Agentic Engineering: What Happens When AI Joins the Dev Team?

What happens when AI stops being a tool engineers use and starts becoming a member of the development team? In this episode of Ship Happens, host Per Krogslund sits down with Scott Gale, CTO and co-founder of Fluency, to explore the rise of agentic engineering and how AI is changing the way software gets built, tested, deployed, and maintained. Fluency has spent years automating digital advertising across platforms including Google, Meta, Amazon, TikTok, and DSPs. Now, the company is bringing AI agents into that workflow—using them to advise customers, enforce compliance, surface optimization opportunities, and help engineers tackle software development itself. Scott walks Per through what that looks like in practice. At Fluency, engineers can use agents to triage bugs, investigate problems, propose fixes, and open pull requests. But the agents don't get the final say. Humans review changes before they make it into production. That human layer becomes especially important when you're operating systems responsible for roughly $3 billion in annual ad spend. Scott explains why agentic engineering requires more than powerful models: it requires strong DevOps foundations, clear guardrails, reliable processes, and a thoughtful approach to what should—and shouldn't—be automated. The conversation also tackles the bigger questions facing engineering teams: If AI can write more of the code, what should engineers spend their time doing? When should you build versus buy? And as AI makes coding easier, does the real engineering challenge shift toward judgment, architecture, and directing intelligent systems? AI may be joining the dev team. The question is what the rest of the team does differently because it's there. What You'll Learn: • What “agentic engineering” looks like in a real production environment • How AI agents can become part of an engineering team's workflow • How Fluency moved from deterministic templates toward AI-powered agents • What happens when agents triage bugs and propose code changes • Why human review remains critical for production systems • How engineers' roles change when AI can write and modify code • Why deterministic processes still matter in agentic systems • The DevOps foundations needed to safely deploy AI agents • How to design effective guardrails around autonomous workflows • How to think about build-versus-buy decisions in an AI-first world • Why token costs and long-term ownership are becoming important engineering considerations • Whether engineers should still be writing code—and what they may increasingly do instead Episode Chapters 00:00 When AI Code Fails 01:03 Meet Fluency CTO Scott Gale 02:30 Fluency Origins 06:06 From Templates to Agents 07:42 Human Control vs. Autopilot 10:00 Automation vs. Persuasion 11:27 Engineers Managing Agents 17:33 Guardrails and DevOps Layers 28:00 Build, Buy, Standardize 32:11 What Gets Expensive Now 33:29 Should You Still Code? 35:24 Closing