10 Lessons from Braze’s CTO on Building an AI-First Engineering Team

Jon Hyman, co-founder and CTO of Braze, has spent nearly 15 years steering the company’s engineering organization through rapid growth. In just a few months, he transformed his team into an AI-first powerhouse. Here are the ten key insights from his playbook—each a critical step for engineering leaders navigating the agentic era.

1. Embrace Agentic Thinking

Hyman realized that traditional automation wasn’t enough. Instead of simply programming predefined rules, his team began building autonomous agents that could make decisions on their own. This shift required engineers to design systems that learn from data and adapt in real time. The result? Smarter, more responsive customer interactions that feel almost human. By fostering a mindset where every problem is viewed through the lens of agentic AI, Braze moved beyond static workflows and into a dynamic, self-improving infrastructure.

2. Speed of Transformation

What typically takes years, Braze accomplished in months. Hyman prioritized rapid iteration over perfection. Small, cross-functional squads were given the freedom to experiment with AI models, fail quickly, and course-correct. This speed wasn’t chaos—it was a disciplined sprint. Weekly showcases kept everyone aligned, and any successful prototype was fast-tracked into production. The key lesson: don’t wait for a grand plan; start with a small agent and scale up as you learn.

3. Culture of Experimentation

Transformation starts with culture. Hyman encouraged a “test and learn” environment where engineers felt safe trying out new AI techniques. He removed the stigma around failure by celebrating experiments—even those that didn’t work. Teams were given dedicated time each week to hack on agentic features, from predictive messaging to self-optimizing campaigns. This culture shift turned skepticism into enthusiasm, and within weeks, the entire organization was buzzing with AI-driven ideas.

4. Leadership Alignment

No transformation succeeds without buy-in from the top. Hyman personally led town halls and one-on-ones to articulate the vision of an AI-first engineering team. He made sure product heads, data scientists, and engineering managers all spoke the same language. By creating a shared roadmap that tied agentic capabilities to business outcomes, he eliminated silos. When the CEO and CTO are visibly aligned, the entire organization moves faster and with more confidence.

5. Reskilling the Workforce

Engineers can’t build what they don’t understand. Hyman invested heavily in upskilling his team. Braze launched internal courses on machine learning and agentic architectures, taught by experts from within the company. Every developer, from frontend to backend, learned the basics of prompting, RAG (Retrieval-Augmented Generation), and model evaluation. This wasn’t a one-time workshop—it was an ongoing learning culture. Within months, even junior engineers could contribute to AI feature development.

6. Data Infrastructure Overhaul

AI agents are only as good as the data they consume. Hyman oversaw a major data infrastructure upgrade to ensure clean, accessible, and real-time data pipelines. The team migrated to a unified data lake, standardized event schemas, and implemented strict data governance. This allowed their agentic models to ingest customer behavior instantly. Without this foundation, the AI-first transformation would have been impossible. As Hyman often says, “Garbage in, garbage out—but gold in, agent out.”

7. Building Internal AI Tools

Instead of buying every external solution, Hyman’s team developed internal AI tooling tailored to Braze’s unique workflows. A custom “agent builder” let product managers define agentic behaviors without writing code. Another tool automated model retraining and deployment. These internal platforms accelerated iteration and reduced dependency on third-party vendors. They also gave engineers deep ownership of the agentic stack, fostering a sense of pride and accelerating the learning curve.

8. Customer-Centric AI

Every AI feature was designed with the end user in mind. Hyman insisted that agentic capabilities must enhance, not replace, human intuition. For instance, Braze’s AI-driven campaign advisor suggests optimal send times and content, but marketers retain final control. The team ran extensive A/B tests to ensure that agentic recommendations truly improved KPIs. By keeping the customer at the center, they avoided the trap of building AI for AI’s sake.

9. Measuring Success

Old metrics like click-through rates no longer sufficed. Hyman introduced new success measures for AI: agent task completion rate, human-in-the-loop intervention frequency, and model drift over time. These metrics gave the team visibility into whether their agents were truly working or just mimicking patterns. Dashboards were built to track agent health in real time. This data-driven approach ensured that the transformation was measurable, not just a buzzword.

10. Future-Proofing

Finally, Hyman emphasizes that the work never stops. He created a continuous learning and adaptation framework where the engineering team regularly scans for new agentic paradigms—like multi-agent systems or reinforcement learning. Periodic “AI innovation sprints” are held to explore cutting-edge research. The goal is to stay ahead of the curve so that Braze’s platform can evolve as fast as the technology itself. In the agentic era, standing still is falling behind.

Conclusion: Jon Hyman’s journey at Braze proves that even established engineering organizations can pivot to an AI-first model in record time. By embracing agentic thinking, fostering experimentation, aligning leadership, and keeping customers at the core, any tech company can survive—and thrive—through the next wave of innovation. The lessons are clear: start small, learn fast, and always keep the agent’s purpose aligned with human needs.