Vanessa Gatihi joined OpenAI as their first customer success hire 18 months ago. Today, she leads a global CS team as Head of AI Deployment & Adoption spanning San Francisco, New York, Munich, Paris, Dublin, London, Singapore, and Korea, serving a thriving enterprise customer base. Her mission: help OpenAI deliver transformational value to customers and retain and grow them in the age of rapid AI progress.
Top 5 Learnings from OpenAI’s CS Playbook:
- Every single person on your team can now be their own biz-ops analyst – When you integrate AI with your data, team members don’t need to wait two weeks for analysis. CSMs are coming to Vanessa with insights about where to focus new assets and enablement. At Slack, she’d submit requests and wait. Now? Immediate answers on feature correlation to retention, regional differences, and growth patterns.
- Map your customer journey to find AI opportunities – A single cross-functional whiteboarding session identifying the top three friction points in your customer journey can unlock massive value. Use AI to automate repetitive tasks, increase CSM productivity, and add personalization at scale.
- Build prompt libraries, not playbooks – Traditional playbooks were always just “okay.” What actually drives operational rigor today is building high-quality prompt libraries and packaged GPTs that teams can use along the entire customer journey for consistency and personalization.
- Race to ROI, not feature adoption – Customers don’t want AI tools, they want outcomes. Stop measuring AI feature usage and start measuring the business outcomes those features drive. Use early AI adoption metrics as a “Trojan horse” to get executive attention, then pivot to true business value.
- The best tiger team members aren’t your most technical people – Early on, OpenAI picked the most technical team members to pilot new AI tools. Wrong move. What works better: medium-technical people who love to talk and build relationships. They create the trust and empathy needed to drive adoption across teams.
Why Customer Success Matters More Than Ever
Vanessa started with a fundamental truth that every SaaS leader knows intuitively: when your customers succeed, your business thrives. They buy more, stay longer, refer more, and cost less to serve. But in the age of AI, the bar has been raised dramatically.
“ChatGPT has changed expectations,” noted Lowe’s SVP of Digital. “Today’s customers demand more value at greater speed and extremely personalized.”
This isn’t just theory. Lowe’s built AI-powered tools on OpenAI’s platform that help customers get personalized expertise for home projects 24/7. Most traffic comes in the middle of the night when stores are closed. The result? Measurable sales increases and repeat customers who trust Lowe’s for their expensive home improvement projects.
For SaaS leaders, the lesson is clear: you need to build things that give customers instant answers, automate support, and accelerate time to ROI.
The Four-Point Blueprint for AI-Powered Customer Success
To avoid getting stuck in what Lowe’s CIO Samantha Nye Godable calls “the AI pilot doom loop,” Vanessa shared a practical four-point framework:
1. Support Automation (The Low-Hanging Fruit)
If you’re not automating at least some tier-one support tickets, start now. The data is undeniable:
- Klarna is triaging over two-thirds of their tickets using OpenAI, reducing average response time from 11 minutes to 2 minutes, and saving $40 million annually.
- Major enterprises like T-Mobile and banks are deploying similar solutions.
But here’s the key insight: the same agentic technology answering tickets can also be surfaced internally as a co-pilot for your go-to-market team. It’s not agents versus co-pilots—it’s both, depending on your use case and where you are in the journey.
2. Onboarding Velocity (The Biggest Opportunity)
Vanessa’s tactical recommendation: Run a customer journey AI exercise.
Here’s the playbook:
- Map out three journeys: end-user, admin, and account-level customer journey
- Identify friction points for both customers AND internal go-to-market teams
- Prioritize the top three friction areas to invest in now
- Project time-to-value reduction if you invest in these areas
- Move quickly – build something, measure it, iterate
Use ChatGPT throughout this process. Take a photo of your whiteboard session (even with bad handwriting), and ChatGPT will translate it into organized action items. No more losing those post-it notes or waiting a week for someone to type up the brainstorm.
