Jeff and Jay get into the gap between vibe coding your own AI tools and building something your whole team can rely on. From PRD skills to master customer data files to ClickUp's "foundry" model — this episode is about what it actually takes to move from single-player AI to enterprise AI, and why slowing down now might be the fastest path forward.

KEY TAKEAWAYS

• PRDs as AI bumpers: A PRD skill forces you to define goals, non-goals, design constraints, and integrations before building — dramatically improving what AI produces.
• Single player vs. multiplayer AI: Personal tools tied to your Gmail account vanish when you leave. Enterprise AI requires shared data layers, authentication, and context.
• MCP vs. curated data: MCPs let you pull from systems in real time, but without a clean master data set, everyone queries the same raw sources and gets different answers.
• The master customer file: One canonical database table of active customers is more token-efficient and reliable than re-deriving data every time an agent runs.
• The foundry model: ClickUp's internal team builds core agentic infrastructure and proliferates learnings org-wide — more than a center of excellence, it actually ships.
• Embed, don't advise: A head of AI sitting in a room advising doesn't work. AI expertise has to work shoulder-to-shoulder with domain experts to build anything real.
• Slow down to speed up: Individual token spend gets you ~15% better. Enterprise data infrastructure + agents unlocks step-function improvement — but requires investing in the foundation first.
• Sell outcomes, not automation: The future is owning an end-to-end outcome (like Fin's "resolutions") and pricing on delivery — not just automating what already exists.

CHAPTERS

• 00:01 - Welcome & World Cup check-in
• 02:35 - The PRD idea: vibe coding needs structure
• 05:59 - Vibe coding vs. production-ready engineering
• 08:00 - Single player AI vs. enterprise multiplayer
• 10:11 - MCP vs. curated data layers
• 15:12 - Master customer data files and token efficiency
• 18:25 - Jeff's PRD skill in action
• 20:57 - Generating tasks from the PRD
• 25:20 - How enterprises are structuring AI teams
• 33:29 - ClickUp's foundry model
• 36:36 - Why infrastructure beats individual token spend
• 39:18 - The ROI problem with AI investment
• 40:42 - AI-native services: selling outcomes
• 43:29 - Wrap up & Uncommon AI community update

About the Show: Chief Customer Officer Podcast is a show about real strategies for customer-led growth in the AI era—from leaders actually executing, not just talking about it.

Your Hosts:

• Jay Nathan – CEO of Balboa Solutions and Co-Founder of ChiefCustomerOfficer.io
• Jeff Breunsbach – Head of Customer Success at Junction and Co-Founder of ChiefCustomerOfficer.io

Jay and Jeff are back with a live build episode — two operators comparing notes on what's actually working with AI. From Jay's Agent Command Center at Balboa to Jeff's Linear task ingestion system and a viral VS Code ad hack, this one's packed with real examples. Plus: why the moat in AI is attention, not technology.

KEY TAKEAWAYS

• AI slop is a real leadership problem: Unedited Claude output is hitting inboxes everywhere. Jeff catches CSM candidates submitting unmodified hiring exercises. Fix: build a "fingerprints on it" culture before anything leaves your hands.
• The Minto Pyramid cuts bloat: Conclusion first, arguments second, details last. Jeff built this as a Claude Cowork skill his team runs before any doc goes to leadership or a customer.
• Agent Command Center over vendor lock-in: Jay's team built their own agent studio instead of using Azure, AWS, or Google — to control business logic, stay model-agnostic, and keep company secret sauce off a vendor platform.
• Models are becoming commodities: The real value is the harness layer — business logic, data connections, process knowledge. Erratic model companies can't be your foundation.
• Agents fill the gap tools never could: Jeff's Claude Code system surfaces emails and Slacks, confirms tasks, and auto-creates Linear tickets — removing the capture burden entirely.
• Show and tell beats mandates: Friday demo sessions at Balboa where team members show what they built create pull, not push.
• Treat AI work like a product backlog: Groom a pipeline of AI projects, sequence by value and dependencies — don't just experiment randomly.
• Attention is the real moat: kickbacks.ai can be copied in hours. The founder's following and first-mover gravity can't be.

