Marketer Turned Builder: How to Ship an AI Project Manager That Unblocks a Lean Team

Looking forward to this session? Share it!

Lean marketing teams are used to doing more with less, but the thing quietly capping their output usually isn't effort or capacity. It's the bottlenecks, the handoffs, approvals, and review loops where a deliverable stops moving and sits waiting on someone. Every stall compounds, and on a small team there's no slack to absorb it. 

In this session, Sabiha shows how she built an AI project manager agent in Python using the Claude Agent SDK – with no engineering team and no budget line – and the results that followed. Deliverable turnaround dropped from about three weeks to one, and review cycles shrank sharply. The team didn't get smaller. It got room to take on the work it had been turning down. This is a practical, build-it-yourself blueprint for spotting the bottleneck that's actually slowing your team, drawing the line between what an agent should own and what stays human, and understanding what a first working agent actually takes in hours and skill.

Key Takeaways:

• Identify the bottleneck that's actually slowing your lean team, by auditing where work waits instead of where your people work.
• Draw the line between what an agent should handle and what stays human, using your review loop as the test.
• Build your first working AI agent using the Claude Agent SDK (or equivalent), following the actual steps involved and a realistic view of what the effort takes in hours and skill.