Transforming Your Engineering Team for the Agentic Era: A CTO's Guide

Introduction

In the fast-evolving world of technology, the shift toward agentic systems—autonomous AI agents that act on behalf of users—is redefining how engineering teams operate. Jon Hyman, co-founder and CTO of Braze, recently shared how he led his engineering organization through nearly 15 years of growth and then pivoted to an AI-first approach in just a few months. This guide distills that experience into actionable steps for any CTO or engineering leader looking to rethink their team for the agentic era. Whether you're scaling a startup or modernizing an enterprise, these steps will help you navigate the transformation with clarity and purpose.

What You Need

• Executive buy-in: Support from your CEO and board to prioritize AI-first initiatives.
• A clear vision: A defined understanding of how agentic systems will enhance your product or service.
• Cross-functional collaboration: Input from product, design, data science, and infrastructure teams.
• Technical expertise: In-house or contracted specialists in machine learning, NLP, and agent architectures.
• Modern tooling: Access to AI platforms (e.g., LLM APIs), version control, CI/CD pipelines, and monitoring systems.
• Culture of experimentation: Willingness to fail fast and learn from iterative development.

Step 1: Assess Your Current Engineering Landscape

Before you can transform, you must understand where you stand. Evaluate how your current engineering processes handle automation, data pipelines, and decision-making. Identify bottlenecks where human intervention is slow or error-prone. At Braze, Hyman began by analyzing their existing customer engagement platform—finding opportunities to embed intelligence that could proactively suggest actions. Use surveys, interviews, and data analysis to map out your team’s strengths and gaps.

Step 2: Define Your Agentic Vision

Clarify what “agentic” means for your organization. Is it about creating autonomous customer service bots? Or building internal tools that automate code reviews? In Braze’s case, the vision was to enable marketers to orchestrate campaigns using AI agents that anticipate user behavior. Write a concise mission statement that ties agentic capabilities to business objectives. Share it across teams to align efforts.

Step 3: Establish an AI-First Mindset

This is the core of the transformation. An AI-first engineering culture treats AI as a fundamental building block, not an add-on. Hyman achieved this by running internal hackathons, dedicating sprint time to AI experiments, and providing hands-on training with LLMs. Encourage your engineers to think in terms of prompts, embeddings, and agent loops. Create working groups that explore how AI can replace routine tasks and enhance complex workflows.

Step 4: Restructure Teams Around Agentic Capabilities

Traditional engineering teams organized by function (frontend, backend, data) may not optimize for agentic development. Consider forming cross-functional squads focused on agent behaviors, memory systems, and orchestration. At Braze, Hyman reorganized engineers into “AI product teams” that owned end-to-end agent features. This reduces dependencies and accelerates delivery. Each squad should include a product manager, AI specialist, and infrastructure engineer.

Step 5: Invest in Agentic Infrastructure

Agents require robust infrastructure: scalable model serving, real-time data streaming, and reliable feedback loops. Assess your current cloud architecture and toolchain. Adopt platforms like LangChain or TensorFlow Agents for rapid prototyping. Implement guardrails for safety and monitoring to detect agent drift. Braze leveraged its existing event streaming system (Braze Currents) and enhanced it with AI inference nodes. Document your stack and ensure it supports rapid iteration.

Step 6: Upskill Your Team Through Continuous Learning

Not every engineer will be an AI expert overnight. Create a learning path: start with fundamentals (e.g., introduction to neural networks), then move to agent-specific topics (e.g., multi-agent collaboration). Use internal workshops, online courses, and pair programming. Hyman emphasized that Braze’s transformation happened quickly because they invested in daily learning sessions and assigned “AI buddies” to less experienced engineers. Celebrate early wins to build momentum.

Step 7: Develop an Agentic Product Roadmap

Transition from monolithic releases to iterative agent experiments. Prioritize features that demonstrate clear ROI, such as reducing manual support tickets or increasing user engagement. At Braze, they launched a beta AI recommendation engine for email timing. Use A/B testing and user feedback to refine agent behavior. Publish your roadmap internally to manage expectations and showcase progress.

Step 8: Foster a Culture of Trust and Experimentation

Agentic systems will occasionally fail or behave unpredictably. Normalize failure as a learning tool. Hold blameless post-mortems and share outcomes across teams. Hyman encouraged engineers to present their agent experiments—even those that didn’t work—in all-hands meetings. This practice builds psychological safety and encourages bolder ideas. Set up a “sandbox” environment where teams can test agents without impacting production.

Step 9: Measure and Iterate

Define key performance indicators (KPIs) specific to agentic systems: agent success rate, user satisfaction, latency, and cost per agent interaction. Track these alongside traditional engineering metrics. Use dashboards to visualize agent performance over time. At Braze, they created an “AI Impact Board” that showed how agent-driven campaigns lifted conversion rates. Based on data, they iterated on agent prompts and models. Schedule regular retrospectives to tweak your approach.

Step 10: Scale and Share Success

Once you have a proven agentic model, scale it across product lines and regions. Automate deployment pipelines to handle multiple agent versions. Document best practices and create an internal playbook. Hyman credits Braze’s fast transformation to a “Share and Reuse” mentality—every team contributed to a central library of agent patterns. Host internal conferences or lunch-and-learns to spread knowledge. Finally, celebrate your team’s journey: recognize individuals who drove the change and build a sense of shared achievement.

Tips for Success

• Start small: Pick one high-impact, low-risk problem for your first agent—like automating status updates or triaging alerts.
• Pair AI expertise with domain knowledge: Your best agents come from engineers who understand both the business logic and the technology.
• Communicate constantly: Keep stakeholders informed about agent capabilities and limitations to avoid overpromising.
• Don’t neglect security: Agentic systems introduce new attack surfaces; implement robust authentication and data privacy controls.
• Stay adaptable: The agentic field evolves rapidly; allocate time for you and your team to stay current with research and tools.

By following these steps, you can lead your engineering organization through a transformation similar to what Braze achieved—shifting from a traditional, manually-tuned operation to an AI-first powerhouse ready for the agentic era. The journey requires commitment, but the payoff—faster innovation, happier users, and a resilient team—is well worth the effort.