The dominant AI narrative in revenue right now is that reps are on borrowed time. Every conference talk, every LinkedIn post, every AI SDR pitch works the same corner: the human seat is the thing to automate.
The company most people would expect to lean hardest into that story is Zapier. Automation is literally the product.
Lindsay Rothlisberger built Zapier's first RevOps function three years ago and now runs their AI transformation across the revenue org. What she's actually building looks nothing like the pitch. Her team recovered up to $2.6M in annual selling capacity per 100 reps, and they did it by leaving the human seat exactly where it was.
I sat down with her recently to walk through the mechanics.
The Reframe She Started With
Most revenue leaders sitting down to plan an AI transformation start with the same question: where can we bolt AI onto the funnel to reduce cost?
Lindsay started somewhere else.
Her opening question was "what is stealing our reps' time from customers?" Not what to remove. What to protect.
The answer, once her team mapped it, was uncomfortable in its specificity. The enemy wasn't headcount. It was the connective tissue between a signal firing and an actual conversation. Spreadsheet exports. List re-imports. Manual sequence enrollment. Pre-call research nobody has time to do properly. Post-call CRM data entry that gets skipped and hurts the next quarter's forecasting.
None of that work is selling. All of it eats the day.
That distinction sits underneath the entire Zapier approach. Automation isn't aimed at the conversation. It's aimed at everything that prevents the conversation from happening.
The One Metric That Anchored Everything
Every AI transformation Lindsay's seen fail chose the wrong KPI.
Time saved per rep, tool adoption rates, campaigns shipped, activity volume — these are the numbers most vendors point at when they claim ROI. They tell you the tool deployed. They tell you almost nothing about whether the strategy is working.
Zapier anchored on a single, harder metric: percentage of a rep's time actually spent in front of buyers.
They were at roughly 50%. The target was 75%. Every AI workflow her team shipped was measured against that one number.
If a workflow saved time but the recovered time didn't show up as customer conversations, it didn't count. If a tool was popular but reps were still doing pre-call research manually, it didn't count. The KPI kept the strategy honest.
This is the shift most revenue orgs still haven't made. Activity-volume metrics were designed for a world where reps were the scarce resource. In an AI-augmented world, the scarce resource is the customer's attention, and the metric has to follow.
The Workflow, Step by Step
Everything below happens before the rep opens their laptop for the day.
A signal fires. A job change, a funding event, a new tech install, an intent signal - whatever the connected data layer surfaces (Lusha and other enrichment sources sitting under it). The trigger is the starting gun.
AI drafts the outreach. An AI step writes a signal-specific email that references the actual trigger. The output doesn't go to a Slack channel. It doesn't fire off a "hey, you should reach out" notification. It drops as a finished draft directly into the rep's outbox. Ready to review, ready to send.
Pre-call prep is auto-generated. Account brief, contact context, signal summary, competitive positioning if relevant. All of it delivered to the rep ahead of the meeting. No manual research window before the call.
Post-call, AI writes the MEDDPIC notes back to the CRM. Structured fields, populated automatically from the call transcript. The rep reviews and edits, they don't rebuild from scratch. Which means the notes actually get written, which means the CRM stays useful, which means next quarter's forecasting doesn't drift.
The end-to-end point of the design: eliminate every manual step between signal detection and a real conversation. Signal-to-contact time at Zapier dropped from days to hours.
The Design Principles That Made It Stick
Lindsay's team built and tested a lot of workflows that didn't stick before they figured out what did. The pattern behind the ones that survived is worth stealing.
"If it needs a new tab, it doesn't ship." This is her design constraint, and it's the reason most of Zapier's early AI investments got abandoned. Standalone AI apps with their own UI, their own login, their own workflow logic all failed the same test. Reps didn't adopt them. The only workflows that survived were embedded inside the tools reps already lived in: their inbox, their CRM, their calendar. Adoption was the actual constraint. Sophistication was easy.
Draft, don't notify. A notification is more work. It says "here's a signal, go do something." A draft is less work. It says "here's the thing done, approve it." Lindsay's team ships finished artifacts everywhere they can. The rep's cognitive load drops because the work arrived complete.
Human-in-the-loop by design. AI executes. The rep approves. Nobody's judgment got removed. Their busywork did. This is the frame Lindsay pushes hardest, because it's the reason the whole system is politically survivable inside a revenue org. The reps aren't being replaced. They're being freed up.
Data quality before prompt cleverness. Her team expected three or more iterations before any workflow produced positive results. The reason wasn't prompt engineering. It was that the underlying data was inconsistent, incomplete, or stale, and no amount of clever prompting fixes that. The teams that won the AI transformation game did the boring data-hygiene work first. The teams still losing keep tweaking prompts against messy data and wondering why the output drifts.
Where the $2.6M Comes From
The recovered capacity math is straightforward once the workflow is running.
Roughly five hours saved per rep, per week. Multiply across 100 reps, value the hours at fully-loaded selling capacity (which varies by segment and average deal size), and the annualized number lands somewhere between $1.7M and $2.6M.
