Episode 009 • Thursday March 12, 2026 • 47:24
Listen on: Spotify | Apple Podcasts | YouTube
About This Episode
No guest this week — Jay Nathan and Jeff Breunsbach dig into what it actually means to be AI-first, and the answer is less philosophical than you'd expect. The conversation opens with a job candidate who sent Jay an app she built in Lovable before the interview — unprompted — and why that's now the bar for standing out. Then it moves fast: the debate over centralizing AI adoption vs. letting it run organically, three go-to-market mistakes leaders are making right now, and Jay building an AI vulnerability matrix live on the call.
The matrix is the centerpiece of the episode. Pulled from a Tomas Tunguz piece, Jay mapped every SaaS company into four quadrants based on how complex their solution is to adopt and how hard it is to replicate. The categories — sitting ducks, protected niches, targets, and fortresses — aren't just theoretical. Jay argues that knowing which one you're in should drive your entire strategy: how you price, how you hire, how fast you ship. He built the visual in Claude mid-conversation and it's worth publishing.
The episode closes with a genuine back-and-forth on whether CSMs are being replaced or transformed. Jeff pushes back on the forward deployed engineer narrative — tools don't work end-to-end yet, private companies are hard to research with AI, and the near-term advantage for CS is leaning harder into relationships, not lighter. Jay agrees, but adds: agent-to-agent purchasing is already happening for simple products, whether we're ready for it or not.
Key Takeaways
The best way to prove you're AI-first is to build something. A candidate for Jay's team at Balboa built an app in Lovable and sent it before the interview. Not a resume. Not a cover letter. A working product relevant to what they'd be doing. Jay was immediately impressed. Jeff's framing: you don't have to go that far — a Loom walkthrough, a workflow document, anything that shows you've put hands on the tools. The point is to stop playing by rules everyone else is following.
Centralize AI where it touches systems of record; let everything else run organically. Jeff's framework from his Spring Health days: when AI tools are plugged into your CRM, customer data, or core platforms via MCP, they need governance, access controls, and oversight. Departmental tools like Gamma (AI presentation builder) don't need the same treatment. The error most companies make is applying one policy to all of it — either locking everything down or opening everything up.
The three go-to-market AI mistakes (Kyle Lacy). First: automating bad processes at scale — if the process was broken before, AI makes it faster and worse. Second: building on generic best practices instead of your own data — if you have every customer call recorded, you have a proprietary corpus that beats any prompt engineering trick. Third: prioritizing internal efficiency over the buyer experience — the bad agentic support experiences Jay and Jeff both had recently are the direct result of this.
The AI Vulnerability Matrix — four quadrants every SaaS leader should know. Jay mapped companies on two axes: how complex is the solution for the customer to adopt, and how hard is it to replicate? Sitting ducks (low complexity, easy to replicate) face the most immediate AI disruption — screen recording tools, basic workflow automation. Protected niches (low complexity, hard to replicate) have data moats like Apollo or ZoomInfo. Targets (high complexity, easy to replicate) are enterprise CRM and ERP — where AI-native entrants are attacking. Fortresses (high complexity, hard to replicate) have the most time. Knowing your quadrant should inform every strategic decision you make this year.
The $700K ARR per employee benchmark changes how you should think about headcount. The old rule of thumb was roughly $200K ARR per employee. AI-first companies are pushing $700K, $800K, and beyond. Jay's take: if you're still staffing at the old ratio, a leaner competitor is already undercutting you on cost structure. This isn't about layoffs — it's about what you're building toward and whether your operating model is designed for where the market is going.
Humans stay in the CS loop longer than the hype suggests — but what the job requires changes fundamentally. Jeff's counterpoint to the "forward deployed engineer" narrative: end-to-end AI replacement isn't working today, tools are being forward-sold on capabilities they don't yet have, and private companies are nearly impossible to fully research with AI alone. His prescription: lean harder into relationships in the near term. Be so embedded in the customer's business that you can recite their strategy. Jay's counterpoint: agent-to-agent PLG is already happening for simpler products — an agent building an app will find and integrate Stripe without a human ever being in the loop. The more complex the solution, the longer the human stays relevant.
Chapters
00:01 – Welcome and personal updates: t-ball season, flu season, Jay's daughter commits to College of Charleston
02:32 – Block's 4,000 layoffs: is AI the real reason, or air cover for over-hiring?
