A knowledge assistant can answer from yesterday's content while the source owner assumes today's update is already live. Freshness is a chain: source publication, ingestion, processing, indexing and retrieval. Ask an agency to make that chain observable and to define when outdated material should stop being used.
Related service: Custom AI chatbot development- 01Publish the source change
- 02Process its version
- 03Verify indexed content
- 04Test retrieval and answer evidence
SECTION 01
Set useful freshness expectations
Identify which sources change, who approves updates and how much delay the business can tolerate. A stable installation manual differs from a live availability field. Where a fact needs an authoritative real-time lookup, repeatedly indexing documents may be the wrong approach.
Microsoft's indexer scheduling documentation describes scheduled execution, not a guarantee that every source update is immediately searchable. Require per-source version evidence and failure reporting. Distinguish time since the last attempted run from time since the last successful ingestion of the specific document. Operators need to know which material is affected.
SECTION 02
Illustrative example: a revised returns exception
A store publishes an approved change to a special-order returns exception. The main policy file updates correctly, but a related attachment fails parsing. The assistant could now retrieve a mixture of old and new rules. A simple green status for the overall job would hide the inconsistency.
The workflow flags the affected source set and runs test questions about the exception. Until the consistent approved version is available, the assistant gives the approved fallback or directs the visitor to a person. This hypothetical example does not imply every source must be synchronized atomically; it asks the buyer to define which documents must agree.
SECTION 03
Test propagation rather than just timestamps
Change a known passage, trigger the supported update path and verify that retrieval uses the new version with the right citation. Test a failed parse, moved file and unavailable source connection. Microsoft's change-detection guidance also shows why deletion needs an explicit strategy; removing a source and updating its contents are related but different cases.
At handoff, request source ownership, sync schedules, version records, alert recipients and a reprocessing procedure. Frequent updates can increase operating work and cost without helping stable content. Choose the cadence and stale-answer behavior by task, and document how staff can temporarily withdraw an affected source while investigating.
SECTION 04
Frequently asked questions
Does running the indexer mean the assistant is current?
No. Verify the relevant source version reached the searchable index and appears in the evidence used for the answer.
Should every source update in real time?
Not necessarily. Match update cadence to the consequence of stale information and use direct authoritative lookups for facts that need live confirmation.
Can the assistant mention its source date?
Yes, if that date is reliably recorded and meaningful. A date label helps users judge context but does not replace a defined freshness policy.
PRIMARY REFERENCES
Sources and further reading
These references cover the standards, platforms, or published prices discussed in the guide. Worked examples and checklists are our editorial guidance.
- Schedule search indexersMicrosoft Learn
- Change and deletion detectionMicrosoft Learn
EDITORIAL METHOD
About this guide
We use AI to assist with drafting and editing. Catapult AI Work is responsible for the published content. Examples illustrate possible approaches; they are not client case studies unless identified as such.
Budget examples are not Catapult package prices. Check linked provider pages for current fees and plan limits before making a purchase.
Read the editorial policy