Cloud Native Summit recently released a video from its 2025 conference in Auckland. The session includes several talks highlighting how organizations can turn their collaboration patterns into results with the Open Practice Library. Andrea Magnorsky discussed the application of library practices to promote the success of enterprise AI by Ahilan Ponnusamy and Andreas Grabner, co-authors of Cloud and Edge Technology Operating Models.
Architect and consultant Magnorsky designed a BYTE size architecture workshop when working with British broadcaster ITV, explaining it as a way to make architecture continuous and comprehensive, using regular short structured workshops to build a common understanding. Her speech, entitled “Deliberate Practices Thinking about Your Systems,” showed that a key differentiator of byte-sized architecture is to avoid enforcing complex outcomes in a single optimistic timeboxed architecture session. She explained how the session integrates architecture alignment and evolution into the rhythm of the team, born from the combination of technical and non-technical partners.
It takes 45-90 minutes. There are strategies to deal with larger groups, but there are less than 10 people in the actual session. And most importantly, it's a recurrence. Therefore, you can run a series of these sessions to achieve certain goals.
Magorsky explained that in a repetitive workshop where participants draw and share their understanding of architectural components, it becomes implicit and explicit, allowing teams to gradually capture insights, surface assumptions and create a living library of architectural knowledge that evolves along the system. She eventually said, “What we built is (an)encoding of what's in the brains of programmers.” And this inconsistency is most prominent during incidents that diverge from when the software is built and affect recovery.
Connecting with these ideas, Ponnusamy and Grabner gave a keynote address entitled Technology Operating Model for Enterprise AI Adoption, introducing a framework designed to guide AI projects into an iterative, resulting deployment of recruitment processes. Based on the open practice library, the model emphasizes stakeholder alignment, platform engineering, and incremental delivery.
Ponnusamy explained that all organizations have a technology operational model, but few are explicit. AI adoption adds urgency as companies are rapidly deploying generated AI, hybrid cloud architectures, and new platforms. He outlined the importance of using a platform engineering approach to AI recruitment.
All tools provide a single point of access, a single source of truth for initiatives, and a guardrail for security and compliance. By operating the platform as a product, we ensure that future flexibility is maintained and improved recruitment and experience.
The Ponnusamy and Grabners operating models are centered around the concept of “streams, dimensions, dimension items.” For example, AI platforms and tenant experience streams are distilled to dimensions within this domain, such as platform onboarding, lifecycle management, and AI operations. Each dimension is split into evolutionary incremental, measurable transition states comparable to milestones, leading to the target state of standardized, automated bias-aware AI operations. Ponnusamy explained it:
You need to know the starting state, define achievable transition states, and build them gradually towards the target state. For example, you might have your customers aware that AI is being used first, capture feedback, implement bias detection, and ultimately arrive at fully automated AI operations.
The technology operational model of AI adoption includes a recurrence process that assesses progress towards a target state, leveraging a mix of established lean processes from the open practice library, including stakeholder and goal adjustment, goal mapping, story mapping and value slicing, and various facilitating technologies, as shown below.
Similar to Magnorsky's BYTE-sized architecture, AI's technology operational model demonstrates that reissues sessions with small groups working together to improve architecture ownership and evolution are equally important to AI adoption. This coincides with the 2025 InfoQ Culture and Methods Trends Report, which states, “in a hurry to adopt AI, the team agreed that the space needs to be saved for human cooperation, reflection and learning.”
This was further reflected by May Xu, head of technology for APAC. Xu proposed strategies focusing on skills, AI literacy, collaborative learning, governance, experimentation, and clear playbooks. Literacy and collaborative learning are closely aligned with byte-sized architectures, but the rest fit well with technology operational models for AI adoption.
Magnoski reminded the audience that they are all “knowledge workers,” which means that they are the most effective when they work together to see the “systems from multiple perspectives” in order to bring value through successful change. She said:
The job of knowledge is to understand and apply knowledge that is all heuristics of your experience, value, context, and your organization can make effective changes.
