Aaron Lu is the AI Product Lead at Notion, where he's built a swarm of AI agents that now run the team's day-to-day ad operations. His agents duplicate creative across hundreds of ad sets, run LinkedIn dayparting the platform doesn't natively support, and monitor spend hour by hour to catch waste a daily report would never show.
Aaron breaks down how he builds self-learning loops where one agent evaluates and rewrites the instructions of another, why some repetitive tasks are better left to a cron job than an AI agent, and how he decides what's actually worth automating in the first place.
He also walks through the project that won Profound's Marketing Engineering Hackathon, a competitor-monitoring tool he built solo by running six or seven parallel agent sessions with one orchestrator managing the rest. He talks through what he'd do differently now, and where he thinks the job goes next as the agents keep getting better.
We discuss:
- How a marketer's job shifted from doing the work to designing AI systems
- Why a good cost-per-lead number can still hide wasted ad spend
- The swarm of AI agents running Notion's ad operations
- How he built an AI agent for LinkedIn dayparting, a feature the platform doesn't natively support
- The creative duplication agent that saves a day of work every week
- What he would do differently on his hackathon-winning project
- How to decide what to automate first
Connect with Aaron
LinkedIn: https://www.linkedin.com/in/aaron-lu/
Notion: https://www.notion.com/
Connect with Profound
Website: https://tryprofound.com
LinkedIn: https://www.linkedin.com/company/tryprofound/
Read the Marketing Engineer Manifesto: https://www.tryprofound.com/marketing-engineer
Profound University: https://university.tryprofound.com/
The Marketing Engineer is a show for practitioners figuring out in real time what it means to market at the frontier. New episodes every two weeks.
