The fastest-growing companies in B2B + AI are not running the customer success playbook the rest of the market spent a decade building. At the FDE / CS Summit at SaaStr AI Annual 2026, leaders from the companies setting the pace got specific about what they’ve thrown out and what they’ve replaced it with.
The lineup:
- Monica Perez, Global Head of Customer Success, Lovable
- Tom Ronen, VP Customer Success, Harvey
- Ryan Seams, Head of Customer Success & Solutions, Assembly AI
- John Gleeson, Founder & GP, SuccessVP
- Bobby Cooper, Founder & CEO, Retention Intelligence (panel moderator)
- Ashvin Vaidyanathan, Chief Customer Officer, LinkedIn
- Ursula Llabres, SVP Customer Success & Support, Content Square
- Daniel Silverstein, VP Customer Success & Head of Business, Carta
Here are the ten things they said that cut against conventional wisdom, with the numbers behind each one.
1. The CSM title itself is a tax on your customer experience
Ryan Seams, who leads customer success and solutions at Assembly AI, went on a listening tour and watched technical buyers physically change their body language the second he said “head of customer success.” They got defensive. The title now reads, to a developer, as a commercially minded message-relayer there to run the QBR and push the renewal. Assembly tried renaming the role to technical account manager and it was a disaster in recruiting, two good candidates in two and a half months. They switched to forward deployed engineer and the pipeline filled with the right people. The actual job barely changed. It was the brand.
Founder takeaway: Audit what your customer-facing titles signal to the person on the other side of the table, especially if you sell to technical buyers. The word on a Slack profile changes whether a customer leans in or shuts down before the first sentence.
2. Saying “AI-powered” out loud signals you’re behind, not ahead
Monica Perez runs customer success at Lovable, the fastest company in the world to $400M. Her most contrarian move: stop talking about AI as the headline. Every deck in the room said “AI-powered,” and her argument is that leading with it signals you’re not thinking ten steps ahead. AI is becoming the baseline, the water the fish swims in, not the story. Lovable’s onboarding doesn’t open with what AI can do. It opens with what the customer will unlock. The moment you stop saying AI in every conversation is the moment you actually become AI-native.
Founder takeaway: Pull “AI” out of your next customer conversation and see if the value still stands on its own. If it doesn’t, you’re selling the technology instead of the outcome, and the customer can feel it.
3. A traditional, on-site, “2001-looking” CS motion can be exactly right
Tom Ronen, VP of customer success at Harvey, made a confession the room did not expect. Harvey runs an old-school motion: executive business reviews, heavy on-site work, change management frameworks that look like they were printed by a Big Four consultant two decades ago. At an $11B company crossing $190M ARR with 100,000+ attorneys across 60+ countries, that is the deliberate choice. Selling AI into a 200-person law firm isn’t selling software. It’s selling a fundamental change in how partners who’ve practiced the same way for 30 years do their work. No automated QBR or health score gets you there.
Founder takeaway: The “AI-native CS playbook” of agents and automated QBRs is right for some products and wrong for others. If your sale requires deep behavior change inside a skeptical organization, high-touch and on-site is a moat, not a relic.
4. Over half of what your CSMs do has no correlation with retention
Bobby Cooper, founder of Retention Intelligence, shared platform data showing that more than 50% of CSM activities have no correlation with retention or the goals on their plate. CSMs accrete work over time, become jacks of all trades, and teams bloat. Ashvin Vaidyanathan at LinkedIn gave the cleanest fix: map every activity, whether it’s a human task, an email campaign, or an agent interaction, to whether it changes the product outcome for the customer inside the intended timeline. If it doesn’t, it has no place in the playbook.
Founder takeaway: Run a time-and-motion study on your post-sales team before you add headcount or buy a tool. You’re likely funding a large block of activity that does nothing for retention, and AI should replace that work, not accelerate it.
5. Adoption is the new starting line, not the finish line
For most of B2B, decent seat utilization plus the same buyers at renewal time meant CS did its job. Ronen’s point is that AI breaks this. A law firm can be logging into Harvey constantly without changing how it operates. Logins are not transformation. Harvey tracks five ROI pillars and asks CSMs to tag value stories against them, because the goal is reducing non-billable work and speeding up matters, not racking up active users.
Founder takeaway: If your health score still leans on login frequency, it’s measuring the wrong thing in an AI product. Define the business outcome the customer bought, then measure progress toward it. Usage is at best a leading proxy.
6. Context plus verifiable correctness equals a unit of work, and that’s what you get paid on
John Gleeson, who took Motive’s CS from $1M to over $300M ARR and now runs the applied-AI fund SuccessVP, gave the sharpest framework of the day. AI succeeds when two conditions are met: high context and verifiable correctness. Code has both baked in, which is why every company hitting $100M ARR in under a year is a developer tool, Cursor in 12 months, Bolt in 14, Lovable in 8, Replit’s agent from $10M to $100M in six. Most domains don’t get those conditions for free. Somebody has to engineer them, and that someone is now the vendor’s CS team. Put the two together and you get a unit of work, and the unit of work is the unit of revenue.
