The 3 episodes below are this week's deep dives; the other 2 are written up in full in the Notion archive.
Shipley's SQL-and-stage discipline (GTM Science) - the cleanest, most portable forecasting hygiene play this week, built to survive the "make a quarter, miss a quarter" whale problem where you only hit plan when one big deal lands.
The framework has three parts, each with a clear "done" test:
First step this week: pick your single SQL definition and audit current pipeline against the 90-day rule; anything stalled past 90 days gets closed or reclassified. Measure by watching forecast accuracy tighten a month out. Failure mode: overloading the CRM so reps stop entering data honestly - keep entry light (use AI call-notes to populate) or you trade one blindness for another.
The dominant thread this week is AI forcing a rethink of where governance and ownership sit in the revenue org, and three shows attack it from different seats. Dinner on The RevOps Review makes the operational case: reps will build with AI whether RevOps likes it or not, so the job is to give them a governed place to do it (read-only scopes, native Salesforce MCP sharing existing permissions, "landing zones" where agents push low-risk UI changes autonomously). Notion's team on [Un]Churned shows the same instinct on the CS side with a company-wide "app store" of rep-built agents and a hall of fame, while Yandel on LeanScale describes an internal app store at Imubit with a "listener" agent checked by a "scrutinizer" agent before anything posts to Slack. The lesson: curated enablement of bottoms-up building beats prohibition.
The second thread is consumption pricing rewriting comp and CS economics. Notion's outcomes architects now own renewals and expand via credit usage rather than seats, feeling "good pressure" to prove ROI credit-by-credit. Topline's hosts frame the risk: if your margins depend on frontier-model tokens, you're hostage to the labs' pricing, and swapping in cheaper open-weight models may degrade the product your usage depends on. Both shows agree the era of assuming 80-90% SaaS margins is over, and RevOps leaders should start instrumenting cost-to-serve per account now.
Where the shows diverge is on gross margin as a signal. Jacobs on Topline holds the hard line: "you're at negative 50% margins, you don't have a business." Zaman argues negative margins can be a rational feature bet if usage is exploding and money is flowing, with secondaries along the way. The stronger read: Jacobs is right that you need a named path off the cost curve, but Zaman is right the market currently rewards the bet, so fund the growth while instrumenting the exit from frontier-model dependency.
The three episodes worth your time this week, with the mechanics behind the takeaways above.
Shipley's diagnostic when growth stalls: teams default to blaming sales execution, but roughly half the time it's product-market fit measured badly - founders calling a revenue number "PMF" instead of checking whether customers use the product, choose it over competitors, and ask for more. His first two asks walking into a company: the best and worst sales calls (command of message, pricing, competition, and real pain show up in one call), and the founder's coachability.
The forecasting spine is SQL-and-stage discipline: one global SQL definition (recorded call showing need/pain plus budget), 90-day decay to kill "forecast furniture," and stage progression only on customer-stated evidence. To get a single version of truth, he pulled sales ops out from under sales to report directly to him for two years. He pairs this with "embedded sales enablement": enablement people listening to calls daily, compensated on making reps better.
The BlackDuck turnaround ($20M to $90M in four years) rested on three changes: ICP narrowing, a pivot toward security-driven demand, and a velocity motion built alongside enterprise, plus developing reps internally rather than buying Rolodexes. His access-to-power point: BlackDuck ran forecast-accuracy simulations split by whether reps were "at power," then trained reps to verify the contact actually held decision authority.
Notion rebuilt CS into "outcomes architecture": they stopped selling software and started selling work, co-building end-to-end workflows in workshop sessions rather than running one-to-many trainings. The mechanics worth stealing: an AI maturity model with four rungs (AI as thought partner, as personal assistant using your context, as teammate, and finally AI-as-the-system), used to diagnose what were formerly success plans, now "AI transformation plans." Each plan documents business objectives, pain points, success metrics, then a mutual action plan of specific workflows to build, drawn from a library of best-in-class workflows tagged with estimated ROI.
Workshops scale from 190-person foundational sessions down to smaller multi-agent workflow builds. Sales hands over one initial workflow plus documented objectives and stakeholders (they run MEDDPICC), and CS grows workflows from there. Underneath, nine sales teams collapsed to four functions, with RM and CSM motions merged so outcomes architects now own commercials.
On consumption pricing, the operational detail worth noting is how they prove ROI: a credit calculator translating agent usage into hours saved times hourly rate. An email agent running under 500 credits/day (roughly $5) saving about 10 hours/week is the worked example. Because pricing is model-agnostic credits, expansion no longer requires seat growth: a contraction in seats can be offset by workflow-driven credit usage, which the team frames as "no ceiling."
Dinner's governance approach: keep any LLM access to the CRM read-only and tightly scoped. As tools make this kind of access easy for anyone to set up, the governance burden drops but a new problem appears: anyone can build what they think is the best version of a workflow, creating duplication and drift.
The sharpest operational take: kill the timed two-week sprint. Requests never stop (new pricing packages, bugs, quality-of-life tweaks), so score everything and split it three ways: 100% human-led critical infrastructure, a middle zone, and "landing zones" where AI can write and push low-risk UI-layer code, with an agent asking "do you want this pushed to prod?" The reasoning: since the industry told everyone to become AI-proficient or lose their job, frontline managers showing up with AI-built tools are just doing what they were told. Give them a governed place to interact or they route around you. His frame: RevOps has always been a yes-shop-or-no-shop trap, and the new answer is a collaborative sandbox.
Full write-ups are in the Notion archive.
Topline: "If the AI Money Dries Up, Which Companies Burn?" The hosts-only format lets Jacobs, Bruno and Zaman genuinely disagree on whether negative-margin AI bets are rational, and hearing the argument unresolved - plus Bruno's candid account of his own stranded 2021 raise - carries more than any summary can.
Get the next one on Monday. Free, one email a week.