A company wants to draft answers to employee questions using approved internal policies.
A reviewer checks the relevant policy passage before publishing. Questions without supporting policy text are routed to a person.
Which is the most appropriate generative AI use case?
Generative AI can draft text and answer suggestions, while important decisions should retain human review.
Detailed explanation
Search can locate evidence, while the requirement also calls for drafting a response to the specific question.
The answer may be fluent, but this omits the required review before publication.
Classification returns a routing label. It does not draft an answer from the policy.
Text generation can assist answer preparation when a reviewer approves the result.
Check official documentation
Compare the following points with the official documentation.
Read the AWS guide to generative AI use cases and the Bedrock Knowledge Bases overview. With a fictional policy requiring travel requests three business days before departure, compare a deadline question with an exception the policy does not mention.
Official references
What to confirm
Draft the three-business-day deadline with its supporting clause. Refer the undocumented exception to a reviewer, who decides what can be published.
Key points
- Text generation
- Business assistance
- Human review
Notes
- Environment: Paper exercise with AWS documentation and a fictional company policy
- For questions with no supporting text, refer the question to a person.
Foundation review
Evidence and approval
Use the relevant policy as context for the draft, then have a reviewer check the clause, version and conditions. Route questions without supporting text to a person.
Checking the draft
A fluent draft can still change a deadline or miss a condition. Before release, a person checks the wording against the approved policy.