Launch Affiliate in 15 minutes

#Prepare three things first

Affiliate begins with reliable product facts, team ownership, and clear human-review boundaries. Creating a Campaign is not a prerequisite. Before the first launch, confirm that:

1. the target Shop is Authorized and Affiliate is enabled from its Overview;
2. the current Desktop is signed in, will remain online, and has a usable model and Affiliate run profile;
3. the team has named the people responsible for reviewing important Agent proposals and sample applications.

If those foundations are not ready, start with Shop Management and Models, devices, and run profiles.

By the end of this guide, you will have one minimum viable loop: a Creator applies for a Sample or sends a message → ownership rules select a responsible BD → eligible relationships go to AI → people review policy-held actions or handle Agent Escalations → outcomes stay in the relationship history. Campaigns are an optional outbound entry; Creators can also arrive directly. See the Affiliate workflow for the complete path.

#1. Enable Affiliate and confirm how it runs

Open Shops, select the target Shop, enable Affiliate from Overview, then open the Affiliate tab. For a first setup:

• keep the built-in Affiliate run profile;
• use the Global default model until the Shop needs a deliberate override;
• bind business handling to a Desktop that will remain online;
• begin with conservative policies that hold important work for human review, then expand automation after the loop is proven.

The Shop Overview is only for authorization and service switches. Shop-level models, devices, and Business instructions belong in the Affiliate tab. Configure BD operating modes, Creator protection, outreach accounts, and Approval Policies in Team & Channels. Confirm which relationships AI may take over before starting automation.

#2. Build Product Knowledge first

Open Affiliate → Product Knowledge and create or choose a knowledge set for the product you plan to promote. At minimum, record:

• positioning, core benefits, and intended audience;
• verifiable specifications, pricing, fulfillment, and sample rules;
• common Creator questions with approved answers;
• useful creative directions and claims that must not be used.

One knowledge set can link the same product across Shops through Seller SKU. Campaigns and the Agent then work from one maintained source of facts.

APP CAPTUREThe English light-theme Creative Cases view in Product Knowledge. Cases organize creative direction, performance, and assets for reuse in Creator outreach and guidance.

Product Knowledge should contain facts the team can verify and rules the Agent can execute. Do not turn an unconfirmed result, inventory level, or platform outcome into a promise.

#3. Optional: create the first Campaign

Open Affiliate → Campaigns, create a Campaign, and define:

1. Product and offer: choose a product with complete Product Knowledge, then enter commission, sample, or other collaboration terms;
2. Target Creators: set market, category, audience size, content performance, and other qualification criteria;
3. Daily target: begin with a small number so the team can inspect match quality and message tone;
4. First message: explain why this Creator is relevant, how the product fits their content, and the next step;
5. Outreach pace: choose Uniform or Fast and verify the daily target against Shop capacity.

APP CAPTUREThe English light-theme Campaign page, showing active Campaigns, daily targets, funnel progress, and replies.

Before launch, check that the target is not too broad, daily volume does not exceed the team’s review capacity, and the message makes no unsupported promise.

Skip this section when Creators already apply for Samples or message you directly. A Campaign invitation does not assign a BD; subsequent inbound contact enters the ownership and AI-eligibility rules.

#4. Complete Agent handoffs in the Affiliate Workbench

Open Affiliate → Agent Workbench. Pending review contains Agent operations held by approval policy; Pending contains issues outside the Agent's capability or business authority; All Agent work provides the AI history. Open the separate Manual Workbench to handle Shop Sample Applications and Creator messages directly.

APP CAPTUREThe English light-theme Agent Workbench with Pending review, Pending, and All Agent work.

Keep “proposal” separate from “executed action.” A proposal is a prepared action that remains on hold until it is approved. During review:

• verify that the recommended action matches the evidence;
• confirm the product, Creator, and Campaign context;
• approve work that is ready to execute;
• rewrite, reject, or request a revision when it is not.

#5. Test human control with a sample review

Open a sample application and compare the Agent recommendation with Creator performance, relationship history, and risk evidence before making the decision.

APP CAPTUREThe English light-theme sample review. The AI recommendation and evidence are shown alongside controls for a person to approve, reject, or add conditions.

Samples, added benefits, and out-of-policy promises create real cost. Keep human approval in place until the team’s decision standard is stable. The result remains in the same Creator relationship and work history, so later conversations can continue with that context.

#6. Close the loop every day

A practical daily rhythm is:

1. review Campaign funnel and execution progress;
2. clear high-priority reviews and takeovers in Agent Workbench, and handle Samples and messages in Manual Workbench;
3. check Creator replies, Samples, and exceptional escalations;
4. add recurring questions to Product Knowledge or Business instructions;
5. increase daily targets or widen auto-execution only after review outcomes are consistently sound.

#Go-live checklist

• The Shop is Authorized and Affiliate is enabled.
• The handling device, run profile, and model are available.
• Product Knowledge contains facts, Q&A, creative directions, and prohibited claims.
• If outbound reachout is needed, the first Campaign’s target, daily volume, and opening message have been reviewed.
• The team knows where to handle held proposals and sample applications.
• At least one “Agent proposal → human review → recorded outcome” test is complete.