Listen now on
Spotify . YouTube . Apple Podcasts
Lindsay Rothlisberger is Director of GTM Innovation at Zapier. Zapier was a billion-dollar company before it had a formal RevOps function. She built it, coming up through marketing ops and lifecycle, then kept going.
Her job now is the layer underneath Zapier's GTM agents: a shared brain of context, a skill that reviews other skills, and a rule that every agent action gets written back to the CRM so it can be measured against deals.
What you’ll learn
How Zapier reached billion-dollar scale product-led, then layered in sales assist and an enterprise motion that 10x'd ACV in three years
The three context layers inside Zapier's shared brain, and why every piece has exactly one named owner
The skill that reviews other skills, and what it caught: whole ICPs and strategy docs stuffed into individual prompts
Why the first call prep agent flopped, and the rule that every agent output has to end in an action
How writing agent actions back to the CRM ties agents to deals, and the renewal sprint that lifted on-time renewals more than 30%
Episode Highlights
01:04 - A billion-dollar company before formal RevOps
10:02 - Three buckets of context, one owner each
12:52 - The skill that reviews the skills
15:01 - When a shared brain is worth building
17:22 - The four-week sprint she would not repeat
26:47 - Enablement is becoming infrastructure
29:08 - The Monday digest and what gets ignored
32:36 - The deck builder the whole sales team adopted
33:30 - The call prep agent that flopped
35:44 - Writing every agent action back to the CRM
36:12 - Closed-lost went from ghosted to real intel
36:45 - The LinkedIn-connection outbound agent
39:24 - One sprint, one metric, renewals up 30%
40:39 - The 10x ACV climb upmarket
42:45 - What to change this week
Key Takeaways
1. A billion dollars first, RevOps second
Zapier's flywheel ran product-led for years. RevOps, sales assist, and product-led sales got built after the company was already at billion-dollar scale, once customers started asking for hands-on help. Pricing moved off pure usage iteratively as dedicated CSMs and build time entered the deal, and ACV went up 10x in three years.
Stable context like company strategy, ICP, and brand, where the ICP has one named owner in product marketing. Operational context like first-call enablement, upcoming events, and approved customer stories. And the data layer, centralized in Databricks and reached over MCP. Every piece has a person who knows they own it.
3. Start with the agents, not the wiki
They ran a four-week move-in sprint converting existing docs to Markdown, and her review is blunt: she would not do it again. The version that works is picking the agents you want to run, asking what context makes them good, and building only that.
4. A skill that reviews the skills
The whole shared brain started because a RevOps teammate built a lightweight review, literally another skill, that checks structure and data policy before anything gets marked gold star and shared. It caught people stuffing entire ICPs and strategy docs into prompts, where the context lives and dies with the skill. What fails review is almost always structure, not data policy.
5. Every output has to end in an action
The first call prep agent flopped: generic summaries, no action to take, no way to measure it. The fix was generating the deck and the pre-call note instead of describing the customer.
"If your context isn't strong and it doesn't have a unique point of view, any output that you create with it is just gonna sound like AI slop."
6. Write every agent action back to the CRM
A Gong extraction agent replaced rep-entered closed-lost reasons, which used to read as everyone ghosting them, with competitor mentions and use cases. A connection agent finds who at Zapier knows someone on an open deal, drafts the outreach, and logs the send. Because every action lands in the CRM, they can run one sprint against one metric, and the renewal sprint lifted on-time renewals more than 30%.
What she's over, and the one thing she wants you to change this week, is in the episode.
If you're enjoying these episodes, reply and let me know what you'd love to learn more about: shared context, skill reviews, agent measurement, going upmarket, anything.
Catch you next week,
Jared
P.S. Lindsay posts what she's learning on LinkedIn, so find her there. Also, always growing our RevGenius community; you can join here to follow along.
Upcoming events
How to Run Job Change Signals Across Your Whole CRM at a Fraction of the Cost
(Thursday, 24th September | 12 PM EST | 9 AM PT)
Job change signals work. The pricing breaks them: you pay just to check, and 97% of your list doesn't move in a given month. Andy Toizer, Head of Growth at Freckle.io, walks through pay-per-change instead, the plays teams run off the data, and what else RevOps should be watching. You'll leave knowing how to run this across your whole CRM without the cost scaling with it.
Can Your GTM Agents Prove They Stayed Inside the Job?
(Tuesday 22 September, 1 PM EST)
Agents that find demand, prepare the response, and turn what they learn into the next test, working inside the tools you already run. Alex McDonnell of Hyperagent shows the handoffs, the evidence trail, and where people stay in charge.
Revolt! 2026 (Digital Conference)
(14-15 October, 11 AM to 5 PM ET)
Two days, dozens of sessions, 9,000+ leaders.
Last year drew registrants from 3,400+ companies across 40+ countries.
New Perks just dropped
Firmable: 20% off any annual plan
AI-native sales platform with company and contact data, buying signals, and agentic AI for B2B sales teams across the US, Canada, and APAC.
TitanX: 20 reserved TitanX 2.0 waitlist spots plus a 10-point bonus
Phone intent platform that tells outbound reps who is likely to answer before they dial.
Want to get your brand in front of 60,000+ revenue leaders?
Partner with us and put your name at the center of the conversation.
