-
When Facilitation Fails_ What to Do Next
When facilitation fails, it can feel like the entire process is unraveling. Whether you’re leading a meeting, workshop, or team collaboration, moments where facilitation doesn’t go as planned are inevitable. However, this doesn’t mean all is lost. It’s simply a cue that the facilitation approach needs to be reassessed, or perhaps even restructured. Knowing how
-
When Architecture Needs to Be Re-Discussed
There are various situations when architecture needs to be revisited or re-discussed within a team or organization. Whether it’s due to evolving requirements, technical debt, or a shift in team dynamics, here are some common scenarios that necessitate re-engagement with the architecture. 1. Changing Business Requirements As businesses evolve, so too must their technical solutions.
-
When Architecture Isn’t a Priority—But Should Be
In many organizations, architecture is seen as a secondary concern—something that comes into play after product development is underway or when there’s a crisis. The reality, however, is that architecture isn’t just a necessary technical layer that underpins systems; it shapes every decision, influences team dynamics, and impacts the long-term viability of any software project.
-
When Architecture Is Too Important to Be Centralized
Architecture is often viewed as a top-down discipline, where decisions are made by a small group of architects, designers, and planners at the highest levels. While centralized control has its place, there are compelling arguments for why architecture should not always be centralized, especially in complex projects or contexts that require flexibility, inclusivity, and long-term
-
What to monitor in machine learning systems in production
When machine learning (ML) systems are in production, monitoring becomes crucial to ensure they continue to operate effectively, efficiently, and without issues. The following are key areas to monitor: 1. Model Performance Metrics Accuracy: Track the overall correctness of the model’s predictions. Precision and Recall: For classification problems, monitor how well the model identifies true
-
What to Do When No One Wants to Own Architecture
When no one wants to own architecture, it can be a challenging and disheartening situation for the profession and the built environment as a whole. Ownership in architecture is multifaceted—ranging from the conceptual ownership of ideas to the tangible ownership of physical spaces. The current state of the architectural field shows a complex interplay between
-
What mutual respect looks like in AI user flows
Mutual respect in AI user flows is about designing systems that prioritize human dignity, autonomy, and understanding in every interaction. It involves ensuring that AI not only responds intelligently but also acknowledges and adapts to the user’s preferences, needs, and emotional states. Here’s how this can manifest in AI design: 1. Active Listening and Understanding
-
What makes an ML model truly production-ready
For a machine learning model to be considered truly production-ready, it needs to meet several key criteria that ensure it functions reliably, efficiently, and effectively in a live environment. Here are the essential components: 1. Robustness and Reliability Resilience to Failures: The model should be able to handle unexpected scenarios without crashing, such as missing
-
What it means to be good ancestors in AI development
Being “good ancestors” in AI development means designing and building artificial intelligence with a long-term perspective, taking into account not just immediate benefits but also the well-being and ethical considerations of future generations. It’s a mindset that prioritizes sustainability, responsibility, and inclusivity in AI creation, ensuring that the technologies we develop today don’t inadvertently harm
-
What inclusive AI design looks like in multilingual societies
In multilingual societies, inclusive AI design is all about ensuring accessibility, fairness, and cultural sensitivity while accommodating diverse languages, dialects, and communication styles. The aim is to build AI systems that can be used by a broad range of people, respecting their linguistic and cultural backgrounds. Here are some key aspects of inclusive AI design