Why Optimizely is Betting on Agentic AI | Cory Liebgott, VP of Product (Optimizely)
From rethinking AI roadmaps to building internal tools like OptiGPT, Optimizely's Cory Liebgott shares lessons on pivoting fast, keeping customers’ trust, and using AI to amplify product teams.
In this episode of LaunchPod, we sit down with Cory Liebgott, VP of Product at Optimizely, to explore how her team is navigating AI adoption with customer trust at the core.
Cory shares how Optimizely avoided a major misstep by stress-testing its AI roadmap early and pivoting from benchmarking data toward agentic AI. She also explains how her teams use AI internally, from PRD writing to OptiGPT, to boost productivity and collaboration.
Here’s what we cover:
Why customer pushback on data sharing reshaped Optimizely’s AI strategy (02:00)
How “agentic AI” powers experimentation ideation, brand-aligned content, and human-in-the-loop workflows (13:00)
Lessons from DocuSign: how not to let the “squeaky wheel” dominate product roadmaps (15:00)
How Optimizely uses AI internally (i.e., Copilot, Claude, and their own OptiGPT) to accelerate product delivery and decision-making (23:00)
1. Rethinking AI roadmaps
Cory learned that the fastest way to derail AI adoption is to assume customers will embrace your roadmap without scrutiny. Her team initially planned to use AI to benchmark data across customers.
But when they floated the idea, the pushback was immediate:
“We made some incorrect assumptions initially […] and the feedback we got was quite the opposite. Customers told us, ‘I don’t want you to couple my data with anyone else, and if that’s what it takes, I’m not interested in using AI.’”
Instead of pushing forward with their initial plan, Cory’s team pivoted to agentic AI, focusing on tools that operate safely within each customer’s own ecosystem.
Things you can do:
Stress-test AI roadmaps with customers early, before investing heavily
Be transparent about data use and set clear governance principles.
Pivot fast when customer feedback shows misalignment
2. Going all-in on agentic AI
Optimizely’s agentic answer was Opal, their AI suite built on Google Gemini. These agents are trained on a customer’s brand guidelines, tone of voice, and best-performing assets.
Examples include:
Experimentation ideation agents that suggest fresh A/B test ideas when customers run out of their own
Content agents that draft emails or blogs aligned with brand voice
Human-in-the-loop workflows where AI assists, but humans approve final output
“We’ve built agents where you can train them on your brand guidelines, your tone of voice, even your best email or blog examples, and say: ‘We want others to look like this.’”
This approach unlocked real value: AI handled repetitive, low-value tasks while marketers and PMs stayed focused on creative and strategic work.
Things you can do
Train AI agents on your unique brand voice and best examples
Apply AI where creativity stalls (e.g., ideation, drafts, repetitive workflows)
Keep humans in the loop for oversight and trust
3. Avoiding the “squeaky wheel” trap
Earlier in her career at DocuSign, Cory ran into a classic product pitfall: over-listening to the most vocal users. Transaction coordinators were eager to give feedback, but real estate agents (the real revenue drivers) remained quiet.
It wasn’t until leaders reinforced the idea “focus on the agents” that the team realized their prioritization was off. The pivot built stronger adoption and kept the most valuable personas engaged.
“It’s really hard to say no, because at heart we’re problem solvers. But if you can explain why you’re prioritizing one persona over another, people are receptive. That builds trust.”
Things you can do:
Don’t confuse volume of feedback with the importance of feedback
Map personas to revenue drivers before prioritizing requests
Explain trade-offs openly; being transparent builds trust even when you say “no”
4. Practicing what they preach: AI at Optimizely
Cory’s team isn’t just building AI for customers; they’re using it daily across product and engineering.
Copilot: Writing PRDs, summarizing meetings, and cleaning up voice-dictated notes into polished documents
Claude and Cursor: Rapid prototyping, vibe coding, and technical translations for non-technical PMs
OptiGPT: An internal tool that lets employees query customer data, product info, HR policies, and more
Things you can do
Host regular “AI show-and-tell” sessions where team members demo tools they’ve tried
Use enterprise tools like Copilot for structured workflows, but don’t ignore emerging ones like Claude for prototyping
Encourage experimentation (even half-baked attempts inspire collaboration and learning)
Chapters
00:00 Introduction
01:52 Implementing Agentic AI at Optimizely
04:58 The Importance of Customer Feedback
06:27 Using AI Tools Internally at Optimizely
15:30 DocuSign and the Squeaky Wheel Problem
30:36 Conclusion
Links
What does LogRocket do?
LogRocket's Galileo AI watches user sessions for you and surfaces the technical and usability issues holding back your web and mobile apps. Understand where your users are struggling by trying it for free at LogRocket.com.