Understory Unfiltered: Marcel Santilli on Building an AI-Native Agency

TLDR: From CMO frustration to codified workflows

Marcel Santilli scaled marketing at HashiCorp (6 to 100 million ARR in two years), led Skill.AI as their first marketing hire, and drove 24x traffic growth at Deepgram. After experiencing the same frustrations at every CMO role, he launched GrowthX: an AI-native agency that codifies marketing workflows like software code.

Listen to this episode to learn why 8-week paid strategy sprints solve the typical agency onboarding problem, how coding agents capture the "messy middle" of expert knowledge work, and why services force you to solve real problems while software alone encourages feature bloat.

Meet the guest: Marcel Santilli, CEO and Co-founder of GrowthX

Marcel built his marketing career at high-growth companies before launching GrowthX, an AI-native agency that treats marketing workflows like software code. His background includes scaling HashiCorp through hypergrowth, leading marketing at Skill.AI during their early stage, and achieving 24x traffic growth at Deepgram, where he spent over a thousand hours building AI workflows.

What makes Marcel's approach unique: GrowthX doesn't just deliver marketing services. They codify every workflow their team builds, capturing the decision-making processes and expert judgment that typically stay locked in individual contributors' heads. The result is a services-as-software model where each client engagement strengthens the underlying platform.

Chapters

[00:21] Marcel's path from Brazil to billion-dollar startups: IBM, HashiCorp, Skill.AI, and Deepgram

[03:50] The breaking point: Why equity schemes and endless meetings drove Marcel to entrepreneurship

[08:00] How solo creators building $10-20M businesses revealed what B2B companies miss about audience trust

[13:38] Services as software: Why GrowthX productizes workflows from human expertise first

[18:00] The quality scaling problem: How GrowthX maintains output quality while onboarding new clients

[20:53] The 8-week strategy sprint: Why every client pays for intensive onboarding before ongoing work

[25:00] Law firm-style pod structure: How GrowthX separates strategy from execution

[26:00] Vibe coding workflows: Why coding agents beat traditional point-and-click automation tools

Key Insights

Every CMO arrives with impossible expectations and no clear path forward.

Marcel's frustration followed the same pattern at every CMO role. "Every time you come into the CMO, you're the most at-risk executive," he explains. "Then you come in and there's crazy unrealistic expectations. You're supposed to come in as a silver bullet." The catch-22 is brutal: spend three months learning the company and the CEO grows impatient, or start making changes immediately without understanding what needs fixing.

This universal CMO experience revealed a market opportunity. Marcel consistently encountered the same pain points around agency quality, freelancer management, and tool proliferation, forming the foundation for GrowthX.

The best services businesses productize the workflows their humans are already building.

As we discuss in the episode, the key insight is simple: humans are already doing these workflows, producing content, and generating results. The opportunity lies in productizing that work to make teams 10x more efficient. Alex frames it as "productizing the workflows that our humans are building" rather than starting with software and hoping it matches real-world needs.

At Understory, we've seen this firsthand with our own GTM engineering work. The workflows we hard-code into our processes came from watching what actually drives results for clients, not from theoretical product roadmaps. Marcel's approach validates this: services force you to solve real problems before you automate them.

Coding agents improve faster than any other AI application because they learn from millions of developer interactions.

The explosion in coding agent capabilities stems from unique training advantages. "Repos are out there, source codes are out there, you can see the pull requests, you can see the change, you can see everything and you can test code," Marcel notes. Tools like Cursor enable what amounts to millions of engineers acting as reinforcement learning loops, continuously validating and improving AI-generated code.

For knowledge work automation, this means treating workflows more like software development than drag-and-drop configuration. GrowthX uses coding agents because code provides better documentation, reproducibility, and the ability to apply developer practices to marketing processes.

The "messy middle" of knowledge work represents the most valuable intellectual property companies fail to capture.

Marcel uses a Michelin Star restaurant analogy: "You have all the pictures, here's every ingredient, you have the inputs and the perfect output. You cannot reproduce that output based on that alone." The messy middle, the thought processes and iteration patterns that transform raw information into valuable work, is what separates great execution from mediocre output.

GrowthX captures this by codifying not just what their experts do, but how they think through problems. This requires understanding where human interventions need to happen, how to evaluate them, and how to learn from them through an expert hub and learning engine.

Hybrid services-software models offer a different path than separating business lines entirely.

Ali raises the comparison to Adam Robinson's Fixer approach, where they run an executive assistant agency separately to fund Fixer AI. Marcel's model differs: GrowthX blends services and software into a single offering where each reinforces the other. The services generate the patterns, the software codifies them, and the cycle compounds.

For SaaS growth leaders evaluating agency partners, this distinction matters. A hybrid model means your engagement directly improves the underlying platform, and you benefit from every other client's learnings. Separate business lines mean you're just another customer for both.

Eight-week strategy sprints solve the agency onboarding problem while generating immediate revenue.

Traditional agency onboarding fails because it rushes to execution without proper foundation work. GrowthX requires every client to complete a paid 8-week strategy sprint. "Everyone comes to us and they're like, we're the same, we just want this thing," Marcel explains. "And then you start and their definition of done is different, their definition of success is different."

Weeks one and two focus on deep company understanding. By week four, most clients are ready to start executing. The sprint generates structured artifacts, including audience personas, brand guidelines, and company context, that feed directly into GrowthX's automated workflows rather than gathering dust in a strategy deck.

Maintaining quality while scaling remains the hardest problem in services, and Marcel admits they haven't fully solved it.

Ali asks directly how GrowthX maintains output quality while signing new clients rapidly. Marcel's honest answer: "It's hard, I don't think we figured it out. Humans are hard, man." This candor matters because most agencies claim seamless scaling while quality quietly erodes.

As we discuss in the episode, the faster you scale, the more clients you onboard, the harder it gets. GrowthX's solution involves law firm-style pod structures with managing directors assigned to clients, separating strategy teams from execution teams. But the underlying tension between growth and quality never fully resolves; it just gets managed more deliberately.

Artifacts built for AI workflows create exponential value for both strategy and execution.

The strategy sprint produces structured company knowledge that can be dynamically ingested into marketing workflows. Marcel describes doing "an entire TAM and ICP analysis with over 50 different sources" for a client attempting a major pivot. "They were kind of shocked. This is so much better than if we had hired a consultant."

These artifacts aren't static planning documents. They're structured inputs that power personalized content, outbound sequences, and paid media campaigns across every engagement. Each workflow improvement discovered through client work becomes available across teams, creating compounding returns.

Vibe coding represents the future of workflow automation for non-developers.

Current workflow creation at GrowthX requires engineering expertise, but Marcel plans to launch "a vibe coding experience for non-developers" that allows domain experts to build sophisticated automation without traditional coding skills. The system uses a four-file structure containing prompts, workflows, activities, and documentation where coding agents can pull existing components automatically.

This approach enables subject matter experts to codify their expertise directly. A successful SDR can document their prospecting process in a way that gets automatically converted into executable workflows for the entire team, eliminating the translation layer between business users and technical implementation.

Want more insight on building AI-native marketing operations? Listen to the full episode on YouTube and subscribe to Understory's podcast for more insights on coordinated SaaS growth.

Looking to eliminate the coordination overhead that's slowing your SaaS growth? Book a call with Understory to explore how expert allbound execution across paid media, GTM engineering, and creative can replace managing multiple specialists.