Episode 006 • February 19, 2026 • 38:24
Jay Nathan, Jeff Breunsbach
Listen on: Spotify | Apple Podcasts | YouTube
About This Episode
Jay Nathan and Jeff Breunsbach go heads-down this week — no guest, just a raw conversation about what they've each been building with Claude Code and Claude Cowork, and what it's revealing about where software, teams, and customer success are actually headed.
Jay has spent the past few weeks building custom AI-powered applications without writing a line of code: a conversational applicant tracking system, an interactive RACI builder for consulting workshops, and Balboa GPT — a client-specific AI assistant that aggregates Slack messages, call transcripts, and emails into private per-customer context buckets that the whole team can query and contribute to. Jeff has been using Claude Cowork to build a "product marketing skill" that pulls from API docs, support portals, Slack, and Notion to generate internal CS documentation on demand — talk tracks, one-pagers, and structured Notion database entries.
They also dig into Jason Lemkin's recent post on the state of SaaS — Anthropic growing from $1B to $14B in 14 months while Salesforce, HubSpot, and Adobe stocks fall — and push back on the doom narrative. Plus: why token costs need a line item in your budget now, and why the real AI advantage belongs to whoever can solve the context problem.
Key Takeaways
Get hands-on with AI or you won't fully understand what's happening. Jay is direct: if you're in leadership and not personally using tools like Claude Code or Cursor, you're not grasping the shift. The gap between idea and execution has never been smaller — but you have to feel that yourself to believe it. Pay $150 for a month. Just do it.
Enabling your team is a different problem than being productive yourself. Jeff flags the pattern he keeps seeing — and admits to doing himself — of discovering an exciting AI tool and then schlepping it over the fence to the team. Change management still applies. Help people get on the change curve, don't just point them at a new tool and expect momentum.
Established SaaS companies aren't dead — they just need to act. Brand trust, enterprise security certifications, and multi-year contracts with durable, predictable cash flow give companies like Salesforce, Workday, and Cloudflare real staying power. Jay compares it to Google post-ChatGPT: the initial narrative was doom, now they're dominant. The question for big SaaS isn't survival — it's whether they'll actually innovate fast enough to catch the next wave.
Point solutions are the most at risk. CFOs are going to start asking hard questions: why are we paying $40K for a screen recorder when someone could vibe code a replacement in an afternoon? Systems of record with billing, forecasting, and compliance dependencies are much harder to rip out. But the category of "nice-to-have point tools" is in real trouble.
The juice is now worth the squeeze on things you never had time to do. Jay's RACI builder is the clearest example: it would have been a spreadsheet before. Now it's an interactive web app with role templates, real-time editing, multiple saved versions, PDF export, and the foundation of a potential marketing tool — built on a weekend, no code written. The effort threshold for this kind of work has collapsed.
It's a context problem, not just an AI problem. The real moat isn't which model you're using — it's who can pull together all the Slack threads, call transcripts, emails, and product usage data into one accessible place. Balboa GPT is Jay's early version of this: private per-client vector stores that prevent data bleeding between customer accounts, with shared conversation history so the whole team can see what's been asked and answered.
Skills democratize expertise across the team. When any CSM can run the product marketing skill and generate a fully structured feature guide without waiting on the one product marketing person, you haven't spread a role thin — you've removed a bottleneck. Jeff's next step: a QPR skill, an onboarding skill, a playbook skill. The consistency compounds too: everyone's using the same basis, not ten different approaches.
Token costs are real and need a line item in your budget. Some engineering teams are already capping AI spend at $100,000 per engineer per year. Token usage should be treated like compensation — planned for, monitored, and optimized. The tools (Cursor, Claude Code) are already building token efficiency into their agentic processes. Pay attention to this now before it becomes a surprise line item.
Chapters
00:00 - Intro and technical issues
00:09 - Getting hands-on with AI: why leadership can't just delegate it
02:48 - The SaaS market landscape: Jason Lemkin's take
08:36 - Why established SaaS companies still have real advantages
12:09 - Point solutions at risk; systems of record are defensible
16:17 - Jeff's product marketing skill in Claude Cowork
20:29 - Democratizing documentation: any CSM can now create the guide
22:48 - Feeding source code as context; building a chatbot on your own codebase
25:28 - QPR, playbook, and onboarding skill ideas
27:19 - Managing token costs and what it means for team budgets
28:38 - Jay's interactive RACI builder built with Claude Code
34:20 - Balboa GPT: client-contextual AI with shared conversation history
Mentioned in This Episode
Claude Code – Anthropic's terminal-based agentic coding tool for building applications without writing code directly
Claude Cowork – Anthropic's AI desktop tool that connects to your files, tools, and browser for team workflows
Cursor – AI-powered code editor popular for vibe coding
ChatGPT – OpenAI's AI assistant; used as the base model layer for Balboa GPT
HubSpot – CRM and marketing automation platform
Salesforce – Enterprise CRM and software platform; referenced for Agentforce
Pendo – Product analytics and user engagement platform
Metabase – Open-source business intelligence and data visualization tool
Workday – Enterprise HR and financial management software
Cloudflare – Network performance and security platform
CrowdStrike – Cybersecurity platform
Notion – Team wiki and documentation platform; used as the output destination for Jeff's product marketing skill
Fathom – AI meeting notes and call recording tool
IGV – iShares Expanded Tech-Software Sector ETF; the fund Jay invested in as a bet on established SaaS resilience
Jason Lemkin – SaaStr founder; his post on SaaS valuation declines vs. AI growth framed the episode's market discussion
About Your Hosts
Jay Nathan – CEO of Balboa Solutions and co-founder of ChiefCustomerOfficer.io. Building custom AI-powered tools using Claude Code — including a conversational ATS, an interactive RACI builder, and Balboa GPT, a client-contextual AI assistant with shared team access and per-customer data isolation.
Jeff Breunsbach – Head of Customer Success at Junction and co-founder of ChiefCustomerOfficer.io. Built a product marketing skill in Claude Cowork that automates internal CS documentation by pulling context from API docs, support portals, Slack, and Notion — and is rolling it out to the full team this week.
