By Hatta Zainal | Head of Product Engineering, Be U by Bank Islam
Three years ago, if you told me our engineering team would spend more time setting boundaries than writing code, I would have laughed. Today, it is the most important thing we do.
At Be U, Malaysia's first cloud-native Islamic digital bank, we went through every phase of modern engineering: writing every line by hand, then pair-programming with AI copilots, then using agentic workflows that build and test for us.
But something has been shifting. Quietly, without a name. And I think it is time we gave it one.
I am calling it Ambient Engineering.
The Journey: How Bank Islam Engineering Evolved
Phase 1: Traditional Engineering (The Craftsmanship Era)
When we first built Be U, every microservice was handcrafted. Every deployment was a ceremony. We wrote the code, tested the code, deployed the code, and watched the dashboards nervously. The engineer was the craftsman, deeply involved in every detail, every line.
This was necessary. We were building a digital bank from scratch inside a regulated institution. You cannot move fast and break things in banking. You move carefully and build things that last.
But it did not scale. Not for the ambition we had.
Phase 2: AI-Assisted Engineering (The Copilot Era)
Then came the copilots. A wave of AI tools that changed how our engineers worked every day. Suddenly, boilerplate disappeared. Code reviews got faster. Junior engineers produced work that looked like it came from mid-levels.
We saw real gains. But nothing really changed at the core. Humans still made every decision. AI was a faster pair of hands, not a different way of thinking. The engineer was still the builder. AI just gave them better tools.
Phase 3: Agentic Engineering (The Orchestration Era)
This is where most of the industry sits today, and where we have been for the past year. In agentic engineering, AI is no longer just helping. It is doing the work. We define tasks, and AI agents write services, generate tests, draft deployment configs, and even suggest architecture changes.
The engineer becomes a conductor. You are running an AI orchestra. Agent A writes the service. Agent B runs the tests. Agent C flags security issues. It is powerful. It saves a lot of time.
But here is what we noticed: we were still the bottleneck. Every agent waited for our instruction. Every workflow started with our prompt. We were faster, yes. But we were still the one slowing things down.
That observation changed everything.
Phase 4: Ambient Engineering (The Emergence Era)
What if the system did not wait for you?
What if, instead of directing agents, you designed the environment where smart systems continuously build, test, deploy, monitor, and improve themselves, all within boundaries you set?
That is the shift. In ambient engineering, you do not build software. You grow the ecosystem where software lives and evolves. Your engineering becomes ambient. Always there, always shaping, but never directly controlling.
Think of it this way. Agentic engineering is conducting an orchestra. Ambient engineering is designing the concert hall, the acoustics, the seating, the atmosphere, and letting the music happen on its own.
The Key Components of Ambient Engineering
Through our exploration at Be U, we found five pillars that make ambient engineering possible.
1. Intent Architecture
The engineer's main output is no longer code. It is intent. You define what the system should achieve, not how it should do it. For us, this means turning product goals, customer outcomes, and regulatory requirements into clear specs that AI systems can understand and act on.
For example, the ambient approach would not say "build a service that tracks user activity." It would say: "customers should feel progress in their journey, the system must follow regulatory guidelines, and the experience should adapt to each person's behavior." The system figures out the architecture.
2. Governance Boundaries
This is where regulated industries like ours have a real advantage. We already think in boundaries. BNM regulations, Shariah compliance, data sovereignty rules. These are not things that slow us down. In ambient engineering, they become the framework that AI systems must work within.
Our Shariah compliance layer is a good example. You do not manually check every transaction. You encode the principles (no riba, no gharar, no maysir) as system-level rules. Every service, whether built by a human or by AI, must follow them. The boundary is ambient. Always there, always enforced. Invisible when things are fine. Immediate when they are not.
3. Observable Emergence
When systems build themselves, you need to see what they become. Observability shifts from "watching what we built" to "understanding what emerged." You need the right tools to watch intelligent systems grow in real time.
The key question changes too. It is no longer "did the deployment work?" It becomes "is the system moving toward the goal we set?" You are watching for drift, for unexpected patterns, for behaviors that are either smart or dangerous.
4. Continuous Boundary Evolution
Boundaries do not stay the same. As the system evolves, the boundaries must evolve too. This creates a loop: the system works within boundaries, produces results, the results tell us how to adjust the boundaries, and the system adapts. The engineer's ongoing job is to keep this loop healthy. Tighten rules where risk shows up. Loosen them where good ideas are being blocked.
