EduCampaign

• Inspiration

I have spent 15 years in search marketing working on some of the largest accounts in the world. Billion dollar budgets. Enterprise clients. Highly regulated industries. Teams of strategists.

My job was simple on paper. Build performance systems that scale.

In reality, I was managing people doing work that should not take that long. Campaign builds that dragged on for weeks, even months. Keyword research that was manual. Ad copy that was repetitive. Structure that depended more on human bandwidth than intelligence.

Then AI showed up.

The first thing it replaced was not strategy. It was execution.

The exact layer I had built teams around.

Over time, I realized something uncomfortable. The job I had been doing for years could be compressed into minutes. Eventually, I was not just replacing junior strategists. I was being replaced too.

This project is my response to that.

Not resisting it. Building with it.

• What it does

EduCampaign AI is an intelligent Google Ads campaign builder designed specifically for higher education.

In under 60 seconds, it takes a university’s enrollment goals, target programs, and audience parameters and generates a complete, publish ready search campaign.

That includes: • Campaign structure • Themed ad groups • Keyword lists with match types • Ad headlines and descriptions • Extensions • Budget framework

Then it pushes everything directly into Google Ads via API.

This is not a mockup. It is a working system.

At a time when universities are under resourced and overpaying agencies, EduCampaign AI gives them a director level SEM strategist on demand at near zero marginal cost.

• How I built it

This is where it gets interesting.

I built the system as a multi stage agentic pipeline using the Jaseci framework, which is based on Object Spatial Programming.

Each part of the campaign creation process is its own node: • Keyword research • Campaign structuring • Ad group clustering • Copy generation • Publishing

These nodes operate independently but pass context to each other in sequence.

Claude acts as the reasoning layer. It interprets the campaign brief and orchestrates decisions across the pipeline.

The Google Ads API is accessed through a Model Context Protocol server, which allows Claude to translate intent directly into executable API actions.

The full architecture looks like this: • Interface: Claude AI • Orchestration: Jaseci agent pipeline • Execution layer: MCP server • Data source: Google Ads API

Everything is auditable. Every step can be paused, reviewed, or overridden.

This is not just automation. It is structured intelligence.

The system I built aligns directly with the design outlined in my Google Ads API integration spec , where campaign creation, GAQL querying, and performance feedback loops are all exposed as callable operations within the agent pipeline.

• Challenges I ran into

The biggest challenge was not technical. It was architectural.

There is no out of the box Google Ads MCP server that just works. I had to bridge that layer myself and map real world SEM logic into callable tools.

Search marketing is full of nuance: • Match type strategy • Query intent segmentation • Budget distribution • Account hierarchy decisions

Getting an AI system to respect those constraints required designing guardrails, not just prompts.

Another challenge was resisting scope creep. Performance optimization, bidding strategies, and cross channel attribution are all possible, but I focused strictly on search campaign generation to make the system demonstrable and real.

• Accomplishments that I’m proud of

I built a fully functional, end to end campaign generation system in 24 hours.

Not a prototype. Not a concept.

A system that: • Takes a real brief • Builds a structured campaign • Pushes it live into Google Ads

I translated 15 years of institutional knowledge into a system that can execute instantly.

More importantly, I proved that the future of SEM is not tools. It is architecture.

• What I learned

AI does not replace strategy. It replaces delay.

The bottleneck in marketing was never ideas. It was execution speed and coordination.

By breaking campaign construction into discrete, intelligent nodes, I can move faster while increasing control and transparency.

I also learned that domain expertise still matters. AI without structure produces noise. AI with structure produces leverage.

• What’s next for EduCampaign

This is just the starting point.

Next steps include: • Performance feedback loops using real campaign data •. Microsoft Network • Budget optimization and bid strategy integration • Multi campaign orchestration across programs and geographies • UI layer for university marketing teams • Expansion beyond search into full funnel campaign systems

Long term, this becomes more than a campaign builder.

It becomes infrastructure for how marketing gets done.

Final note

This project is personal. I spent years building systems around human execution. Then I watched that layer disappear. EduCampaign AI is not about replacing people. It is about rebuilding the role of the strategist.

From operator to architect. From executor to system designer. This is version 2.0.

Built With

• claude
• google-ads
• google-cloud
• gpt
• jacesi