At OpenAI, the CS team focuses on three things:
- Increasing CSM productivity through co-pilots and tools that increase capacity
- Automating repetitive tasks like data pulls, customer settings, and mundane work
- Adding personalization at scale – creating “wow” experiences that make customers feel at the center of your world
The results? OpenAI’s CSMs can create highly customized launch plans in seconds, pulling data from Gong, web searches of recent earnings calls, and customer-specific information. What used to take hours now takes clicks.
3. Data-Driven Decision Making (The Game Changer)
“This for me has been such a game changer in the way that I lead my business,” Vanessa emphasized.
When ChatGPT is integrated with your data, you become your own biz-ops team. At Slack (where Vanessa spent 5 years before OpenAI), she’d submit a request to the biz ops team, wait days for clarifying questions, then finally get analysis two weeks later.
Now? She runs the analysis herself immediately. Even better, individual team members are coming to her with insights about where the business should focus new assets and enablement.
“It’s so powerful when now every single person can have these insights and help shape the company in a way that is putting our resources in the most impactful areas.”
4. Voice of Customer Programs (The Easy Win)
With access to modern AI, you can spin up a robust voice of customer program easily. Integrate your Gong calls, pull in data from your community wherever it lives, and let AI synthesize patterns and insights.
OpenAI is heavily optimizing their VoC program right now because it’s such a rich source of data for product velocity and customer-driven innovation.
The OpenAI Iteration Loop
OpenAI’s competitive advantage comes from their speed of iteration. Here’s their model:
Research (the “beating brain”) → Applied team builds products → Go-to-market deploys to customers → Learn and feed back to research
Vanessa recommends mimicking this loop for AI adoption in your own organization:
- Ideate (You’re your own research lab): Monthly team meetings to identify friction points along the customer journey
- Pilot (Small tiger team): Test with a small group of CSMs in different segments who are close to real customer questions
- Measure (Track time savings and new capabilities): What can they do now that they couldn’t before?
- Iterate and scale: Enable people on the new tool, celebrate learnings, and roll out
Critical insight on tiger teams: Don’t just pick your most technical person. Pick someone medium technical who loves to talk and build relationships. They’ll create the trust and empathy needed to drive AI adoption across the team.
Key Guardrails as You Scale:
- Keep human in the loop, especially early on
- Get buy-in from teams you’ll need to partner with
- Choose great champions (see above)
- Don’t do 1,000 pilots that go nowhere – stay focused
Racing to ROI: Focus on Outcomes, Not Features
Customer value is tricky, and it’s gotten harder with the explosion of AI products and features with different pricing models (consumption, token-based, etc.).
The fundamental truth: Your customers and employees don’t want AI tools. They want outcomes.
Stop measuring whether customers used your AI feature and had X logins. Measure the outcome that AI feature is driving. Even when it’s hard, stay laser-focused on true business value.
Pro tip: Executives actually like talking about AI feature adoption right now. It’s still new and sexy. If that’s your Trojan horse—either internally or with customers—to get to a CEO and then pivot to true business outcomes, use it to your advantage. OpenAI has seen success with this approach.
The OpenAI Value Journey for ChatGPT Enterprise
OpenAI’s value journey shows how to hack quick ROI:
Phase 1 (Immediate): Measure increase in tool usage, understanding, and literacy (surveys, adoption metrics)
Phase 2 (Short-term): Time savings, productivity gains, AI fluency
Phase 3 (Long-term): Transformative business change
The key insight? Phase 3 takes time. By showing immediate ROI in Phase 1, you buy yourself time while working toward the transformational business change that really matters.
Real example: The San Antonio Spurs launched multiple GPTs focused on customer retention. OpenAI measured AI fluency, literacy, and time savings while working toward the ultimate metric—customer retention uplift. (Check out their blog where they’ve hyperlinked their entire seven-step enterprise AI journey manifesto.)
