In the fast-paced world of regional marketing, where the difference between a successful campaign and a missed opportunity often comes down to the precision of administrative tasks, Tomoko Tanaka, a regional marketing lead for GitHub in Japan and Korea, has found a solution to the "manual labor" trap. By leveraging her background as a former engineer and utilizing modern AI-driven tools, Tanaka has moved beyond traditional marketing automation, creating a self-sustaining system that manages event planning, execution, and reporting through the very same pipelines used by software developers.
For marketing teams, events like webinars, community meetups, and executive sessions serve as the heartbeat of engagement. However, the process of executing these events—coordinating schedules, managing registration lists, sending invitations, and generating post-event reports—is often fraught with repetitive, manual tasks. These duties, while essential, are prone to human error, such as mistyping a campaign name, using an incorrect link, or failing to update downstream reports. Recognizing this, Tanaka realized that if she could articulate her workflow as a set of instructions, she could, in effect, turn that workflow into code.
Treating an Event as a Project
Tanaka’s approach centers on the philosophy that a marketing event should be treated with the same rigor as a software project. At GitHub, the team already utilized the practice of opening a single GitHub Issue for each project, serving as a centralized hub for planning, discussion, and status tracking. By transforming these issues into actionable units of work, Tanaka successfully integrated her marketing operations into the existing developer infrastructure.
This system relies on three core GitHub primitives: issues for planning, labels for workflow state, and Actions for execution. By housing the marketing workflow within a repository, the team gained access to the benefits of version control: comprehensive history, transparency, reviewability, and a permanent URL for every decision made.
A critical requirement for this level of automation is accessibility. Whether through an API or a command-line interface (CLI), the software tools used by the marketing team—from event platforms to CRMs—must be scriptable. Tanaka notes that even when an official API is absent, a CLI can often bridge the gap, as many tools can handle authentication through existing browser sessions. This realization is a game-changer for marketers across various disciplines: if your tools allow for programmatic input, your repetitive workflows are prime candidates for automation.
While some might argue that dedicated marketing automation platforms are the standard solution, Tanaka points out the limitations of such tools in diverse markets. The Asia-Pacific (APAC) region is not a monolith; it is a complex collection of distinct markets, each with unique linguistic, cultural, and operational requirements. A "one-size-fits-all" software solution often requires expensive consulting hours and lengthy development roadmaps to accommodate these variations. By building a custom workflow within GitHub, the team ensures that any change—whether it’s a shift in regional requirements or a new definition of a "lead"—is simply a pull request away. This allows for agility and precision that packaged software often struggles to provide.
The Role of Conversation in Planning
The automation pipeline begins long before a project is technically underway. By using GitHub Copilot, Tanaka initiates the planning phase through natural language, such as stating her intent to host a webinar on a specific topic. This process is governed by a file named AGENTS.md located at the repository’s root. This document serves as a team runbook, codifying essential details like campaign naming conventions, fiscal calendar mappings, regional time zones, and email templates.
GitHub Copilot reads this runbook, identifies relevant historical data, and drafts the necessary materials, while simultaneously prompting Tanaka for any missing information. This conversational approach strikes a delicate balance between total automation and human oversight. While fully automated systems can be rigid and inflexible, and fully manual systems are prone to human error, this hybrid model allows the AI to handle the heavy lifting while ensuring that a human makes the final decision on every critical detail.
This interaction, once confined to the terminal via the GitHub Copilot CLI, has evolved into a more accessible experience through the GitHub Copilot app. This shift in accessibility has allowed more team members to engage with the automated workflow, lowering the barrier to entry from "terminal-proficient" to simply being able to type.
Execution Through Automation
Once a project is initialized and the appropriate label—such as event-setup—is applied to an issue, the system takes over. GitHub Actions automatically handles tasks that previously consumed nearly a full day of labor, including the creation of registration pages, the drafting of email sequences, and the synchronization of data across systems.
Registration screening, another time-consuming task, is managed via a cron-triggered workflow. This routine fetches new registrants daily, cleans the data, and screens participants against specific criteria for invite-only events. This ensures that the team can focus on the strategic aspects of the event rather than the mundane verification process.
A key feature of this system is the DRY_RUN switch, a repository variable that acts as a rehearsal mode. By enabling this, the team can test the entire pipeline without actually interacting with external systems. This safety mechanism is essential for experimentation, allowing marketers to refine their workflows without the fear of causing accidental disruptions or errors in live data.
Post-Event Reporting and Governance
The conclusion of an event was historically the most labor-intensive phase, involving the export and reformatting of attendee data and the manual creation of performance reports. Today, this is handled through "slash commands" that trigger specific GitHub Copilot agent skills. These skills are essentially written procedures, formatted in Markdown, that outline exactly how to fetch metrics, shape data for CRM uploads, and generate reports directly within the event’s issue.
Because these skills are defined in prose, they remain highly adaptable. Different regional teams can modify their specific runbooks to reflect local conventions and requirements without disrupting the core infrastructure. These skills are treated like code: they are submitted via pull requests and subject to review by team maintainers, ensuring that marketing automation is governed by the same rigorous standards as software development.
Guardrails and Continuous Improvement
Integrating customer data and API credentials into an automated system requires robust security, and Tanaka emphasizes that many of the necessary guardrails were already embedded within the platform. By leveraging standard developer practices—such as code reviews, test suites, and strict access controls—the marketing team has built a secure, transparent, and resilient system.
However, the experience has not been without its learning moments. Tanaka notes a failure in the registration screening workflow that went unnoticed for several days, highlighting the necessity of robust monitoring. She stresses that automation without oversight is a liability; therefore, any scheduled task must have a mechanism to report errors or anomalies effectively.
For those looking to adopt a similar strategy, Tanaka’s advice is simple: start with a single, repetitive task. By identifying a process that touches tools with an API or CLI and building a small, manageable version of an automated workflow, marketers can begin to reclaim their time. As she suggests, if you can write down how you perform your work, you have the foundation to automate it. By moving from manual labor to an automated, code-like workflow, Tanaka has shown that the future of marketing isn’t just about using tools—it’s about designing systems that allow us to focus on the human decisions that truly matter.
