sns-ops-ai-team

name

sns-ops-ai-team

description

Stand up a role-based "AI operations team" for social-media / SNS content so the AI does the operating work (strategy, research, planning, copywriting, traffic + sales design, task tracking) and the human only reviews and approves. Use this WHENEVER the user wants to "build an SNS/social-media team", "组建运营 团队", "帮我组建SNS运营团队", run social accounts with AI, "let AI run my content ops", delegate account operations, or asks how to stop doing social-media busywork themselves and become the final decision-maker. Also use it when someone has one overloaded social account and wants to hand off pieces of the workflow to AI one role at a time. Distilled from @wanerfu's thread on delegating SNS operations to Claude Code.

SNS Ops AI Team

Turn one person doing all of social-media operations into a role-based AI team where the human is only the final approver. The distilled idea: instead of doing the operating work, define the operating roles, hand each one to an AI agent, and keep the human in the "拍板" (final decision) seat — reviewing and deciding rather than producing.

The deeper lever from the source: don't try to build the whole team at once. Start by authorizing AI for one role, watch it, then add the next. The win isn't just less work — it's that the human's time concentrates on judgment and decisions instead of production.

When to use

Trigger when the user wants to:

• Build / configure an "SNS 运营团队" or "social-media team" (real or AI).
• Delegate running a content account (小红书/Xiaohongshu, LinkedIn, X, IG, Threads, TikTok, a blog) to AI and only sign off on outputs.
• Stop personally doing research/planning/writing/scheduling and move to a review-and-approve role.
• Hand off just one part of their social workflow to AI as a first step.

Also fires on phrasings like "帮我组建SNS运营团队", "AI 帮我运营账号", "我只想负责 审核和拍板", "let AI handle my content operations".

Do not use this for one-off single-post requests (just write the post), for paid-ads account audits (use an ads skill), or where the user wants a bespoke brand strategy deliverable rather than an operating system.

The six operating roles

Define the account's work as six roles. Each is a hat the AI wears; each produces an artifact the human reviews. Adapt names to the platform, but keep the coverage.

Role	中文	Owns	Produces
Strategy Lead	战略负责人	Publishing direction & positioning	Content strategy / calendar direction
Researcher	调研员	Competitor analysis & market scan	Trend + competitor brief
Planner	企划员	Content planning & topic selection	Ranked topic/angle list
Editor	编辑	Copywriting & optimization proposals	Draft posts + revision options
Marketer	市场营销	Traffic funneling (导流) & sales design	CTAs, funnels, conversion hooks
Secretary	秘书	Task & progress management	Status tracker / next-actions list

The pipeline runs in that order: Strategy sets direction → Researcher gathers evidence → Planner picks topics → Editor drafts → Marketer wires conversion → Secretary tracks it. The human reviews at each handoff and does the final 拍板.

The operating procedure

1. Name the target account and goal. One platform, one audience, one primary objective (reach, leads, sales). Everything downstream inherits this.
2. Pick the delegation depth — start with ONE role. Do not spin up all six on day one. Choose the single role that eats the most of the user's time (usually Editor or Planner) and authorize AI for just that. Prove it, then add roles.
3. For each active role, run its loop:
   • Give the role its input (the account goal + upstream role's output).
   • Have the AI produce the role's artifact.
   • Human reviews and approves or redirects — this is the only mandatory human step. Approval, not production, is the job now.
4. Chain approved outputs down the pipeline. The Strategy artifact feeds the Researcher; the Planner's approved topics feed the Editor; etc. Never let a role run on an unapproved upstream artifact.
5. Let the Secretary hold state. Keep a running tracker of what's drafted, approved, scheduled, and published so the human sees progress without doing the coordination.
6. Expand authorization gradually. Once a role is trusted, add the next one. The end state: AI runs all six roles, the human only reviews and decides.

Guidelines / rules of thumb

• The human's job becomes review + 拍板, not production. Design every role to end in an approval gate, not an auto-publish.
• One role at a time beats a full team. Incremental authorization surfaces where the AI is weak before you've bet the whole account on it.
• Delegate the role, not the task. "Be the Editor for this account" is reusable; "write this one post" is not — that's what makes this an operating system rather than a prompt.
• Optimize for concentration, not just hours saved. The point is freeing the human to focus on judgment; measure success by decision quality, not only volume.
• Keep artifacts reviewable. Each role should output something a human can scan and approve in minutes (a ranked list, a draft with 2–3 options), not a wall of text.

Pitfalls

• Standing up all six roles before trusting one. You get volume you can't vet and can't tell which role failed. Ladder up instead.
• Removing the approval gate to "save more time." Auto-publishing loses the one advantage of this model — the human as quality backstop.
• Confusing this with content generation. This is a delegation structure; if the user just wants a post written, skip the roles.
• Fixed role names. On a sales-heavy account the Marketer role dominates; on a thought-leadership account the Strategy/Editor roles do. Keep the six functions, flex the emphasis.

Example

Input: "帮我组建SNS运营团队 for my Xiaohongshu account, goal = drive DMs to my service."

Output: A six-role setup (Strategy → Researcher → Planner → Editor → Marketer → Secretary) scoped to Xiaohongshu + DM-conversion, with a recommendation to start by authorizing the Editor role only (the user's biggest time sink), a review gate after each role's artifact, and a Secretary tracker — then add the Planner and Marketer once the Editor loop is trusted.