your AI. It may contain approved information about your business that the assistant can use when answering customers, depending on the platform's retrieval and processing capabilities.
General AI knowledge
Language, broad concepts and common knowledge the AI already has from its model.
Your business knowledge
Specific facts, rules, offers, procedures and details that you supply and maintain.
2. Why business knowledge matters
Without business-specific context, an AI assistant may be able to converse naturally but may not know what your business actually offers. A well-prepared knowledge base can help it respond to recurring questions with information aligned to your business.
More relevant answers
Provide information about your actual services, products, policies and business processes.
Less repetitive work
Give the assistant a reference for common customer questions so staff do not have to repeatedly explain the same details.
Consistent information
Keep approved details in one maintained reference source and update it when the business changes.
Better customer experience
Help customers get clear answers about what you offer, how to proceed and where to find support.
Results depend on the clarity of your content, the AI configuration, how information is indexed or retrieved, and the user's question. Uploading more files does not automatically make answers better.
3. Prepare your documents before uploading
Preparation is one of the most important steps. Review your documents first and remove content that is obsolete, contradictory, confusing or not intended for customers.
Before you add a file
- Use the latest version of the document.
- Make sure the text is readable and the document is not corrupted.
- Use clear headings and short, specific explanations.
- Check prices, dates, addresses, contact details, product descriptions and policy wording.
- Remove duplicate documents and old versions that conflict with current information.
- Do not upload sensitive information or data that should not be accessible to the AI system.
Example: improve the document
Less useful
“We deliver quickly. Delivery depends on location.”
More useful
“Orders within Nairobi are usually delivered within 1–2 business days after payment confirmation. Delivery timelines for locations outside Nairobi are confirmed with the customer before dispatch.”
The second version gives more concrete guidance. Only publish service promises that your business can actually meet.
4. What business information should you add?
Start with information customers frequently ask about. You do not need to upload every document your business has. Prioritize material that is accurate, useful and suitable for customer-facing assistance.
Business profile
Business name, location, operating hours, contact channels, service areas and a short overview.
Products and services
Descriptions, variations, specifications, inclusions, exclusions, availability rules and pricing where accurate.
FAQs
Answers to common questions, how-to instructions and basic customer guidance.
Policies
Returns, refunds, cancellations, delivery, booking, warranty or other customer policies.
Procedures
How customers place orders, book services, make payments, request assistance or follow up.
Support information
Escalation steps, support hours, contact information and cases that require a human.
5. Uploading knowledge into Shilte AI
Use the Knowledge Hub area in your Shilte workspace to add your business information. The exact controls and accepted file types can depend on the current version of the platform; follow the file upload interface and any displayed requirements.
Sign in to your workspace
Open Shilte AI and sign in using the account associated with your business.
Open your Knowledge Hub
Navigate to the knowledge or file-upload area for your business workspace. Confirm you are working in the correct business account.
Select the prepared document
Choose a file that you have reviewed and cleaned up. Use a descriptive filename so you can identify its purpose later.
Upload and wait for processing
Follow the interface instructions and allow any upload or processing step to complete. If a status or confirmation is displayed, check that the file is accepted before continuing.
Confirm the file is listed
Look for the uploaded file in your knowledge area. If it does not appear or shows an error, do not assume the information is available to your assistant.
Review and test
Ask questions that the document should answer, then compare the response against your approved source information. If available, test in a preview or controlled conversation before customers use it.
6. Organize your knowledge for easier maintenance
Make your Knowledge Hub manageable as your business grows. Use filenames and document structure that tell you what a file contains, who owns it and when it was last reviewed.
Filename pattern example
Category - Topic - Version or Review Date
Products - Main Catalogue - Nov 2026
Product names, descriptions, key details and current prices if maintained.
Customer Support - Delivery Policy - Nov 2026
Current delivery areas, timelines and customer instructions.
Business - Frequently Asked Questions - Nov 2026
Common customer questions with concise approved answers.
Operations - Booking Procedure - Nov 2026
How customers book, what details they need and what happens after submission.
A simple document structure
- Document title and review date.
- Purpose of the document.
- Clearly labeled sections with descriptive headings.
- One topic or policy per section.
- Plain language, clear steps and consistent terminology.
- Owner or responsible team, if helpful to your internal process.
Keep an authoritative copy of each policy or catalogue. When replacing information, remove or retire older versions if your workflow allows it. If the platform does not offer file versioning or deletion controls, keep an external change log and contact support for the appropriate management process.
7. Test what your AI knows
Testing is essential. Do not assume that an upload has improved your assistant until you ask real questions and review the responses. Test both straightforward questions and situations where the answer is not present in the documents.
Build a small test set
| Question to test | What you are checking |
|---|---|
| “What services do you offer?” | Does it explain your actual service list? |
| “How much does this service cost?” | Does it use current pricing or appropriately ask for clarification? |
| “Do you deliver to my area?” | Does it reflect your service coverage and avoid guessing? |
| “What is your refund policy?” | Does it explain your actual policy accurately? |
| “Can I get a discount that is not listed?” | Does it avoid inventing offers and route the request appropriately? |
| “What if you do not know the answer?” | Does it acknowledge uncertainty and provide an appropriate next step? |
How to review an answer
- Compare the answer against your source document.
- Check that facts, prices, conditions and dates are correct.
- Look for invented details, overconfident claims or missing caveats.
- Check that the assistant does not disclose internal-only material.
- Check that it offers a human handoff when necessary and that your handoff process works.
- Record weak questions and revise the source content or configuration accordingly.
