The Shift from Mobile Revolution to Agentic Awakening
Jon Hyman, co-founder and CTO of Braze, has spent nearly fifteen years navigating technological upheavals. He lived through the mobile explosion that reshaped how brands reach consumers. Now he watches a new transformation unfold. The engineering landscape is tilting again, this time toward autonomous agents that write code, ship features, and even question architectural decisions. Hyman calls the current moment the agentic era, and he believes engineering teams must rethink nearly everything about how they build software.
Braze is a customer engagement platform that started in the early mobile days. Over the past year, the company transformed into an AI-first organization in just a few months. More than sixty percent of Braze’s committed code is now AI-generated. That shift did not happen by accident. It happened because Hyman and his team challenged long-held assumptions about developer workflows, code quality, and the role of human judgment.
Hyman describes himself as an “on the ground general.” He stays deeply involved with the technical underpinnings of the product. He does not sit in a metaphorical Pentagon issuing orders from a distance. He works alongside his engineers, understands the operational challenges firsthand, and makes decisions informed by real-world complexity. That perspective shaped the five ways Braze rethinks engineering for the agentic era.
1. Model Quality Outranks Mandates
When Braze began pushing AI-assisted development, some engineers resisted. They worried about code quality, security, and losing control. Hyman could have issued mandates. Instead, he let the models prove themselves.
The team focused on selecting and tuning language models that delivered consistently reliable output. They ran controlled experiments. Engineers who tried the tools on small, low-risk tasks saw the results. The code was clean, well-documented, and often better than first drafts written by humans. Word spread organically.
Hyman noticed a pattern: model quality, not executive directives, won over skeptics. When engineers saw that AI-generated code passed tests and met style guidelines, adoption followed naturally. Within months, most of the three-hundred-person engineering org was using AI tools daily.
This approach challenges the common belief that cultural change requires top-down enforcement. In the agentic era, the best tool wins. Engineers are pragmatic. Show them a model that improves their output, and they will adopt it without being told.
2. Ship Trust Before You Ship Scale
One of the most telling moments came when Hyman watched his team ship an MCP server six weeks ahead of schedule. That early win built confidence across the organization. It proved that agents could handle real production work, not just prototypes.
The lesson is simple: start small, prove value, then expand. Braze did not roll out AI across every team at once. They identified a few isolated challenges where agents could contribute quickly. The MCP server was one example. Once the team saw it work in production, the skeptics softened.
Hyman warns against the temptation to go big immediately. “Vibe-coding your way to scale is folly,” he says. It is easy to generate thousands of lines of AI-written code in a weekend. It is much harder to ensure that code runs reliably at scale, handles edge cases, and can be maintained by humans later.
In the agentic era, engineering leaders must resist the urge to ship everything at once. Build trust incrementally. Let each successful deployment pave the way for the next.
3. Measure Business Value, Not Just Developer Velocity
Many teams celebrate AI tools by counting lines of code or pull requests per day. Hyman finds those metrics hollow. The real question is whether AI generates measurable business value.
Braze tracks how AI contributions affect customer outcomes. Does a feature built partially by agents reduce churn? Does it increase engagement? Does it lower the cost of serving a customer? These questions matter more than output volume.
One challenge Hyman identifies is the cost of inference at scale. Running large language models for every code suggestion, every review, every test generation adds up quickly. Braze monitors inference costs closely and optimizes model selection for each use case. A lightweight model might handle a simple refactoring task, while a more expensive model only runs when complex reasoning is needed.
Measuring business value also requires a shift in how engineering managers evaluate performance. Instead of asking “How many features did you ship?” they ask “What measurable impact did those features create?” This shift aligns perfectly with agentic era engineering, where the goal is not more code, but better outcomes.
4. Rethink the Human-Agent Partnership
Hyman emphasizes that agents are not replacements for engineers. They are collaborators. The best results come when humans and agents work together, each doing what they do best.
Autonomous agents can now build features overnight. An agent might draft the core logic, write unit tests, and even create documentation. But the human engineer must still review the design, ensure it fits the larger architecture, and verify that edge cases are handled. The human sets the direction and catches subtle mistakes the agent misses.
Braze encourages engineers to treat agents like junior teammates. You would not give a junior developer full autonomy on a critical payment system. You would assign smaller tasks, review their work closely, and gradually increase responsibility. The same logic applies to agents.
This partnership model requires a new skill set. Engineers must learn to write clear prompts, verify agent output efficiently, and recognize when an agent is going down the wrong path. Hyman calls this “prompt engineering as a core competency.” In the agentic era, being good at coding is not enough. You must also be good at teaching an agent to code alongside you.
You may also enjoy reading: PIC Technology: How EMCORE Revolutionizes Inertial Navigation.
5. Stay Technical, Stay Grounded
Hyman’s leadership philosophy centers on staying hands-on. He still dives into code reviews, debates architectural trade-offs, and experiments with new tools. He believes that CTOs who retreat to strategy alone lose touch with reality.
When Braze transitioned to an AI-first team, Hyman got his hands deep in the AI systems. He evaluated models, tested prompts, and watched how agents behaved in production. This direct involvement gave him the credibility to guide the organization. When he spoke about the benefits of AI, engineers knew he had seen the results himself.
The same principle applies to the agentic era. Leaders who understand the technical details can make better decisions about where to invest, when to trust an agent, and how to mitigate risks. Hyman advises engineering leaders to resist the pull of the executive suite. Sit with the team. Write some code. Debug a tricky issue. That operational understanding pays dividends when you need to rally the organization around a new approach.
The Hard Truths About Agentic Era Engineering
Hyman does not sugarcoat the challenges. The cost of inference at scale is real and growing. Teams that treat AI as a free resource will burn through budgets quickly. Measuring business value requires discipline and a willingness to say no to flashy but low-impact uses.
Another hard truth is that vibe-coding — generating lots of code without deep understanding of how it works — produces brittle systems. An agent might produce code that passes tests but crashes under real-world load. Braze invests heavily in observability and testing to catch these failures before they reach customers.
Security is another concern. Malicious actors can also use agents to write exploit code. Braze has tightened its code review process and added automated security checks that run on every AI-generated commit. The goal is not to block agents but to ensure they operate within safe boundaries.
Finally, Hyman notes that the pace of change will only accelerate. Agents that build features overnight are here. Soon, agents will manage deployments, monitor systems, and even trigger rollbacks. Engineering organizations must build the cultural and technical infrastructure to handle that pace without breaking.
What Comes Next for Braze
Braze is not slowing down. The team continues to push the limits of what agents can do. They are exploring multi-agent systems where different agents handle different parts of the stack — one for frontend changes, another for backend logic, a third for database migration.
Hyman is particularly excited about agents that can learn from production incidents. Imagine an agent that watches how a human debugged an outage, then applies that knowledge to prevent similar issues in the future. That kind of institutional memory could transform incident response.
He also expects the role of engineering managers to evolve. Managers will spend less time assigning tasks and more time guiding agents, reviewing their output, and ensuring alignment with business goals. The human touch remains essential for strategy, creativity, and ethical oversight.
For Hyman, the agentic era is not a threat to engineering careers. It is an opportunity to focus on higher-value work. Engineers who embrace agentic era engineering will find themselves solving more interesting problems and building better products faster.
The key is to approach the shift with the same mindset that took Braze from startup to global leader: stay curious, stay technical, and let quality speak louder than any mandate.