Adapting to Rapid AI Progress: Change Management That Works
Things feel crazy with AI right now. The velocity and speed of change is something we haven’t seen in a while. But as leaders, we need to create flexible processes that help us stay on top of adopting this technology.
The good news? The change management playbook isn’t that different from what we’ve seen before:
- Have great executive sponsorship
- Communicate the why and the vision (What will you do with time savings? How will you show up better for customers?)
- Give people time to experiment and play with tools—in off-sites or hackathons
- Reward experimentation
Critical insight: Build AI automation not just for customer outcomes, but to free up your team’s time to learn new AI features. The landscape will keep changing, and your people need capacity to stay enabled on new progress.
How Vanessa Personally Stays on Top of AI
There’s no magic bullet. It takes time. Here are her tactical tips:
- Block time to use the tools – Get hands dirty with different models. Understand their power and limitations as they exist today.
- Use voice mode religiously – During commutes, washing dishes, folding laundry. Having a personalized tutor explain complex technical papers (like the new inference time scaling paper) while you do other tasks is a game-changer.
- Listen to quality AI podcasts – Vanessa recommends one in particular (check the QR code from her presentation) that helps her stay current.
- Use ChatGPT Projects – If you’re not using Projects in ChatGPT Enterprise, start now. It’s a great way to 10x your work. OpenAI’s free Academy has videos on business workflows and use cases.
Her Final Message: Start Today
“This will not slow down, and each of you really needs to use AI as a competitive advantage. How can ChatGPT or other AI tools help you ideate, pilot, and scale for your customers? You need to do it now.”
Vanessa’s own first-day story embodies this urgency: She launched two customers on day two at OpenAI. They probably weren’t the best launches. But by Friday, after iterating, learning, asking questions, and using ChatGPT at every step to understand customer businesses, she was already dramatically better.
Speed matters. Learning quickly matters. And using AI to help your customers and teams is becoming an essential skill.
Top 5 Mistakes Vanessa Made Building CS at OpenAI
During the Q&A and throughout her talk, Vanessa was refreshingly candid about what hasn’t worked. Here are the key mistakes she and the OpenAI team made along the way:
1. Picking Only the Most Technical People for Tiger Teams
Early on, OpenAI selected the most technical team members to pilot new AI tools. Mistake. What actually works better: choosing people who are medium technical but love to talk and build relationships. They create the trust and empathy needed to drive adoption across teams. Technical skills matter less than communication skills and relationship-building.
2. Running Too Many Pilots That Went Nowhere
The “AI pilot doom loop” is real, and OpenAI wasn’t immune. They ran numerous pilots without clear focus or commitment to scale. The lesson: be ruthless about prioritization. Pick your top three friction points, commit to them, and see them through to scaled deployment. Don’t spread resources across dozens of experiments.
3. Over-Indexing on AI Feature Usage Instead of Outcomes
Initially, OpenAI measured whether customers were using AI features—logins, activation rates, feature adoption. But customers don’t care about using features. They care about outcomes. The team had to shift from “are they using it?” to “what business results are they seeing?” It’s a harder metric to measure, but it’s the only one that matters for retention.
4. Not Building Enough Personalization into Early Customer Experiences
OpenAI’s first customer onboarding experiences were generic. Vanessa admitted her day-two launches “probably weren’t the best.” What changed everything was using AI to create hyper-personalized experiences—custom demos for specific customer teams, tailored training for their exact use cases, dummy data specific to their industry. This created “wow” moments that built true partnership. The mistake was not building this personalization from day one.
5. Underestimating the Time Needed for Team Enablement
As new AI features and models shipped rapidly, OpenAI initially didn’t give their go-to-market teams enough time to learn and experiment. The team was so focused on customer outcomes that they didn’t build enough internal capacity for continuous learning. The fix: intentionally automate work to free up team time for hackathons, experimentation, and enablement sessions. Your team can’t help customers adopt AI if they’re not staying current themselves.