Founder takeaway: Find your unit of work, the discrete thing a customer pays you to produce. Then ask what context the system needs and how you verify the output is correct. Engineering those two conditions is the actual job now, and it’s where revenue gets created.
7. NPS is dead, and the panel said so unanimously
Cooper’s panel with Vaidyanathan (LinkedIn), Daniel Silverstein (Carta), and Ursula Llabres (Content Square) reached a rare consensus: kill NPS. Response rates are low, and the people not responding are often the ones churning. The survey catches customers at an arbitrary moment in time. And there’s no clean correlation to gross revenue retention. Llabres named the tell directly, a great NPS score sitting right next to weak retention. They’re loving you in the survey and leaving with their wallets. Silverstein can’t fully kill it because the board still asks, so he supplements with CSAT and CES and aggregates rather than trusting one number.
Founder takeaway: Stop treating NPS as a leading indicator of revenue. If you must report it for the board, surround it with transaction-level signals (CES, CSAT) and product telemetry that actually correlate with renewal.
8. One change-management document drove 2x more revenue
Ronen cited a Harvard study of 1,515 startups. Both groups got the same AI tools and the same training. One group additionally got a single document describing how teams like theirs had used the tools successfully. That one piece of change management produced 2x more revenue and made the group 18% more likely to acquire paying customers. A Microsoft survey of 500 enterprise decision-makers landed in the same place: AI vision and change-management confidence correlated with success above the technical factors. As Box’s Aaron Levie put it, the winners deliver the change management that drives workflow change, not the best model or UI.
Founder takeaway: Change management is not the soft, unmeasurable part of the sale. It’s one of the highest-return inputs to revenue you have, and a small, deliberate investment compounds. Write the simple “here’s how teams like yours succeed” document and put it in front of every new customer.
9. Moving “closed-won” into implementation crushed churn
Cooper described a shift Weave made on its way from $8M to $200M ARR through IPO. They stopped letting sales or RevOps mark a deal closed-won at signature. A deal only counted as booked once the customer crossed a defined success threshold inside implementation. That single change fixed the handoff problem (a signed deal that never kicked off was never counted), drove sales to qualify harder for customers likely to succeed, and took churn from 4% per month down to roughly half a percent while scaling. The whole company aligned around the moment value actually started, not the moment paper was signed.
Founder takeaway: Look at where you draw the closed-won line. Pulling it into implementation, tied to a real activation milestone, aligns sales and CS around the same outcome and surfaces failure-to-launch churn before it compounds.
10. You can rebuild your entire CS stack yourself, and version it weekly
Perez replaced Gainsight with a customer success command center her team built on Lovable. It tracks the portfolio in real time, surfaces risk and expansion before they become fires, recommends next steps per account, and generates individualized living hubs for every customer. When it stops working, they version it weekly. The point that lands hardest: she’s not technical and never has been. Her reasoning is that the team closest to the customer should own the tool, and waiting for a vendor’s 12-month roadmap is now the constraint, not the engineering budget. The build cost has collapsed.
Founder takeaway: Before you renew a rigid CS platform, ask whether your own team could build the 80% that actually matters. If your tools move slower than your team, the tool is the bottleneck, and that’s now a choice rather than a given.
Have a question for Dear SaaStr? Submit it at saastr.ai/ai-mentor
10 More Quick Take-Aways From the FDE / CS Summit:
- Onboard during the interview. Monica Perez (Lovable) ran a one-week work trial that had her onboarding a real Fortune 100 customer before she was hired, then met every customer in her first 30 days. CS ships day one.
- Give every customer their own scoped AI bot. Perez (Lovable): hundreds of individualized Slack bots, each scoped to one customer’s instance, projects, and contract terms, no harder to build than one.
- Demystify the model before anyone logs in. Tom Ronen (Harvey) runs an LLM and prompt-basics session first, rooted in ADKAR, because people don’t skip change-management steps.
- Win with the crown-jewel group first. Ronen (Harvey): land the most prestigious or revenue-driving practice group so the rest of the firm follows.
- Measure deflection honestly. Ryan Seams (Assembly AI): the real number is pickup rate times end-to-end resolution, not the headline. An 88% deflection rate on 20% of issues isn’t what it sounds like.
- “Does it come with an FDE?” is now a buying question. Seams (Assembly AI): some buyers won’t sign for software unless a forward deployed engineer comes with it.
- Multi-thread to the managers, not just the buyer. Ashvin Vaidyanathan (LinkedIn): the first- and second-line managers of your end users make or break agentic adoption.
- Health scores are dying as outcomes go on-demand. Daniel Silverstein (Carta): with plain-language MCP access, and Claude deployed to all 2,000 employees, the only questions left are was it right and was it fast.
- Kill time-to-live too. Bobby Cooper (Retention Intelligence): it’s an arbitrary finance milestone with no correlation to value. Track time-to-value instead.
- Services are the new software. John Gleeson (SuccessVP): roughly $6 of services spend for every $1 of software, per Sequoia. Applied AI is going after the work, which is why CS becomes a toll booth, not an annuity.
