Dialogue with frontline architecture expert Christian Ciceri: Subverting traditional perceptions, in the eyes of top architects, what determines the upper limit of a career is not technical ability

In the context of the rapid evolution of software development intersecting with the wave of artificial intelligence, the role and methodology of software architects are undergoing unprecedented tests and transformations.

In the past, the core responsibilities of architects mainly focused on system design, module division, and technical decision - making, emphasizing stability, maintainability, and control over the technology stack. However, with the popularization of cloud - native architectures, microservices, large - scale distributed systems, and low - code/no - code platforms, the complexity of software systems has increased exponentially. Architects are now facing not only the issue of technology selection but also the challenge of maintaining architectural health and team efficiency in an environment of rapid iteration and continuous delivery.

Meanwhile, the rise of artificial intelligence has injected unprecedented tools and capabilities into software development. Technologies such as automated code generation, intelligent testing, and AI - assisted design enable some traditional architectural tasks to be completed quickly by algorithms. Architects no longer need to draw all dependency graphs or manually analyze performance bottlenecks themselves but can use AI to quickly identify potential problems and optimization opportunities. While this change improves efficiency, it also poses new questions: Decision - making power, system understanding, and technical judgment still highly depend on human experience and insight. How to ensure that AI tools assist rather than replace humans has become the core issue that architects must face.

It can be said that software architects are in a transformation period of integrating technical capabilities, business understanding, and data - driven decision - making. In a rapidly iterating environment, they need to master new technologies, understand the evolution laws of complex systems, and guide the team to form a shared vision and architectural culture.

Recently, InfoQ exclusively interviewed Christian Ciceri, the co - founder of Apiumhub and a well - known software architecture expert, to lead readers to deeply explore the concept of "measurable and evolvable architecture" and the impact of AI and modern software engineering tools on architectural practice based on the practical experience of front - line architects. This interview not only reviews Ciceri's personal growth process but also provides rich wisdom in architectural practice, enabling readers to understand how to maintain architectural quality and team adaptability in a rapidly changing technological environment.

Ciceri's career path is quite representative: He accumulated experience from front - line software development and architectural design practices and witnessed common challenges in large enterprises, such as lack of flexibility, long delivery cycles, and low process efficiency. In 2014, he co - founded Apiumhub with Evgeny Predein in Barcelona, determined to closely integrate agile methodology and software architecture into the core of business operations. It was through long - term practice that Ciceri gradually formed the concept of "measurable and evolvable architecture" and condensed this concept in his book "Software Architecture Metrics". He emphasizes that building a stable and adaptable system can not only improve the quality of software delivery but also ensure that the system grows in sync with business needs.

In the interview, Ciceri elaborated on "Observability" and "Architectural Governance". He pointed out that the quality attributes of a system during runtime are only part of the overall quality. True architectural governance requires continuous monitoring of all software attributes. A good way to achieve this process is to adopt relevant technologies in "Evolvable Software Architecture", and the most important concept among them is fitness functions.

Another highlight of the interview was Ciceri's insight into the role of AI in software architecture. He clearly stated that AI can assist in analyzing indicators and providing possible improvement solutions but cannot replace human judgment and decision - making. He emphasized: "Only when driven by humans can artificial intelligence become a real productivity tool in software design, rather than the other way around." Regarding various AI - generated architectural suggestions currently available, he still positions them as "assistants" rather than "partners" and reminds architects to remain rational when embracing AI.

From Architect to Founder

InfoQ: Hello, Mr. Ciceri. After reading your book, I learned that you have always been an advocate of "measurable and evolvable architecture". Could you first talk about your experience of growing from a front - line software architect to the co - founder of Apiumhub and writing "Software Architecture Metrics"?

Christian: The development of my career benefits from years of practical experience in software development and architectural design in large enterprises. During this period, I encountered common challenges in the industry, such as lack of enterprise flexibility, long delivery cycles, and low process efficiency.

In 2014, I co - founded the Apiumhub company with Evgeny Predein in Barcelona. Our core goal is to place agile methodology and software architecture at the core of business operations.

As time passed, I gradually realized that to create long - term value, we must focus on quantifiable and evolvable architectural design. Building a system that is both stable and adaptable enables the team to deliver high - quality software and ensures that the software can grow in sync with business needs. This concept ultimately prompted me to write the relevant chapters in "Software Architecture Metrics".

InfoQ: In this book, I noticed that many architects emphasize "vision", while you emphasize "measurement" and "indicators". Why is measurability so crucial in modern software architecture?

Christian: Indicators and measurements are effective ways to make disputes more objective. However, it should be noted that the architectural vision shared with the team is still an extremely important part of the daily work of software architects.

InfoQ: AI tools are automating some design, coding, and testing work, but architecture often still depends on human judgment. Which aspects of architectural design do you think can be truly enhanced by AI? Which should still be led by humans? In the context of the rapid development of AI - assisted development and intelligent engineering tools, what new changes are there in your architectural philosophy?

Christian: Although artificial intelligence is expected to be a very useful assistant in the work of software architects, I really think it cannot replace the technical decision - making process. Technical decisions must be led by human capabilities and experience.

In other words, I firmly believe that only when driven by humans can artificial intelligence become a real productivity tool in software design, rather than the other way around.

