CMO Shares How To Use AI As A Learning And Experimentation Loop
Lomit Patel explains how AI helps marketing teams replace campaign cycles with continuous experimentation, turning customer signals into faster, smarter growth decisions.
Lomit Patel
CMO at TYB (Try Your Best)
AI Integration: AI shifts marketing from campaigns to continuous learning systems, enhancing signal interpretation and strategy.
Real-time Learning: Marketing now prioritizes speed and adaptability in interpreting user behavior and adjusting strategies.
Dynamic Segmentation: AI enables behavior-based clustering, overtaking traditional stable user segment models in marketing.
Experimental Shift: Always-on experimentation replaces static campaign cycles, improving message adaptation and performance.
Human Role: Human oversight remains crucial for strategic decisions, narrative consistency, and prioritization in marketing.
Lomit Patel is a growth and marketing executive, currently acting as CMO at TYB (Try Your Best). He's also the author of Lean AI.
We caught up with Lomit to learn how he's turning marketing operations into a continuous learning and experimentation loop. Here's what he said.
Marketing is being redefined
I’ve spent the last 20+ years helping to scale companies like Roku, IMVU, TrustedID, and Texture. Currently, I'm CMO at TYB (Try Your Best), where I lead a lean but high-velocity marketing organization built around one core belief: Community is not a channel, it is a go-to-market system.
Create a Free Account to Read More
You need to sign up for a free membership to complete reading this article:
Have an account? Log In
Across each of these roles, I consistently built systems that connect customer insight to measurable business outcomes. Early in my career, growth was largely channel-driven. You optimized campaigns, tested creatives, and scaled what worked. But even then, I focused on something deeper: how to use data and experimentation to remove guesswork and increase decision velocity across the entire funnel.
That thinking evolved into a broader perspective I explored in my book, Lean AI, where I focused not just on using AI for efficiency, but on redesigning how growth systems operate.
At IMVU, for example, we moved from manual optimization to AI-driven orchestration of user journeys, creative testing, and lifecycle engagement. My goal was always the same: increase the speed and quality of learning, while improving retention and LTV.
Today, as CMO at TYB, that shift has fully accelerated. AI is no longer just a tool in the stack; it is becoming the operating layer for modern marketing. We are moving from campaign thinking to system thinking, where signals, automation, and creative intelligence continuously interact in real time.
What excites me most is that marketing is being fundamentally redefined. We are moving from channel operators and storytellers to architects of intelligent growth systems that learn, adapt, and scale impact across acquisition, engagement, and monetization.
How AI has turned marketing into a real-time learning system
AI has challenged three big assumptions in marketing:
- In an AI-enabled environment, channel efficiency becomes more of a baseline than a bottleneck. The system around it differentiates performance — how quickly can you detect intent, interpret behavior, generate hypotheses, and adapt messaging or product experiences in response.
- AI showed us that segments are not stable enough to be the foundation of marketing strategies. Traditional segmentation models assume relatively fixed user groups for consistent messaging. With AI, we now see that user behavior is far more fluid. Intent shifts quickly, and static segments often lag reality. This pushed us toward dynamic, continuously updating, behavior-based clustering.
- Campaigns are no longer the core unit of marketing execution. Campaigns assume a start, end, and evaluation cycle. AI pushed us toward always-on systems where experiments continuously run, learning is incremental, and optimization is constant rather than episodic.
In other words, marketing is a real-time learning system where speed of interpretation and adaptation matter more than static planning or perfect targeting.
How AI reshapes marketing processes
Marketing is no longer organized around campaigns. It is organized around a continuous learning loop where AI helps us detect signals, generate hypotheses, and scale what works in near real time.
So, we've shifted from static campaign planning to always-on, signal-driven experimentation across the entire marketing system.
Historically, we planned campaigns in cycles. We’d define messaging, build creative sets, launch, and then analyze results after a meaningful delay. That model is too slow in an AI-native environment because user behavior and intent signals evolve quickly.
We introduced an AI-assisted growth workflow that continuously turns real-time behavioral signals into triggered experiments and adaptive messaging across lifecycle, content, and performance channels. Instead of waiting for a campaign to finish, we now treat every meaningful user signal, onboarding behavior, community participation, and creator activity as inputs to immediately inform the next best action.
Practically, this has reshaped how teams operate. Creative iteration is faster because AI helps generate and test variations continuously. Lifecycle marketing is more dynamic because messaging is no longer tied to fixed segments but to evolving behavioral clusters. And in-product intent signals increasingly inform performance marketing, not just platform-level targeting.
The bigger change, though, is cultural. Marketing is no longer organized around campaigns. It is organized around a continuous learning loop where AI helps us detect signals, generate hypotheses, and scale what works in near real time.
An AI-assisted growth workflow
Here's the workflow.
- It starts with real-time behavioral signals inside the product and community layer. We track actions like onboarding progression, participation depth, creator activity, and engagement patterns. The key shift is that these signals are no longer just analytics dashboards; they are structured inputs into our marketing system. Tools: Segment for event collection and routing, Amplitude for behavioral analytics, and Snowflake as the data warehouse backbone.
- Next, AI helps us cluster and interpret those signals into dynamic intent segments. Instead of relying on static personas or fixed lifecycle stages, we use AI to identify patterns like “high-intent but low activation,” “active participants with low retention risk,” or “new users showing early advocacy behavior.” These segments constantly evolve based on live behavior. Tools: BigQuery and Python/ML models to build behavioral features, with OpenAI (GPT) used to label and interpret emerging intent patterns.
