Skip to main content
Your AI Experience is only as reliable as the information it can use. As your services, methodologies, policies, or educational materials evolve, review and update the documents in its knowledge base.
Updating a knowledge document is different from changing system instructions. Documents provide reference information; system instructions guide how the assistant behaves, including when it should acknowledge uncertainty or recommend speaking with a person.

​
When to update your knowledge

Review your knowledge whenever you:
  • Introduce or discontinue a service, program, or offering.
  • Change eligibility criteria, policies, processes, or FAQs.
  • Revise your methodology or publish updated educational material.
  • Discover a factual error, contradiction, or missing explanation.
  • Notice visitors asking questions your current materials do not answer well.
  • Reach a scheduled review date for time-sensitive information.
For regulated or frequently changing subjects, such as insurance, legal requirements, or financial products, establish a more frequent review process and have an appropriate subject-matter expert approve changes.

​
Step 1: Identify what changed

Before uploading anything, make a short change record: This helps you avoid adding new information while leaving contradictory older guidance in place.

​
Step 2: Update the source document

Edit the original document so that it clearly describes the current information.
  • Use descriptive headings and concise sections.
  • Replace outdated statements rather than merely appending corrections.
  • State the effective date when the timing matters.
  • Separate confirmed facts from examples or hypothetical scenarios.
  • Remove unnecessary personal, confidential, or sensitive information.
  • Check other documents for statements that conflict with the update.
Do not treat a document upload as a real-time data connection. An assistant should not claim to know current inventory, availability, prices, regulations, or other changing facts unless the information is supported by an appropriately current source or a connected real-time capability.

​
Step 3: Update the file in Agentsy

  1. Sign in to your Agentsy workspace.
  2. Open the AI Experience you want to update.
  3. Navigate to its knowledge or document-management area.
  4. Find the document that contains outdated information.
  5. Use the available document-management controls to remove or replace outdated content and upload the revised file.
  6. Wait until the new file has finished processing, if a processing status is shown.
  7. Confirm that the current version is the one available to your assistant.
The exact control names may vary by workspace and product version. If your interface does not offer an in-place replacement action, use the supported remove-and-upload workflow and verify that the old version is no longer being used.
Keep a dated copy of the previous document outside Agentsy if your organization needs an audit trail. Avoid retaining obsolete versions in the active knowledge base unless there is a clear reason and the documents explicitly distinguish historical from current information.

​
Step 4: Check your system instructions

Some updates affect not only what the assistant knows but also how it should respond. For example:
  • A new appointment process may require a revised human-handoff instruction.
  • A change in service scope may require updated assistant boundaries.
  • A new compliance requirement may change which questions the assistant can address.
If the assistant’s behavior needs to change, update its system instructions separately. See System Instructions.

​
Step 5: Test the updated knowledge

Ask questions that should produce the new answer, and questions that might still surface outdated information. If an answer is wrong, investigate whether the updated document was processed, an old file remains active, other documents contradict it, or the system instructions need adjustment.

​
Step 6: Establish an ongoing review routine

Assign responsibility for reviewing the knowledge base. A simple tracking table is often sufficient: These dates are examples, not prescribed review intervals. Choose a cadence appropriate to the subject matter and how quickly it changes.

​
Example: A boat advisory experience

A boat broker uses Agentsy to educate prospective buyers about vessel types, ownership considerations, and buying processes. The broker’s available inventory changes frequently. A good update strategy is to:
  1. Keep durable educational guidance in the knowledge base.
  2. Update buying-process documents when the process changes.
  3. Avoid uploading inventory snapshots as if they were continuously current.
  4. Instruct the assistant to refer requests for currently available boats to a human broker unless a verified live-inventory integration is available.
  5. Test that the assistant distinguishes general buying advice from actual availability.

​
Troubleshooting

Check whether the old file remains active, the new file finished processing, or another document contains conflicting information. Retest after making changes.
Confirm that the revised file was uploaded successfully. Make the relevant section easier to locate with a descriptive heading and direct wording. See File Search and Retrieval.
Remove or clearly label outdated material and review related documents for contradictions. Keep one authoritative current explanation where practical.
Review the source’s effective date and add explicit instructions about time-sensitive claims and when to defer to a human or live source.

​
Update checklist

  • Identify the changed information and affected documents.
  • Update the authoritative source and remove contradictions.
  • Check that no unnecessary sensitive information is included.
  • Upload the revised file and verify processing.
  • Ensure obsolete content is no longer active.
  • Update system instructions if behavior or handoff rules changed.
  • Test both new and previously supported questions.
  • Record the review date and next review owner.
Next: Learn how to structure content for reliable answers in Knowledge Best Practices.