← Business Operations

Brightline: Biz Ops case

v1.39 Nov 2026Business OperationsHiring step 2
Download the case foldergops-case-1.3.zip

Brightline runs TikTok Shop affiliate programs for five consumer brands. Every day, creators message us about samples, commission and what to film. We're an AI-first team. An agent drafts every reply by running skill files. The GOPS runs the agent, checks what comes back and makes the calls the agent shouldn't make alone.

This case is a Monday on one pod's affiliate inbox, on made-up data. All names, brands and numbers are fictional. It's the same case our Growth Operations Manager candidates take, with one extra deliverable for Biz Ops: the bottleneck you'd fix next.

You're the pod's Growth Operations Manager (GOPS). The inbox is yours: 50 messages across five clients. Marcus, your account manager, is at an offsite today. He'll read one note from you at 16:00.

You need an agent that works on files

The setup is the same as for the short screener.

  • Claude Cowork is the minimum, and Claude Code is ideal.
  • Any other agent works, as long as it can open this folder, follow the skill files and write files back.
  • Plain chat won't get you through 50 messages.

The rules

  • Plan on about 2 hours. Stop by 3 hours at the latest. Tell us roughly how long you spent and what you didn't get to.
  • Point your agent at tasks.csv and let it run. 50 messages is too many to do by hand, and that's on purpose. Your time goes to checking, fixing and deciding.
  • You own every reply that goes out. Some drafts will be wrong, and some messages need more than a reply. When a draft is wrong, fix it. Then trace the mistake to where it came from: the data, a skill file or the AI making something up.
  • Every client is different. Each one has its own preferences. Your replies and actions should follow them.
  • Decide or escalate. pod.md says what's yours to decide and which expert handles what. Escalate only what truly needs someone else, and give them what they need to act fast. Marcus has 15 minutes for decisions today, so choose what reaches him.
  • Don't edit the skill files. Ops owns them. Write what should change, with the evidence, in requests.md.

What's in the folder

  • pod.md: who's who, the experts you can escalate to, what you can decide, and a map of every file. Start here.
  • CLAUDE.md: tells your agent how to work in this folder. Claude Code reads it on its own. Tell any other agent to read it first.
  • skills/: the skill files the agent runs:
    • the reply skill
    • how we take care of creators
    • how Brightline sounds
    • each client's preferences
  • brands/: one brief per brand. The briefs are the source of truth for rates, sample rules and claims.

What to send back

  1. out/: one file per task, named by task ID. Each holds the action and the reply, ready to send.
  2. status.csv: one row per task, with:
    • the status: done, escalated, held or skipped
    • how it got done: the agent as is, the agent plus your fix, or by hand
    • a few words on why
  3. escalations.md: one note per person you need something from, as you'd post it on Slack. For each item: what happened, what you already know, what you need from them and by when. Marcus gets five items at most, ranked.
  4. requests.md: each mistake you found, with:
    • where it came from: data, skill or AI
    • the evidence: the file and row, or the skill and line
    • the change you'd ask for
    • then anything you'd change about how the agent runs
  5. bottleneck.md: one page on the bottleneck you'd fix this month for the whole pod, not just today's inbox:
    • what it costs today, in numbers from the folder
    • the playbook you'd write
    • how you'd train the team on it, and how you'd know it stuck
  6. A 2-minute recording (Loom or similar): show your agent running one task, then talk through the call you were least sure about today, and why you made it. Use your own voice and your own reasoning.
  7. The session export, if your agent can make one (in Claude Code, /export).

Send one link to your working folder (a shared Drive folder or a GitHub repo) and the link to your recording. Make sure anyone with the link can view both. Tell us how long you spent and what you didn't get to.

Then we'll set up a 30-minute working session. It isn't a presentation. We'll drop 10 new messages into your folder and watch you run them with your setup.