Home/News/A GitHub marketing lead runs event ops as code with Copilot and Actions

AI

A GitHub marketing lead runs event ops as code with Copilot and Actions

2026-09-11 AIHOT 精选 AI
A GitHub marketing lead runs event ops as code with Copilot and Actions
示意图(AI 生成,非现场图片)

GitHub published a walkthrough from Tomoko Tanaka, its regional marketing lead for Japan and Korea, who moved her entire event operation into a GitHub repository: issue forms to capture the brief, labels to trigger Actions, and Copilot to execute runbooks written in Markdown. She wrote almost no code herself.

Key facts

  • When: A published case study built on features that already exist. You can copy it today
  • Who it hits: Solo operators running several lanes at once, especially for work like launches, events and seasonal promos that is nearly the same every time yet gets rebuilt from scratch every time
  • What to do: Pick your single most repetitive process and try it once. Write the runbook first, automate second

What happened

The workflow runs in three phases. Planning: Copilot drafts campaign names, email templates and event details from team runbooks stored in an AGENTS.md file, and she reviews and approves before anything proceeds. Registration: applying a single event-setup label fires GitHub Actions that duplicate landing pages, generate UTM-tagged links, create invitation emails, populate project boards and screen registrants daily against set criteria. Follow-up: two slash commands close it out — /lead-upload formats attendee data for CRM import, /event-report produces the attendance summary. The pieces are Issues and Issue Forms, Actions, Copilot in the CLI and the app, Copilot Agent Skills written as Markdown, Projects, and a DRY_RUN repository variable for safe testing. On results, roughly "a couple of days" of manual assembly collapses into minutes, and the post-event work she calls "the worst part" now takes two commands. Her own summary of the split: "GitHub Copilot drafts; I decide."

What it means for independent store owners

The value here is not GitHub. It is that she puts the steps of AI automation in the opposite order from how most people attempt it. The usual approach is to ask the model to do the work, discover you re-explain the context every single time, and give up. Her order is: write down the process that already lives in your head and your chat history as one Markdown document, then let the model execute against it. For a store, the documents worth writing first are the obvious repeat offenders — what actually changes when you launch a SKU (product page, collection, internal links, structured data, social teaser), what a promotion requires (landing page, discount code, three emails, ad creative sizes, the post-mortem you always skip), and the pre-peak-season checklist. You do these quarterly and reconstruct them from memory every time. Writing them down costs little, and it is what makes the work handable to a model or a new hire. Two details are worth stealing outright. One is the DRY_RUN switch: anything that sends outward — emails, live page edits, social posts — should have a print-only mode before it has a run mode. The other is the drafts-versus-decides split, which matters most wherever price, stock commitments or public statements are involved. Do not let those run end to end.

Someone should be watching these changes for you

I run AI automation for independent stores — content, SEO/GEO, product listing, social, and keeping up with policy shifts like this one.

See how it works