Lindsay led with the $1.7M floor in the panel because it's the honest lower bound. The $2.6M ceiling assumes a higher per-hour valuation of selling capacity, which is defensible in enterprise segments and less defensible in SMB. Both numbers are real. Which one applies to your org depends on how you value a selling hour.
The point of the range isn't the number itself. It's that recovered capacity is a P&L line item, not a productivity stat. When the CFO sees seven-figure recovered capacity showing up quarter after quarter, the AI investment stops being a cost center and starts being a revenue lever.
The Replication Checklist
If the mechanism transfers to your org, here's the sequence Lindsay's team ran, compressed into something a head of revenue could work against on Monday:
Pick your one metric. Percentage of rep time in front of customers is the clearest anchor. It forces the rest of the strategy into shape.
Instrument the signals you can actually act on. Job change, funding events, tech installs, intent signals. Ignore the ones you can't act on. A signal without a workflow is noise.
Automate the draft, not the alert. Deliver finished work into the tool they already use. If your team is still sending "you should follow up on this" pings to reps, you're doing it wrong.
Auto-generate pre-call prep and post-call CRM notes. These are the two biggest time sinks in the workflow. Fixing them alone will recover most of the five hours per week.
Keep a human approval step on everything. AI drafts, rep approves. This is the guardrail that keeps quality intact and keeps reps in control of their own book of business.
Expect three iterations. Fix the data before you blame the prompt. The teams that skip the data-hygiene phase are the same teams complaining about AI hallucinations six months in.
What This Actually Means
The narrative that AI is coming for sales reps is the wrong narrative. It's not what the best revenue orgs are building, and it's not what Zapier just proved out at scale.
The narrative that's actually playing out is quieter and more useful. AI is coming for the connective tissue that surrounds a sale. The exports, the prep, the notes, the handoffs. The work that never should have been rep work in the first place.
Lindsay's team built the system that separates the two. If you want to see the full breakdown, the panel is worth watching in full.
- Jared Robin, CEO and Co-Founder of RevGenius
Want to learn from more Revenue Creators like Lindsey? Join the RevGenius community and be part of the movement rewriting the GTM playbook.
Next Coffee Talk: Thursday, July 9th at 12pm ET.
Bring something you're building, something that's stuck, or just bring questions. The point is to show up.
Upcoming Events
Demo Day
(Tuesday, 21st July | 12 PM EST)
Experience the future of GTM with an exclusive look at the essential tech stack for 2026. This no-fluff showcase delivers high-value, 10-minute demos from brands solving real technical challenges, paired with lively community networking and exclusive partner discounts. It’s a straightforward, product-first event designed for immediate impact.
From Workflow Chaos to Operational AI : Real RevOps Use Cases
(Thursday, 22nd September | 1 PM EST | 10 AM PT)
In this session, we're showcasing AI agents built specifically for the event, deployed against real broken RevOps workflows: lead routing, onboarding, forecasting, handoffs, and resurfacing opportunities before they slip. Live builds grounded in real challenges, not generic hype.
Fresh Perks Just Dropped!
Our perks page keeps growing. It's where we collect member-only deals on tools we've hosted, demoed, or run ourselves, each one vetted before it goes up. Five new names just joined, so here's your first look.
Explorium
A B2B data platform that surfaces the right accounts and enriches your pipeline with signals that matter.
Your perk: 20% off the Growth Plan
Airspeed
The commercial brain for revenue teams, deploying AI agents that turn calls, emails, and tickets into action across your CRM.
Your perk: 2-month free trial
Vibe Prospecting
AI-powered prospecting that helps you find and reach the right buyers faster.
Your perk: 20% off the Ultra Plan
FirstTouch
Sales tooling built to help reps start better conversations and move deals forward.
Your perk: 10% off for the first 12 months
TextUs
Business texting that gets your outreach the replies email keeps missing.
Your perk: 10% off your first three months
We're adding new perks every month, so there's always something new worth claiming. Have a look at the full lineup whenever you're planning your next tool decision.
Check them all out →
Put your tool in front of 65k+ buyers
If you run a company our community would benefit from, this is a chance to get in front of 65,000 revenue creators who are already here looking for their next tool.
Add a member perk and you land in front of buyers at the exact moment they're deciding what to try next.
GTM Hot Jobs from RevOps Pipeline
Strategic BD/Sales Head - Citrusbug Technolabs - INR 300K - Apply
Senior Account Executive (Enterprise) - FlyBase Labs - INR 200K - 300K - Apply
Revenue Operations Analyst, Revenue Intelligence & AI GTM Systems - HSI - $115K - Apply
Enterprise Account Executive - Paraform - $125K - $150K - Apply
Brand Partner Marketing Manager - Open AI - $374K - $415K - Apply
Art Director, Claude - Anthropic - $305K - $385K - Apply
Are you hiring and want your job featured? Submit a role here.
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.