- After forming those clusters, AI assists us in generating hypotheses and creative direction. For each segment, we identify the message most likely to influence behavior, and AI helps generate variations across copy, hooks, and content angles. This is not fully automated output, but it dramatically increases the speed of exploring messaging paths. Tools: ChatGPT and Claude to generate messaging angles, hooks, and positioning variations, with Notion AI to organize and maintain a living creative library.
- We then feed those outputs into always-on experimentation across lifecycle, paid, and community touchpoints. Instead of running isolated campaigns, we constantly deploy small, iterative tests, adjusting messaging, creative, and timing based on real-time performance signals. Tools: Braze for lifecycle experiments, Optimizely for product and web A/B testing, and LaunchDarkly for feature flagging and controlled rollouts.
- Finally, and most importantly, the system closes the loop with AI-assisted learning synthesis. We don't just analyze what performed better; we use AI to surface why certain messages or journeys worked for specific segments. Those insights then directly inform updated segmentation logic, messaging frameworks, and product-informed marketing decisions. Tools: Amplitude for behavioral insights, Looker for structured reporting, and Claude to synthesize learnings into narrative and decision frameworks.
How AI overhauls benefit marketers
AI-driven clustering and insight generation allow us to move toward dynamic audience definitions, which better reflect how users behave rather than how we historically bucketed them.
On the positive side, the biggest impact of AI has been speed and throughput of experimentation. We’ve meaningfully increased the number of creative and lifecycle experiments we can run in a period, because AI reduces friction when generating variants, testing messaging, and iterating based on early signal performance. This has translated into faster learning cycles across both acquisition and retention programs.
We’ve also seen improvements in conversion efficiency driven by better signal usage. By connecting in-product and community behavior to marketing activation, we respond to intent much earlier in the user journey. That has improved the quality of traffic-to-activation flow and strengthened downstream retention because users receive more contextually relevant touchpoints.
Another clear benefit is reduced dependency on rigid segmentation and manual analysis. AI-driven clustering and insight generation allow us to move toward dynamic audience definitions, which better reflect how users behave rather than how we historically bucketed them.
Why moving too quickly on AI insights can be a risk
On the more nuanced side, we’ve also had to adjust expectations around signal noise and over-automation.
Early on, we tended to act on too many AI-generated insights too quickly, creating fragmentation in messaging and experimentation overload. We had to step back and introduce stronger human filters for prioritization and narrative consistency.
We also underestimated the need for discipline in defining success metrics upfront. AI makes it very easy to iterate quickly, but if you don’t anchor experiments to clear behavioral or revenue outcomes, you can end up optimizing for engagement within a narrow slice of the system rather than meaningful business impact.
If I had understood that earlier, I would have spent less time scaling experimentation volume and more time designing the operating model around AI-driven decisions, including tighter hypothesis frameworks, clearer ownership of decisions, and stronger alignment between marketing signals and business outcomes.
What AI handles vs. where humans must maintain oversight
We draw a clear line between what AI informs and what humans own, and the distinction comes down to one question: Is this about optimization and pattern recognition, or is it about judgment, context, and taste?
We increasingly rely on AI to power signal processing, segmentation, and experimentation at scale. That includes identifying behavioral patterns across community and product activity, clustering users into dynamic segments based on real-time actions, and generating and iterating on creative and messaging variations. AI also heavily analyzes performance data across channels, surfacing non-obvious insights faster than manual review ever could.
AI is especially powerful at compressing the loop between signal and action. For example, it can detect shifts in engagement behavior inside the product or community, suggest hypotheses, and even generate testable creative or lifecycle variants. That allows us to run far more experiments than a traditional team structure would support.
Humans own positioning, narrative, and creative direction. AI can generate variations, but it cannot define what we stand for or how we want to show up in the market. That requires context about brand, audience psychology, and long-term differentiation that goes beyond pattern recognition.
In community-driven environments, specifically, AI has not replaced authentic human interaction and trust-building. It can assist in scaling engagement and identifying patterns, but it cannot replicate the credibility that comes from real participation, leadership presence, and human judgment in community moments.
Humans also own prioritization and strategic tradeoffs. As I said before, just because AI can surface dozens of opportunities does not mean we should act on all of them. Deciding what not to do, and what aligns with the broader growth thesis, is still a deeply human responsibility.
Finally, humans own interpretation of meaning. AI can tell us what is happening, but humans are responsible for understanding why it matters and how it should reshape the strategy.
How CMOs should approach AI transformation
I'll share four pieces of advice:
- Don’t treat AI as a set of tools to plug into your existing marketing machine. Treat it as a reason to redesign the machine itself. Because this isn't about productivity, the real shift happens when you move from campaign-based operating models to continuous, signal-driven systems of growth where AI is embedded into how you detect intent, make decisions, and learn in real time.
- Rethink what “scale” means. In the past, scale involved channels and budget efficiency. In an AI-driven world, scale is about learning velocity. The teams that win will not be the ones that run the most campaigns, but the ones that can convert signals into experiments, and experiments into decisions, faster than everyone else.
- Be very intentional about where humans stay in the loop. As AI takes on more execution and analysis, the human advantage becomes sharper, not weaker. Leaders need to double down on narrative, positioning, and prioritization. If everything is measurable and optimizable, then judgment becomes the primary differentiator.
- Invest in systems thinking over channel thinking. AI breaks down silos between performance, lifecycle, product, and community. The opportunity is to design a unified growth system where signals flow across all of them, rather than optimizing each in isolation.
I’ll share four pieces of advice: Don’t treat AI as a set of tools to plug into your existing marketing machine…Rethink what “scale” means…Be very intentional about where humans stay in the loop…Invest in systems thinking over channel thinking.
Follow along
You can follow along with Lomit Patel on LinkedIn and his personal blog. And check out his book, Lean AI.
More expert interviews to come on The CMO Club!
