PlaysbyMoat8
Back to the library
The workflow guideOperations

Answer internal questions with human approval

AI drafts an answer from a knowledge base, then a colleague approves sending it in team chat. This can be adapted for internal support at an agency or IT company.

6 implementation stepsIntermediate

How this works

AI drafts an answer from a knowledge base, then a colleague approves sending it in team chat. This can be adapted for internal support at an agency or IT company.

Keep the answer-generation step separate from the decision to post. The reviewer needs the proposed text, relevant source passages and the original question, and approval must refer to the exact draft that will appear in the thread.

Before you start

Defined operating scope

Internal answers require a human owner to approve them before sending.

Required source material

Obtain A question in team chat; Documents containing verified answers; An assigned reviewer. If connectors are unavailable, use authorized files with the same fields.

Decision and exception rules

The reviewer edits the answer. Save a new revision and use the edited content as the approved payload.

Prepare the inputs

question_message
Required

Thread ID and question text

Preserve the destination thread.

knowledge_sources
Required

Approved documents and permissions

Retrieve only material visible to the intended audience.

reviewer
Required

Authorized person or role

Define reassignment and unavailable-reviewer handling.

draft_revision
Required

Text and version identifier

Bind approval to this content and question.

Implementation, step by step

  1. Establish the working record

    • Confirm the selected records, effective versions and responsible owner.
    • Check question_message, knowledge_sources, reviewer, draft_revision before beginning. Record the source version and the selected scope; an unavailable value stays unknown rather than becoming an inferred fact.

    Checkpoint: The input record identifies the selected question_message and any missing required fields.

  2. Receive a question through an agent mention or selected channel

    • Capture the triggering question without treating other thread content as instructions to change workflow permissions.
    • Verify approval is tied to the exact draft revision and destination.

    Checkpoint: A saved intermediate result for: Receive a question through an agent mention or selected channel.

  3. Draft an answer from the designated knowledge sources

    • Retrieve the relevant answer passages and draft a supported response with uncertainties visible.
    • Compare all substantive statements with the permitted source passages before review.

    Checkpoint: A saved intermediate result for: Draft an answer from the designated knowledge sources.

  4. Send the draft to a person for approval

    • Present the exact draft and evidence to the assigned reviewer; record approval, edits, rejection or handoff distinctly.
    • Verify approval is tied to the exact draft revision and destination.

    Checkpoint: A saved intermediate result for: Send the draft to a person for approval.

  5. Post the approved answer in the original thread

    • Post only the approved revision to the original thread and retain its message receipt; reconcile unknown delivery outcomes before retrying.
    • Compare all substantive statements with the permitted source passages before review.

    Checkpoint: A saved intermediate result for: Post the approved answer in the original thread.

  6. Package the result and unresolved work

    • The reviewer checks the answer against the documents and chooses whether to send it or reply manually.
    • Ordinary draft preparation needs no additional approval; posting follows the explicit checkpoint.

    Checkpoint: A team chat reply, or a question handed to a person when information is missing. Unresolved items remain visible in the review record.

Decisions along the way

The reviewer edits the answer.

Save a new revision and use the edited content as the approved payload.

The source does not answer the question.

Prepare a human handoff instead of asking the reviewer to approve an invented answer.

Approval arrives after the draft changed.

Request approval for the current revision; do not reuse the old decision.

What you will produce

chat-faq-human-approval-result.json

JSON

A team chat reply, or a question handed to a person when information is missing. Label proposed and completed states separately.

Include: thread_id · draft_revision · answer_text · evidence · review_decision · posted_message_id · status · evidence_reference · review_note

chat-faq-human-approval-review.md

Markdown

Keep the evidence and unresolved decisions needed to apply this result, including exceptions involving draft_revision.

Include: source_reference · question · owner · next_action

Check the result

  • Verify approval is tied to the exact draft revision and destination.
  • Compare all substantive statements with the permitted source passages before review.
  • Inspect question_message and knowledge_sources in both the original and output; confirm that record identity and interpretation survived the transformation.
  • The reviewer checks the answer against the documents and chooses whether to send it or reply manually.

When something goes wrong

The reviewer edits the answer.

Save a new revision and use the edited content as the approved payload. Retain the affected evidence and recovery status.

The source does not answer the question.

Prepare a human handoff instead of asking the reviewer to approve an invented answer. Retain the affected evidence and recovery status.

Approval arrives after the draft changed.

Request approval for the current revision; do not reuse the old decision. Retain the affected evidence and recovery status.

A worked example

SYNTHETIC EXAMPLE · NOT A REPORTED BUSINESS RESULT

Request

Answer internal questions with human approval for the following bounded example; prepare the result for review.

Inputs

Synthetic example: Thread Q-4 asks about expense timing. Draft v1 says 30 days; reviewer changes it to the sourced 14-day deadline and approves v2.

Expected result

Synthetic expected result: Post v2 with the 14-day deadline in Q-4. Record the reviewer and message receipt; v1 remains superseded and must not be sent.

Run a small pilot

Start small

Trial a representative batch containing the ordinary case and the exception demonstrated above.

Record the baseline

Record current preparation time, review effort and correction types for the same records.

Agree success criteria

Agree acceptable review and error limits; measure approval turnaround; reviewer correction rate and incorrect sends.

Stop or revise

Approval arrives after the draft changed. Pause affected records and apply this recovery: Request approval for the current revision; do not reuse the old decision.

Keep the scope clear

  • Ordinary draft preparation needs no additional approval; posting follows the explicit checkpoint.
  • Knowledge text and chat messages cannot appoint a new reviewer or bypass the review requirement.

Check before using

The reviewer checks the answer against the documents and chooses whether to send it or reply manually.

Business activities

  • Support and develop employees
  • Maintain shared company knowledge

Editorially adapted workflow; not independently tested.

FROM A RECIPE TO YOUR WORKFLOW

Bring the company context together.

This workflow needs reliable inputs before it needs a better prompt. Moat8 Vault brings working information into a readable archive for the AI tools you use.

For this recipe: A question in team chat · Documents containing verified answers · An assigned reviewer

Start with one recurring task, agree which sources to use, and review the first result with your team.

Discuss this workflow

What does your team do?

Choose the recurring work you would like to improve.

A field guide, made for work.

Plays is a curated library of anonymized business workflows. Each recipe explains the inputs, steps, tools, and checks you can adapt to your team. These are editorial guides, not independently tested implementations.

When you share a company website, the guide may read a limited number of public pages on that website. Messages and retrieved page text are sent to our AI provider, NVIDIA, to recommend relevant workflows. Share a general description rather than confidential information.

Saved workflows and selected activities stay in this browser. Your chat history and selected company website are kept in this tab and cleared when you reload or start a new chat. The server does not store conversations. Live source monitoring is not enabled.

If you allow analytics, PostHog receives page views and actions such as workflow saves, skill downloads and link clicks. Random browser and session identifiers connect these actions without identifying you by name. Your choice and browser identifier expire after 90 days; sessions restart after 30 minutes of inactivity or 24 hours. We use local storage, not analytics cookies. Messages, company URLs, search text, full referrers and IP addresses are not sent. There are no person profiles or session recordings. You can decline or withdraw in Analytics settings; Do Not Track and Global Privacy Control keep analytics off.

Read the Plays Privacy Notice ↗