05:12 – What Jay and Jeff are actually seeing: AI replacing parts of jobs, not whole jobs
07:32 – Jack Dorsey's thesis: smaller teams plus autonomous tools equals better outcomes
09:00 – Jeff's MCP story: mapping HubSpot to PlanHat fields in a Sunday afternoon
12:37 – What MCPs are and why they matter for every CS team
15:27 – Token costs and the new economics of AI-powered operations
16:50 – Are UIs becoming obsolete? Natural language as the future interface
17:08 – Jay's SaaS self-disruption thesis: cut your own prices before someone else does
21:15 – Adobe's "swallowing the fish" and what SaaS incumbents can learn from it
22:39 – M&A in the AI era: are acquisitions easier to integrate now?
25:39 – Palo Alto Networks' acquisition playbook: build the roadmap before you close
27:12 – Jeff's CS AI build: deal staging from call transcripts via Fathom and PlanHat
30:28 – Three advantages of AI-driven deal staging: less data entry, consistent criteria, better conversations 32:37 – Creating hundreds of renewal records in one afternoon with Claude Cowork
34:30 – Building a renewal forecast and running CSM pipeline reviews
35:10 – Future idea: an AI agent that recommends renewal pricing and deal options
36:23 – When to carve off renewals into a dedicated team (and when it's a privilege of scale)
37:40 – Tech stack consolidation: fewer tools, tighter AI integration
39:41 – The clearest definition of customer journey vs. service blueprint on this show
42:10 – Wrap-up and preview of an upcoming guest building the future CS platform
Mentioned in This Episode
Block / Square – Jack Dorsey's fintech company; laid off approximately 4,000 employees, sparking debate about AI-driven displacement vs. COVID over-hiring correction
Claude / Claude Cowork – AI assistant and desktop automation tool from Anthropic; used by both hosts for MCP-based workflows and HubSpot operations
HubSpot – CRM platform used by both hosts; central to Jeff's MCP-based pipeline and renewal automation work
PlanHat – Customer success platform with an MCP server; Jeff is using it as the hub for AI deal staging and renewal forecasting
Fathom – Call recording tool used by Jeff's team to capture transcripts that feed into the AI deal staging workflow
MCP (Model Context Protocol) – The protocol that allows LLMs to connect to and interact with external platforms like HubSpot, PlanHat, and Pendo in plain English
Monday.com – SaaS project management platform; CEO discussed on 20VC about the future of software companies in the AI era
Salesforce / AgentForce – Referenced for their aggressive AI acquisition strategy (6–7 acquisitions in recent months) and their agentic platform
Palo Alto Networks – CEO's M&A playbook discussed: build the combined product roadmap before closing the deal, structure equity to retain founders
Adobe – Referenced for their "swallowing the fish" transition from perpetual licenses to SaaS subscription as a model for self-disruption
20VC – Podcast where Jay heard the Monday.com CEO interview
All-In Podcast / Jason Calacanis – Referenced for the point about token costs becoming part of employee cost calculations
College of Charleston (CFC) – Jay's daughter committed to attend; Jeff is also a CFC grad
About Your Hosts
Jay Nathan – CEO of Balboa Solutions and co-founder of ChiefCustomerOfficer.io. Jay has spent his career leading customer-facing and product organizations at SaaS companies and is one of the most widely followed voices in customer success leadership.
Jeff Breunsbach – Head of Customer Success at Junction and co-founder of ChiefCustomerOfficer.io. Jeff is currently deploying MCP-based AI automation across his CS and revenue operations stack and building AI-powered deal staging and renewal forecasting workflows.
EP006 — The SaaS Shift: What We're Actually Building with AI — closely related, covers the same AI-in-the-trenches territory
EP005 — Building Predictive Customer Success with AI w/ Justin Chappell — AI in CS workflows, very on-theme
EP007 — The Wisest Person in Your Company You're Not Talking To w/ Jon Olson — adjacent episode from the same run
Subscribe
Chief Customer Officer Newsletter Actionable insights for senior customer success leaders. Join 4,000+ SaaS executives getting weekly strategies on retention, growth, and CS leadership. www.chiefcustomerofficer.io/subscribe
