10 Key Insights from Braze’s CTO on Engineering in the Agentic Era

Jon Hyman, co-founder and CTO of Braze, has spent nearly 15 years steering the company’s engineering organization through explosive growth and rapid technological shifts. In a recent conversation, he shared how his team transformed into an AI-first powerhouse in just a few months—and what that means for the future of engineering. Below are ten essential takeaways from his journey, offering a playbook for leaders navigating the agentic era.

1. Embrace the Agentic Shift Early

Hyman emphasizes that engineering leaders must recognize the move from human-directed workflows to agentic systems—where AI agents act autonomously. For Braze, this meant pivotaling quickly to embed AI into every layer of their platform. By treating agentic capability not as a feature but as a core architectural principle, they accelerated delivery and improved decision-making. The lesson: don’t wait for the perfect plan; start experimenting with small agentic modules to build organizational muscle.

2. Cultivate a Culture of Rapid Experimentation

Braze’s transformation to an AI-first team didn’t happen overnight—it happened in a few months because the culture already prized rapid experimentation. Hyman notes that his team ran dozens of small trials, discarding what didn’t work and doubling down on successful prototypes. This iterative approach reduced risk and allowed engineers to upskill organically. Leaders should create safe spaces for failure and reward learnings over rigid adherence to plans.

3. Rethink Team Structures for AI

Traditional engineering teams often separate data, machine learning, and product development. Hyman broke down those silos by forming cross-functional squads that owned end-to-end agentic features. Each squad included data engineers, ML experts, and backend developers, fostering holistic ownership. This structural change was crucial for Braze’s speed, as it eliminated handoffs and empowered teams to iterate faster. Consider revisiting your organizational chart to align with agentic workflows.

4. Invest in Foundational Infrastructure

Before deploying AI agents at scale, Braze had to modernize its data infrastructure. Hyman explains that reliable, real-time data pipelines are the backbone of agentic systems. They invested heavily in event-driven architectures and robust model serving layers. Without this foundation, agentic features would be brittle and slow. For any engineering leader eyeing AI adoption, prioritize data quality and latency before adding intelligence.

5. Shift from Feature Factories to AI-Enabled Platforms

In the agentic era, Hyman advises moving away from building endless features and instead constructing platforms that enable AI agents. Braze transformed its product into a set of composable API services that agents can orchestrate autonomously. This shift reduced the engineering burden of maintaining feature lists and opened up new possibilities for customer personalization. The goal is to build a Lego set of capabilities, not a monolithic application.

6. Prioritize Observability and Safety Guards

When AI agents act autonomously, errors can cascade quickly. Hyman stresses the need for comprehensive observability—logging every decision, confidence score, and fallback path. Braze implemented layered guardrails, including human-in-the-loop checks for high-stakes actions. Engineers must treat agentic systems as production-grade components from day one, with monitoring metrics that detect drift or malicious behavior. Safety is not optional; it’s table stakes.

7. Upskill Your Engineers—Fast

Transforming to an AI-first team requires upskilling existing engineers, not just hiring new talent. Hyman launched internal bootcamps, pairing backend engineers with ML specialists to learn prompt engineering and agent orchestration. Within months, many senior engineers were comfortable designing agent loops and fine-tuning small models. The key is to make learning a part of daily work, not a separate track.

8. Redefine Success Metrics

Traditional metrics like feature velocity or bug count become less relevant when agents perform tasks. Hyman introduced agentic KPIs: task completion rate, autonomy ratio (percentage of tasks handled without human intervention), and user satisfaction scores. These metrics drove prioritization and helped the team see value quickly. Engineering leaders must shift from measuring output to measuring outcomes enabled by agents.

9. Maintain a Human-Centric Design Philosophy

Despite the autonomy of agents, Hyman insists that the user experience must remain human-centric. Every agentic feature was built with a clear off-ramp for users to override or inspect actions. Braze’s design team ran extensive user testing to ensure agents felt like helpful assistants, not opaque black boxes. Trust is hard to earn and easy to lose—always give end users control and transparency.

10. Lead with Vision, Not Fear

Finally, Hyman reflects that the greatest challenge wasn’t technology—it was leadership mindset. He consciously chose to frame the agentic shift as an opportunity for engineers to do more meaningful work, rather than a threat to their roles. By painting a compelling vision and backing it with resources, he inspired his team to embrace change. In the agentic era, CTOs must be storytellers and coaches, not just architects.

Conclusion: Jon Hyman’s journey at Braze shows that rethinking engineering for the agentic era is both a technical and human endeavor. By embracing experimentation, restructuring teams, investing in infrastructure, and leading with empathy, any engineering organization can make the leap. The agentic era is here—and the insights from Braze’s CTO provide a practical roadmap for navigating it successfully.