5. Human-in-the-Loop Governance (Not Execution)
The human never leaves the picture. But what the human does changes completely. You are not in the execution loop anymore. You are in the governance loop. You review what the system has become, not what it is building right now. You step in at the strategy level, not the task level.
This matters a lot in regulated financial services. The regulator does not need an engineer watching every deployment. The regulator needs confidence that the governance framework is solid, that boundaries are enforced, and that the system behaves the way it should. Ambient engineering is built for exactly this.
Our Observations So Far
We are not fully in the ambient engineering era yet. Nobody is. But the journey has taught us a few things worth sharing.
The hardest shift is in the mind, not the tech. Engineers who spent their careers getting good at code find it hard to accept that their biggest contribution now is setting constraints, not writing logic. It feels like giving up control. But actually, it is the highest form of engineering. You are designing systems that can do more than you ever could alone.
Regulated industries will lead this shift. This surprised us. We thought heavily regulated environments would be the last to try this. Instead, we found that years of compliance work (encoding rules, checking behavior, making sure outcomes are correct) gave us exactly the skills ambient engineering needs. Banks, insurance companies, healthcare. They already think in boundaries. They just need to apply that thinking to AI-native systems.
Observability becomes your most important investment. When you stop building every part yourself, your ability to see what is happening becomes everything. In an ambient world, observability is how you understand what your system is turning into.
Small teams benefit the most. Ambient engineering is a great equalizer. It lets small, focused teams handle very complex systems. Not by working harder, but by designing environments where complexity takes care of itself.
Why We Keep Going Down This Path
The honest answer: because the other option does not work long term.
The number of services will grow. Complexity will get deeper. Customer expectations will move faster. Regulations will multiply. No amount of hiring, no number of AI agents, will keep up if the model stays "humans direct, machines follow."
We need systems that take care of themselves. That fix themselves. That get better on their own. Not in some science fiction way. In a practical, measurable, regulated-banking way. We need our services to adjust based on real traffic. We need resources to shift based on what the system observes. We need compliance rules to spread automatically when BNM puts out new guidelines.
That is ambient engineering. Not a buzzword. A necessity.
What We Are Looking Forward To
We are early. The tools are not mature yet. The patterns are still forming. But we can see the shape of what is coming, and it gets us excited.
A new kind of engineer. The engineers who will do well in the ambient era think differently. They are part systems designer, part governance architect, part ecosystem thinker. They do not ask "how do I build this?" They ask "what conditions let the right system emerge?" We are already shaping our training programs around this mindset. Not teaching engineers to use AI tools, but teaching them to think in environments, boundaries, and emergence.
Islamic finance as a testing ground. Shariah compliance is one of the most principled, boundary-driven frameworks in global finance. We believe Islamic digital banking will be one of the first places where ambient engineering fully works. The governance framework already exists. We just need to make it native to intelligent systems.
APAC as the frontier. Southeast Asia is building digital financial services at amazing speed. New digital banks, new regulations, new customer expectations, all showing up at the same time. This is the perfect environment for ambient engineering. The companies that learn to grow software ecosystems instead of just building software products will lead the next decade.
From engineering teams to engineering environments. The way teams are organized will change. Instead of squads built around features, we will see teams built around system boundaries, emergence patterns, and intent. The "ambient engineer" is not just a new title. It is a new way of working.
A Final Thought
Every generation of engineering is defined by what the engineer lets go of.
Traditional engineers let go of machine code. They moved to compilers.
Assisted engineers let go of writing everything themselves. They used AI help.
Agentic engineers let go of doing the work directly. They started orchestrating.
Ambient engineers let go of orchestration itself. They embrace emergence.
It is the hardest thing to release. And I believe it is the most important.
We are not there yet at Bank Islam. But we can see it from here. Every service that heals itself, every boundary that enforces itself, every system that adapts without being told, brings us closer.
The future of engineering is not about building more. It is about growing better.
Welcome to the ambient era.
Hatta Zainal is Head of Product Engineering at Be U by Bank Islam. With over two decades in digital banking and fintech, he writes about AI, engineering leadership, and regulated financial services in APAC.
#AmbientEngineering #DigitalBanking #IslamicFinance #EngineeringLeadership #AI #FutureOfEngineering #Fintech #APAC #BankIslam #BeU
Top comments (0)
Some comments may only be visible to logged-in visitors. Sign in to view all comments.