CHAPTERS

• 00:00 - Intro & new baby update
• 02:30 - kickbacks.ai: the VS Code ad hack
• 09:00 - Attention is the moat, not the tech
• 12:00 - Claude Cowork as a paternity leave to-do list
• 16:30 - The AI slop problem hitting leadership inboxes
• 19:00 - The CSM hiring fingerprints test
• 21:30 - The Minto Pyramid as a team skill
• 25:00 - True personalization vs. segmentation
• 28:30 - Jeff's Linear task ingestion agent
• 31:00 - Jay's Agent Command Center at Balboa
• 37:00 - Build vs. buy: why they went custom
• 39:30 - Models as commodities, harness layer as moat
• 43:30 - Keeping AI momentum inside your team
• 46:00 - AI work as a product backlog

About the Show: Chief Customer Officer Podcast is a show about real strategies for customer-led growth in the AI era—from leaders actually executing, not just talking about it.

Your Hosts:

• Jay Nathan – CEO of Balboa Solutions and Co-Founder of ChiefCustomerOfficer.io
• Jeff Breunsbach – Head of Customer Success at Junction and Co-Founder of ChiefCustomerOfficer.io

Jay and Jeff kick off the show by getting into the real stuff: managing agent sprawl, why most teams aren't ready for multiplayer AI, and whether tech layoffs actually have anything to do with AI efficiency. Unfiltered and practical.

KEY TAKEAWAYS

• Agent Sprawl Is Everyone's Problem: Agents are spinning up in every tool—Planhat, HubSpot, Gainsight, Claude. Without a team-level agent command center, you're burning tokens on experiments nobody's watching.
• Single Player vs. Multiplayer AI: Most teams are in single-player mode—each person in their own context window. The unlock is shared agents, shared data, and shared outputs.
• Verified Data Sets = Trust + Efficiency: If your team doubts an agent's output, they revert to manual work. Pre-aggregated data builds trust and cuts token costs.
• Jevons' Paradox in Real Time: Token prices are falling, but usage is exploding. Total AI spend is going up, not down.
• Model Matching Matters: Don't run a daily briefing on Opus. Use Haiku for simple tasks; save big models for high-value work.
• Rolling Out AI Right: Canva gave 5,000 employees a week to learn AI—they froze. Fix: verify tools and data before the hackathon, then let people explore.
• Layoffs Aren't What They Seem: Companies citing "AI efficiency" for cuts are mostly rationalizing. Engineering hiring is up.
• Every Job Is Changing: The highest-paid ops role will be the AI agent builder. Lean in or get left behind.

CHAPTERS

• 00:00 - Intro & Jeff's baby is coming
• 00:53 - NanoClaw: Secure open-source personal agents
• 03:47 - Meet Maverick, Jay's AI podcast producer
• 05:11 - Agent sprawl and the containment problem
• 06:20 - Building a team-level agent command center
• 13:44 - Token costs, Jevons' paradox & model matching
• 17:07 - Data centers, energy, and the physical bottleneck
• 20:44 - How to roll AI out to teams (the Canva lesson)
• 22:58 - Verified data sets: Why trust and efficiency go together
• 35:02 - Sierra AI: $15.8B valuation, 100x revenue
• 38:35 - AI in customer support: Back-end before front-end
• 41:16 - ClickUp layoffs and the 10x vs. 100x mindset
• 42:18 - Tech layoffs: Is AI really the reason?
• 45:19 - Every job is changing—lean into it

About the Show: Chief Customer Officer Podcast is a show about real strategies for customer-led growth in the AI era—from leaders actually executing, not just talking about it.

Your Hosts:

• Jay Nathan – CEO of Balboa Solutions and Co-Founder of ChiefCustomerOfficer.io
• Jeff Breunsbach – Head of Customer Success at Junction and Co-Founder of ChiefCustomerOfficer.io