In this episode of the Paid Search NYC Podcast, host Matt Shenton is joined by Stewart Dunlop (Founder, PPC.io) to break down what AI is actually good at today - and where it still falls short.We go beyond the hype into real workflows: how to use tools like ChatGPT and Claude Code effectively, why context is the most important input, and how marketers can start building repeatable systems using AI - not just one-off prompts.Stewart also shares a behind-the-scenes look at building PPC.io - a platform designed to run AI agents across paid media accounts — and what it takes to turn AI from a novelty into something genuinely useful.🔑 Key Insights from the Episode- Why most AI outputs are weak (and how to fix it with better context)- The difference between chat-based AI and agent-based workflows- How tools like Claude Code enable faster, more scalable execution- Why “skills” and structured instructions outperform one-off prompts- How to turn your SOPs into repeatable AI workflows- Where AI adds the most value (and where it still needs human input)- Why documentation is the biggest unlock for using AI effectively- How agencies can use AI to scale work across multiple clients- The concept of “roundtable agents” and stress-testing decisions- Why execution is becoming table stakes — and where real value shifts⏱ Key Sections & Timestamps00:00 – Intro01:30 – AI hype vs reality: where it actually works05:00 – Why context is everything in AI prompting08:30 – Improving outputs with better inputs12:00 – Chat AI vs agent workflows explained16:00 – Introduction to Claude Code20:30 – Building workflows inside VS Code26:00 – Using AI for landing pages and reporting32:00 – Where AI still falls short36:00 – The importance of documentation and SOPs40:00 – From workflows to agents: the next step 45:00 – Building PPC.io (behind the scenes)50:00 – Testing agents, prompts, and models55:00 – Multi-agent systems and “roundtable” decision making59:00 – Final thoughts on AI in marketing