Humans in the Loop: Engineering Leadership in a Chaotic Industry
At QCon San Francisco 2025, the atmosphere buzzed with anticipation as Michelle Brush, engineering director of Site Reliability Engineering (SRE) at Google, took the stage. Her closing keynote, titled “Humans in the Loop: Engineering Leadership in a Chaotic Industry,” resonated with a room full of software leaders on November 19, 2025. Michelle’s insights explored the dynamic shifts occurring within software engineering, systems thinking, and the nature of leadership amid an increasingly complex landscape.
Navigating Uncertainty in Technology
Michelle opened her keynote by addressing a common sentiment among practitioners: uncertainty. This unease is a shared experience in today’s fast-paced technological environment. Drawing on her contributions to “97 Things Every SRE Should Know,” she emphasized that while the nature of software engineering is evolving, it is far from disappearing. With the rise of AI and automation in software development, engineers now face an array of sophisticated challenges that demand new skill sets and mindsets.
“Large language models are, by their nature, unconsciously competent—they know a lot, but they can’t explain it,” Michelle articulated. This statement encapsulates a core issue many engineers face as they integrate AI into their workflows.
The Dilemmas of Automation
Citing Lisanne Bainbridge’s influential paper, “Ironies of Automation,” Michelle explained that automating tasks often leaves humans with even more complex responsibilities. As AI takes on routine work, engineers must engage in monitoring, debugging, and validating these automated systems. Using a relatable analogy, she likened automation to dishwashers—while they simplify the washing of dishes, they don’t eliminate the need for human oversight and intervention.
“Our brains are going to start working on higher and higher abstractions,” she stated, urging attendees to recognize the cognitive shift required in modern development practices.
Understanding AI’s Limitations
In her talk, Michelle elaborated on the behavior of large language models (LLMs). These AI systems, though capable of generating impressive outcomes, operate with “unconscious competence.” They produce results but often lack the necessary transparency and self-awareness. As she put it, “They don’t know what they don’t know,” and this inherent limitation can lead to errors, or what she termed hallucinations.
In contrast, human engineers possess “conscious competence.” This means they can both understand and articulate their knowledge, a critical skill in the realms of mentorship and validation of machine outputs.
More Read
![Exploring the Generalized Information Bottleneck Theory in Deep Learning: Insights and Applications [2509.26327]](https://aimodelkit.com/wp-content/uploads/2025/10/1760666434_arxiv-logo-fb-150x150.png "Exploring The Generalized Information Bottleneck Theory In Deep Learning: Insights And Applications [2509.26327]"){kind=logo}
The Art of Chunking
A recurring theme throughout Michelle’s presentation was the concept of “chunking,” which refers to cognitive encapsulation as engineers tackle increasing complexity. She urged attendees to develop the capability to navigate between different abstraction layers while still drilling down into underlying systems.
“All abstractions leak, especially our hardware abstractions,” she reminded the audience, reinforcing the importance of awareness in system design.
Economic Pressures and Engineering Demands
Michelle addressed the current economic landscape, marked by the end of zero-interest rates and intensified competition for specialized AI hardware like GPUs and TPUs. Referencing Jevons’ paradox, she argued that as AI technologies accelerate software development, organizations will not accomplish less work—but rather, much more. This surge in demand compels engineering leaders to employ techniques such as non-abstract system design, facilitating a comprehensive understanding of compute, storage, and reliability costs.
Leadership in Complexity Management
As Michelle delved deeper, she framed reliability and complexity management as critical leadership responsibilities. Drawing upon Richard Cook’s essay “How Complex Systems Fail,” she highlighted how outages often stem from hidden interdependencies—systems that function as intended yet still fail. To mitigate these risks, she advocated for investing in “generic mitigations” such as safe rollbacks and traffic shedding to restore service swiftly even in the face of unknown failure modes.
“Experimentation, hypothesis-driven change, and deliberate mentoring are essential ways to grow the next generation of engineers," Michelle asserted, promoting a proactive approach to leadership.
Lessons from Google’s History
To conclude her keynote, Michelle shared a poignant case study from Google’s extensive experience with SRE. She recounted a significant outage in 2019 that impacted two data centers due to runaway automation. The assumption that geographic distribution alone was adequate for resilience turned out to be flawed when a third data center failed under recovery load.
The takeaway? Understanding the need for broader capacity and smarter design practices is crucial. Through initiatives like latency injection testing and intent-based rollout systems, Google has learned to uncover risks before deployment, ensuring robust and reliable services.
Embracing the Future of Engineering
As the closing keynote wrapped up, Michelle Brush left the audience with invaluable insights: the evolving landscape of software engineering necessitates a shift toward higher cognitive demands, better embrace of abstraction, and a focus on mentorship and leadership. Developers eager to learn more can follow InfoQ’s comprehensive coverage of this event, with videos anticipated to roll out in the coming weeks.
For those involved in engineering leadership, Michelle’s observations serve as a roadmap for navigating the complexities of a rapidly changing industry. Emphasizing the synergy of humans and technology, she underscored the pivotal role engineers will play in shaping the future of software development.
Inspired by: Source