8. Improve your knowledge base over time
Your knowledge base is a living business resource, not a one-time upload. As the business changes, the documents that support your AI should change too.
Use feedback as a maintenance signal
- Notice questions the AI repeatedly gets wrong or cannot answer.
- Find the relevant source document, or confirm that the answer is missing.
- Update the source with clear, approved information.
- Remove or replace outdated material where possible.
- Retest the same question and a few related questions.
- Record what changed and when it was reviewed.
Review regularly
Set a review schedule that fits how frequently your business changes. Product prices, promotions and operating hours may need more frequent review than general company background.
Assign ownership
Decide who in your team is responsible for keeping each important document accurate and approved.
When you add new documents, check whether they conflict with existing content. Two different refund policies or price lists may confuse your AI retrieval system and make answer quality unpredictable.
9. Privacy, security and responsible knowledge sharing
Upload only information that your business is authorized to share with its AI system and that is appropriate for the AI to use in customer conversations.
- Do not upload passwords, access tokens, secret keys, payment credentials or private security instructions.
- Do not include customer personal data, private employee records, identity documents or confidential case files unless there is a clear, approved, secure purpose and suitable safeguards.
- Separate internal staff procedures from customer-facing information. Do not assume uploaded documents are automatically restricted by audience.
- Remove confidential material from documents before upload, including hidden text, comments, tracked changes and metadata where relevant.
- Use approved information and review the platform's privacy, data handling and access controls before uploading sensitive content.
- Test how the assistant responds to requests for confidential information and configure access and escalation appropriately.
10. Recommended workflow for a business owner
Use this practical routine when setting up your Knowledge Hub for the first time.
- Start small: Begin with your business overview, FAQs, products or services, and the most important customer policies.
- Clean your source documents: Make sure they are clear, accurate and current.
- Upload: Add the approved files to the correct workspace following the interface.
- Test questions: Ask 10–15 real customer questions, including edge cases.
- Review: Compare the AI answers to your approved source content.
- Improve: Fill knowledge gaps, clarify confusing sections and retire outdated content.
- Monitor: Keep reviewing customer conversations and update knowledge as your business changes.
Suggested starter knowledge set
| Document | Priority | Purpose |
|---|---|---|
| Business profile and contact details | High | Establish basic business context |
| Products or services overview | High | Answer questions about what you sell or provide |
| Frequently asked questions | High | Cover recurring customer queries |
| Customer policies | High | Clarify delivery, returns, cancellations or booking rules |
| Process and support instructions | Medium | Explain next steps and when to contact staff |
11. Troubleshooting knowledge upload and AI answers
| Issue | Possible cause | What to try |
|---|---|---|
| File upload fails | Unsupported format, file size limit, connection or processing issue | Check the upload requirements, try a smaller or clean file, and retry. If it continues, contact support. |
| AI cannot answer a question that is in a document | Content may be hard to extract, retrieval may not find it, or the question may be phrased differently | Use clearer headings and simpler text, check upload status, test different phrasing and report persistent retrieval issues. |
| AI gives outdated pricing | Old documents or conflicting sources remain available | Update the authoritative price list and retire old copies where possible; retest. |
| AI invents details | Answer may be unsupported by your content or system instructions | Clarify relevant source information and instructions, test unknown-answer behavior, and review the response before deployment. |
| AI gives vague answers | Source content is general or missing specifics | Add clear, actionable information, relevant conditions and examples. |
| Two answers contradict | Documents disagree or instructions conflict | Identify the current approved source, remove ambiguity and test both questions again. |
12. Frequently asked questions
Does uploading documents train or retrain the AI model?
Generally, adding documents to a knowledge base provides information for the AI to reference. It is not the same as fine-tuning or retraining the underlying model.
Will the AI know everything in my uploaded files?
No. Uploading a document does not guarantee that all its content will be retrieved for every question. Test important information directly.
Should I upload all my business files?
No. Start with relevant, approved customer-facing content. Avoid unnecessary, confidential, duplicate or outdated files.
Can I upload a price list?
You can use an approved price list as business knowledge if supported by the platform. Keep it current and test price-related questions. For products with live inventory or changing prices, confirm that your AI is connected to the correct live source before allowing it to quote definitive availability or prices.
How many documents should I upload?
There is no universal ideal number. A few well-structured, accurate documents are often a better starting point than a large collection of duplicates or conflicting files.
What if a customer asks something not covered?
Review the AI's behavior for unsupported questions. Ideally, it should acknowledge when it lacks reliable information and guide the customer to a human or an appropriate next step, rather than guessing.
How often should I update the Knowledge Hub?
Update it whenever business information changes, and schedule regular reviews. For fast-changing prices, availability, promotions and policies, review more frequently.
Can I add internal documents?
Only if they are appropriate for the AI's audience and access model. Keep private internal information separate unless access controls and the intended use have been reviewed and approved.
13. Video walkthrough
Use the video walkthrough below when it is available to follow the process on screen. This placeholder can be replaced with your actual recorded tutorial.
Video recording will be added here
For a useful recording, demonstrate preparing a sample FAQ, uploading it, confirming the file is listed, asking a few test questions, and showing how you update a document when an answer needs correction. Use sample business data and avoid exposing private customer information or credentials.
Ready to put your knowledge to work?
Your AI assistant becomes more useful when the business information behind it is clear, approved, current and regularly tested. Start with a small set of essential documents, evaluate the answers and build from there.
This guide explains knowledge-base best practices. Features, accepted file formats, upload limits, processing behavior and retrieval controls may vary by current Shilte AI product version.