Episode 014 • April 23, 2026

Listen on: Spotify | Apple Podcasts

The CSM role is splitting in two. Chad Hornfeld's framing from Avoca — TAM + forward-deployed engineer, with commercial motion lived elsewhere — lines up with what Jay and Jeff are seeing across their own portfolios. The argument: asking one person to go deep technically and carry a number produces mediocrity at both. The work doesn't disappear; it gets disaggregated and specialized.

Technical fluency is the new table stakes for CSMs. Jeff is hiring for two things at Junction: technical aptitude and commercial instincts. In interviews he expects candidates to have used AI tools to build something real, and to have read Junction's public API docs well enough to teach him something back. "Junction is the healthcare data infrastructure company" is not the answer he's looking for.

Onboarding should follow the customer's journey, not the product. Junction is replacing a product-tour onboarding with a patient-journey map — defining exactly when a customer should hit specific API endpoints across a patient's lifecycle. The payoff Jeff is already seeing: fewer support tickets caused by configurations that the product allows but that aren't a best practice.

Internal apps are the new dashboards. Jay built a fully functional HubSpot pipeline dashboard — with heat maps for deal health, overdue flags, and two-way edits back into HubSpot — in a weekend using Claude Code. He didn't need to pay HubSpot for the reporting module. The broader lesson: once a department can build its own tooling, the role of traditional BI (Looker, Tableau) and of expensive vendor add-ons starts to look different.

Security is still engineering. Jay cites that roughly 45% of Claude-written code ships with significant security vulnerabilities. The answer isn't to stop building — it's to set up a central "AI Center of Excellence" that puts guardrails around deployment, data handling, and certifications while still letting every department move quickly.

Claude is winning the B2B market. Jeff's theory: Anthropic nailed the naming (Chat for casual users, Cowork for non-technical knowledge workers, Code for engineers), and an early, heavy investment in Claude Code built a flywheel. OpenAI, meanwhile, is optimizing for B2C scale at a billion users and burning through the largest investment round in history. Jay predicts Chat/Cowork/Code ultimately converge into one product called Claude.

Jevons Paradox, not doomerism. Jay revives the telephone-operator analogy: 800,000 U.S. jobs disappeared when switches replaced operators, and the broader economy exploded because long-distance calling became cheap. Expect AI to unlock a wave of product management and engineering roles in non-technical industries that could never previously justify the investment.

Avoca – AI voice platform for home services; Chad Hornfeld leads customer success there.

Claude Code – Anthropic's agentic coding tool, used by Jay to build the HubSpot dashboard.

Claude (Cowork) – Anthropic's desktop app for non-technical knowledge workers.

HubSpot – CRM platform; now "headless" with most configuration accessible via APIs.

Salesforce – CRM platform; announced a fully headless, API-first posture this week.

Pendo – Product analytics platform where Jay has colleagues building AI tooling.

Looker / Tableau – Incumbent BI tools discussed in light of custom Claude-built dashboards.

Planhat – Customer success platform Jeff's team uses for renewal and expansion data.

Junction – Healthcare data infrastructure company where Jeff leads customer success.

Balboa Solutions – Jay's firm, working with SaaS companies on AI-driven operations.

Chad Hornfeld's Avoca post – The LinkedIn post that anchored the CSM-role conversation.