Manual Workbench

#What belongs in Manual Workbench

Open Affiliate → Manual Workbench. Its two tabs are Sample Applications and Creator messages: a central place for direct staff operations across Shops, not AI review or Escalation queues. This includes work for protected Creators and Human-only BDs.

The Manual Workbench brings human-owned back-office tasks together across Shops. These tasks do not indicate that AI is running, and they are not Agent Escalations. Viewing or processing them does not start Agent work.

#1. Sample Applications

This tab directly aggregates Sample applications that still need manual handling across Shops. A Sample can be reviewed here even when the Agent did not discover or advance it.

Before acting, check Product, Shop, expiry, Creator performance, and current platform state. Approve, reject, or ignore updates the real Shop back-office workflow.

The Sample list is both a live queue and a history view. Its status filter includes All, Pending review, Ignored, Approved, Rejected, Cancelled, Expired, and Sync issue. It opens on Pending review; switch to another status when investigating history. Ignored applications show whether the Agent or staff ignored them and can be reopened before platform expiry.

Sort oldest or newest first; the default prioritizes older applications. For Sync issue, refresh and verify platform state before attempting another approval or rejection.

Narrow the list by Shop, BD, Creator, protection state, time range, or Product. Product filtering can select Shop Products directly or use the Internal product code from Product Knowledge to match the same business Product across linked Shops. Internal-code lookup uses active bindings inside the current Shop scope. If it returns nothing, check that Product Knowledge has the code and at least one Product binding.

Applications with a confirmed, elapsed platform approval deadline leave both Pending review and its count. Expiring soon counts only future deadlines within 24 hours. An unknown deadline is not proof of expiry. Use All or the appropriate historical state to review old applications.

#Choose the region and Shops, then read Creator performance

Search the Shop picker by name or alias, select all Shops in one region, or choose several manually. Changing Shop scope clears selected catalogue Products; select them again under the new scope. A BD can select only accessible Shops within their assigned regions.

Newer Sample lists show Creator performance directly: GMV, Sample-to-post rate, average video views, and units sold. Sample-to-post rate is TikTok's sample publication rate, not parcel delivery. Values use the Sample Shop's market. GMV may specify 30 days, 90 days, lifetime, or an unspecified period; other metrics do not automatically share that period. A dash means missing, not zero.

APP CAPTUREThe English light-theme Manual Workbench Samples list with multi-Shop scope, review status, and Creator performance.

The application-time snapshot in Analytics can differ from currently available Workbench performance. Copy Sample order IDs from the list or detail; an application ID is not a substitute when an order ID is missing. See Sample shipping for tracking and five-day warnings.

#2. Creator messages

This tab directly aggregates Creator messages that need a human response across Shops, allowing a multi-Shop team to work centrally instead of opening each Seller Center.

When product facts are incomplete, update Product Knowledge first. When a Creator asks for commission, Samples, or terms outside the authorized offer, an owner should handle the request within the business boundary.

#Query handled messages and conversation history

Select Pending, Handled, or All, then narrow by channel, BD, region, Shops, and date. Platform chat belongs to a specific Shop. WhatsApp and Email belong to contacts under the seller account and should not be assumed to carry the same Shop scope.

Pending messages support longest-waiting-first or newest-first order; Handled/All use the corresponding message time. Prioritize longest waiting to clear a backlog and newest first when investigating recent interactions.

Channel counts follow the selected status and owner scope but are not narrowed by the current date and channel selections, so they can exceed the visible result count. Message content loads when a conversation is opened or more history is requested. A listed conversation does not mean its entire history is already loaded.

APP CAPTUREThe English light-theme Manual Workbench messages page with pending, handled, and all statuses plus region, Shop, and BD filters.

For a TikTok sending restriction, read the displayed platform reason and suggested next step. A rejected send is not a successfully sent message awaiting refresh. When replying to a public Creator through an existing WhatsApp/Email contact, select the correct contact and sending account explicitly. Working a public item does not transfer the account to you.

#Distinguish list dates from the AI handling window

Manual Sample dates filter when the system first observed the application. Pending messages use the latest unanswered-message time; Proposals and Escalations use creation time. Workbench timestamps use the operator's local timezone. This differs from Campaign Shop-local days and Analytics application-date cohorts.

The Unreplied-message window in Team & Channels controls AI eligibility rather than access to old records. Staff can still handle messages outside that window manually.

#Trace who acted, reviewed, and sent a message

Samples and the timeline identify the actual operator. All Agent work shows who decided a Proposal, and messages identify the sender. The responsible BD is the relationship owner, not necessarily the person who performed an action. Analytics Details also supports reviewer filtering.

After reassignment, existing Proposals retain their original reviewer scope rather than automatically moving to the new BD. Ask a supervisor to coordinate read-only colleague records or inaccessible old Proposals; repeated reassignment does not replace permission checks.

When no reply is required, open the conversation and select No reply needed · Mark handled to finish the human task without sending anything to the Creator. If the same message still has a Pending Proposal, decide that proposal first so the system does not keep a prepared action while the message is marked complete.

#Replying: channel, attachments, and send results

Open Communication in Creator detail, confirm the recipient and Shop, then select reply/default channel, platform chat, WhatsApp, or Email. For setup, see WhatsApp & Outlook.

• Platform chat sends through the selected Shop. For WhatsApp and Email, also check the contact, BD-owned account, and preferred sender.
• A new Email thread requires a subject. Attach files; Email images can be inline. Messages can combine text and files, subject to the selected channel's format and limits.
• An existing public-Creator contact can belong to another BD. Selecting it does not transfer that account or Creator to you.
• A failed send on the selected channel does not silently switch to another channel. Read the error before changing account, permissions, content, or attachments.

Submitting a message and refreshing its history are separate steps. If submission succeeded but history refresh failed, refresh history rather than sending again just to make it appear. A platform rejection, by contrast, is not delivery: follow the displayed reason.

APP CAPTUREEnglish light-theme Creator communication with channel history and reply controls.

#Do not confuse the two kinds of human work

• Handling an original application or message here is a direct staff operation.
• For an AI-prepared action held by policy, use Pending review in Agent Workbench.
• For an issue the Agent actively hands off, such as deciding a replacement shipment, use Pending in Agent Workbench.

A protected Creator's Sample is decided manually here; an AI-prepared Sample decision held by policy needs review instead. The word “Sample” alone does not identify the right queue.