InfoQ: Nowadays, there are more and more AI - generated architectural suggestions (such as microservice decomposition and dependency graph optimization). Do you think these tools are currently reliable enough to be "architectural partners" or are they still at the "assistant" stage?

Christian: In my opinion, these tools are helpful in proposing possible solutions and exploring new possibilities, but as I said before, they cannot replace the human decision - making process. In my opinion, these tools are and will always be valuable "assistants".

InfoQ: Apiumhub cooperates with many large enterprises in architectural transformation. In these projects, where do you think the biggest resistance comes from - the technical level or the cultural level? How can a sustainable "architectural culture" be established within an enterprise, rather than just setting up an "architecture department"?

Christian: In an old article by Martin Fowler titled "Who Needs Architects?", he distinguished two roles of traditional architects - decision - makers and guides, and the latter is the best way to "improve the team's level". This approach helps to establish a real software architectural culture within the team, working side by side with developers, especially during the modeling stage.

However, I firmly believe that software architecture is an evolving science within the broader field of engineering. I mean, anyone can become an architect, but this is not something that can be done casually. To become an effective architect, you need to study a large amount of scientific literature and continuously deepen your understanding of the field.

Continuous Architecture and Architectural Observability

InfoQ: How can "observability" be integrated into system design so that architectural quality is no longer just a document indicator but is real - time visible and verifiable?

Christian: "Observability" usually refers to the quality attributes of a system during runtime, which is only part of the overall system quality. Generally speaking, introducing architectural governance or architectural design to a system means continuously controlling all software attributes. A good way to achieve this process is to adopt relevant technologies in "Evolvable Software Architecture", and the most important concept among them is fitness functions.

InfoQ: Facing complex ecosystems such as multi - AI agents, low - code components, and distributed systems, how do you suggest architects build an "observable toolbox" at the architectural level? What signals can remind the team that their architecture is becoming more and more "unobservable" and difficult to evolve?

Christian: When we think about "observability" from the perspective of architectural governance, an architectural error should ideally be detected through failed architectural unit tests. However, the signs of architectural degradation usually appear gradually, such as a slowdown in development speed, an increase in defects and runtime problems (such as performance issues), difficulties for the system to handle higher loads, and other similar symptoms.

InfoQ: In your book, you proposed to connect architectural quality attributes (such as scalability, maintainability, and modularity) with specific quantifiable indicators. How well has the team accepted this concept in practice?

Christian: According to our experience, this actually depends on the specific situation of the team. Generally speaking, indicators should not be forcibly set as goals but should be carefully introduced and promoted in combination with the team's cultural construction - this is the real starting point. Another key point is that the use of indicators should be based on real and recognized pain points, that is, problems that all team members can feel and unanimously agree need to be focused on.

InfoQ: What architectural indicators have you seen misused or misunderstood in actual projects? Can you give a specific example to illustrate how architectural indicators help discover "architectural degradation" problems in the system or guide architectural refactoring decisions?

Christian: I think the most misused indicator is test code coverage. The problem is that this indicator can hardly tell us whether the test strategy is really effective - the test effect mainly depends on the design quality of the tested modules (such as classes, methods, etc.). However, when the code coverage figure is very low, it is still a useful signal because it usually clearly reflects insufficient team productivity or problems in the development process.

InfoQ: With AI entering the field of software analysis, is it possible to have an "intelligent architectural monitoring" mechanism that not only observes system indicators but also automatically optimizes them?

Christian: No, in my opinion, this still belongs to the category of "science fiction". Although artificial intelligence can indeed help in analyzing indicators and proposing potential improvement suggestions, it cannot replace human judgment or the meticulous decision - making process required in software architecture. Software architecture involves trade - offs, understanding the business context, and predicting future needs, which are currently difficult to fully encode and implement in an automated system.

The Golden Survival Rules for Architects in the AI Era

InfoQ: In your opinion, what are the most important qualities of an excellent architect today? Is it analytical ability, leadership, empathy, or curiosity?

Christian: All of these three qualities, including analytical ability, leadership, and empathy, are very important. Analytical ability is of course the key to understanding complex systems and making sound architectural decisions. But I think curiosity is equally precious - it drives continuous learning, helps you continuously master new technologies and practices, and often leads to creative solutions that cannot be obtained by analysis alone. In many ways, it is curiosity that enables architects to grow and adapt in a constantly changing field.

InfoQ: If you're willing, could you recommend one or two books or resources that you think are or will influence "architects' thinking" to our readers?

Christian: All well - known books on domain - driven design (DDD) are very valuable. However, if you want to gain more in - depth and contemporary insights into software design, I recommend reading Vladik Khononov's "Balancing Coupling in Software Design".

InfoQ: If there is a "golden rule for architects" that still applies in the intelligent era, what would you say it is? Why do you say so?

Christian: The core "golden rule" in our industry is: Remove the "I" from "architecture". Architecture is a shared vision, and you cannot make decisions based solely on your own understanding of the domain. Real architectural work should involve the entire team to ensure that all decisions are collective.

This article is from the WeChat official account "InfoQ" (ID: infoqchina), author: Dongmei, published by 36Kr with authorization.