Episode 011 • April 2, 2026

Listen on: Spotify | Apple Podcasts

Most companies are still in early adopter territory. Jeff attended a room of 60 CS leaders in Boston and walked away with a calibrating realization: even people who talk about AI daily are likely further ahead than the vast majority of companies. The adoption curve from Crossing the Chasm still applies — there's a significant majority that is evaluating, not acting. The actionable implication: small, focused workflow changes now will compound into real organizational advantage.

You can't automate a process you haven't documented. Both Jay and Jeff land on the same diagnosis: before you can leverage AI in your workflows, you have to actually understand what those workflows are. Jay made the point that building a Claude skill forces you to articulate exactly what you're trying to do on a repeatable basis. Jeff's takeaway from the Boston event was similar — he's planning to map every CS process in Figma so the team can actually see and inspect where technology could be inserted.

"Just use AI" is not an AI adoption strategy. The companies at Jeff's event struggling with team-wide adoption shared a common pattern: a top-down mandate with no specificity. What works instead is a three-part approach — demonstrating AI behaviors at every layer of the organization (CEO included), pointing teams toward 2–3 specific problems rather than open-ended exploration, and creating regular show-and-tell moments where people can see what colleagues are actually building.

Call transcripts are your company's most valuable untapped asset. Jay called Fathom recordings "the most valuable thing we have in our company" — and both hosts are actively building pipelines around them. Jeff is exploring an automation that reads Fathom transcripts for customer feature requests, then auto-generates a Linear ticket in PRD format with the relevant call timestamp attached. Jay is using transcripts as the primary source for website content, knowledge base articles, and client-facing artifacts.

Claude skills turn individual expertise into team infrastructure. Jay built an executive readout coaching skill that captures his own feedback patterns and makes them available to every member of his Balboa team. Instead of prepping a deck together, prep calls can now focus entirely on engagement strategy — who plays what role, what relationships to activate. Jeff built a parallel product marketing skill that takes Notion source content and generates a branded, customer-facing one-pager without the team ever touching a design tool.

The CSM of the future is technically fluent. Multiple CS leaders at Jeff's Boston event flagged the same shift: the role is converging toward something closer to a technical account manager. Jeff is already incorporating technical curiosity into his hiring rubric — asking candidates to explain a complex product simply, and asking what they've done with AI outside of work. Jay's take: technical fluency shouldn't need to be in the job posting. If someone isn't curious enough to use AI in their personal life, that tells you something.

Forward deployed engineering could redefine voice of customer. Jay made the case that giving CSMs a local build of a product — and the ability to spin up a feature-based pull request from a customer conversation, reviewed and refined before submission — is entirely possible today. Jeff's friend Spencer is already doing this at his startup, using Linear to generate dev environments from PRD-style tickets. The companies executing on this are capturing customer insight at a scale and speed that wasn't conceivable two years ago.

A center of excellence is how you sustain AI adoption at scale. Rather than an ongoing mandate, Jay describes the center of excellence model as a small group of people embedded with executive sponsorship — vetting and standardizing tooling, running discovery sessions and show-and-tells, pushing teams toward maximal use rather than individual use, and measuring outcomes. It doesn't need to be formal. It just needs to be intentional.

Planhat – Customer success platform that organized the 60-person CS leader event in Boston

Claude (Anthropic) – AI assistant platform used by both hosts; team version includes organizational skill sharing

Claude Code – Anthropic's coding tool, used by Jay to rebuild the Balboa Solutions website and generate HTML glossary pages

Fathom – AI meeting recorder and transcript tool referenced as the source for call transcript automation

Linear – Project management tool referenced for auto-generating PRD-style feature request tickets from call transcripts

Notion – Internal documentation tool; Jeff's team uses it as source content for the product marketing skill

Figma – Design tool Jeff is exploring for visualizing and mapping CS workflows

Balboa Solutions – Jay Nathan's consulting firm; rebuilt website and partner portal using Claude Code

Junction – Jeff Breunsbach's company; building Claude skills for internal CS team use

EP008: Cut Your Own Prices Before Someone Else Does — The self-disruption thesis: why SaaS companies need to move before AI does it for them. Pairs directly with this episode's data-layer argument.

EP006: The SaaS Shift — What We're Actually Building with AI — Jay and Jeff on the AI tools they're actively building with, including early MCP workflows and the Balboa GPT.

EP005: Building Predictive Customer Success with AI w/ Justin Chappell — Justin Chappell on moving from reactive to predictive CS — and the data infrastructure that makes it possible.