Evolving the Kubernetes User Experience:​ More intuitive, more extensible, more agentic ​

Keynote presented at the Cloud Native and Open Source AI (Community Stack conference), London, June 11, 2026
Abstract: Kubernetes has become the default platform for modern workloads and increasingly for AI systems. But as adoption grows, its user experience (UX) must serve two very different audiences simultaneously: humans learning the platform, and AI agents operating within it. Grounding the discussion in a real example of deploying an agent-based system on Kubernetes, we will explore recent developments in the Kubernetes user experience and the implications for both newcomers to the platform and experienced users, with a focus on usability, governance, and production readiness.

Transcript

1. Evolving the Kubernetes User Experience More intuitive, more extensible, more agentic Community Stack Cloud Native & Open Source AI Conference 11 June 2026 BrainStation London, UK
3. London, 10 March 2016
4. Kubernetes back in 2016
5. Cloud native developers (+116% y/y) 20 million Source: CNCF + SlashData State of Cloud Native Development Report, Q1 2026
6. experienced developers hobbyists / students IT admins The NEXT 20M cloud native users
7. User Experience (UX) … is how a user interacts with and experiences a product, system or service … includes a person's perceptions of utility, ease of use, and efficiency. https://en.wikipedia.org/wiki/User_experience
8. % brew install kubernetes $ apt-get install kubernetes
10. Storage Service Mesh Networking Security Observability GitOps AI Container Management
11. “… a suggested path for companies that want to modernize their infrastructure, but nobody is using all 16 CNCF projects as part of their enterprise computing stack…” The Cloud Native Trail Map (2018)
13. A Rich, Diverse Ecosystem Lots of consistency CRDs YAML Controller pattern Helm charts ... Standalone UX (for good reasons) Specialized focus Independence and autonomy Customization Choice Is good, right?
14. Challenges of UX Fragmentation Context Switching Switching between different tools may impact the user’s productivity Inconsistency May lead to frustration or not be compatible with the users’ workflows/systems Learning Curve Users must learn how to run and use different UIs, commands, etc. Time consuming Not novice friendly
16. Don’t managed services simplify all this?
17. 10 Million 147 Million 1994 1998 Worldwide Internet Users
18. In-cluster Web Portal (a “Kubernetes Dashboard ++”) Local UI+K8s (a “Kubernetes Desktop”) Unified Management UI for Multiple Remote Clusters
20. The new Kubernetes UI, replacing Dashboard Modern Web UI with Local App option Multi-cluster management Highly customizable and extensible
21. https://headlamp.dev/
22. https://learn.microsoft.com/en-us/azure/aks/aks-desktop-overview
23. Contributing organizations, last 12 months Headlamp has been embraced by the community
24. Headlamp plugins AI Assistant Kaito Beacon Flux KubeScape KubeVirt MiniKube KubeFlow Kompose Knative OpenCost Radius Strimzi Trivy Volcano Keda Cert-manager Cluster API Inspektor Gadget Crossplane Backstage ArgoCD Karpenter Prometheus … and many more in CNCF Artifact Hub
25. Obligatory transition to talk about AI
26. Kubernetes is the primary platform for AI Source: 2025 CNCF Annual Survey 66% using Kubernetes to host GenAI workloads
27. Workloads are changing 12% 12% 12% 16% 24% 24% 28% 40% 40% 44% 48% Feature engineering Automated retraining Hyperparameter tuning Model monitoring and/or drift detection Data labelling Training large-scale models Batch model inferencing Data pre-processing Batch jobs for AI/ML pipelines Real-time model inferencing Model experimentation Source: 2025 CNCF Annual Survey
28. From Cloud Native to AI Native 2029 “New normal” of enterprise applications Agents will be created on the fly by humans, and humans and AI will collaborate in new ways. 2028 AI agent ecosystems across multiple applications A network of AI agent ecosystems will evolve to leverage agents that can act dynamically to changing scenarios. 2027 Collaborative AI agents within an application Multiple single AI agents will collaborate, with diverse skills for complex tasks within an application or data environment. 2026 Task-specific agent applications Enterprise apps will integrate task-specific agents for each application. 2025 AI assistants for every application AI assistants will be embedded in most enterprise applications (precursor to agentic AI). Source: Gartner (Emerging Tech: The Future of Agentic AI in Enterprise Applications, 2025) 40% of enterprise apps will include agents by end of 2026
29. The agentic challenge: security and trust
30. The agentic challenge: security and trust Main obstacles preventing organizations from reaching fully scaled agentic AI 9 23 28 32 34 36 38 38 52 Lack of executive support Organizational resistance Immature vendor or ecosystem landscape Unclear or insufficient business value Resource or budget constraints Gaps in responsible AI tooling and control Technical limitations Regulatory uncertainty Security and risk concerns Source: McKinsey AI Trust Maturity Survey, Jan 2026
31. Like toddlers, agents need sandboxes
32. Microsoft Agent Framework Framework for building production-grade AI agents and multi-agent workflows in .NET and Python github.com/microsoft/agent-framework Microsoft contributions to agentic open source Microsoft Agent Governance Toolkit A policy-driven runtime governance layer that enforces zero-trust controls, auditing, and safe execution for AI agent actions github.com/microsoft/agent-governance-toolkit Agent Reference Stack for Kubernetes A Kubernetes-native reference stack for easily and securely running AI agents with built-in sandboxing, governance, and execution infrastructure github.com/azure/kars
33. Agent Reference Stack for Kubernetes (KARS) Lifecycle management Execution sandbox Agent mesh Inference router Any agent OpenClaw • Hermes • OpenAI Agents SDK • Microsoft Agent Framework • LangGraph • LangGraph.js • Anthropic Claude Agent SDK • Pydantic-AI • Bring your own fleets of agents and subagents isolation, confidential, seccomp, … secure, encrypted agent-to-agent comms enforcement point for governance policy
34. Demo
35. experienced developers hobbyists / students IT admins agents AI won’t just run on Kubernetes, it will operate it!
36. • Incident triage & auto- investigation • Logs & Metrics query generation • Troubleshooting guide search • On-call knowledge transfer Example (from Microsoft): A personal AI agent running on AKS for every AKS engineer
37. What changes when the user is an agent? Machine Interfaces • APIs / CLIs • Model Context Protocol (MCP) • Documentation Policy and governance Scale and automation
38. A couple of open source MCP projects Inspektor Gadget MCP Server Azure Kubernetes Service MCP Server A powerful suite of “Gadgets” for observability, troubleshooting & more • Low-level visibility into system calls, block i/o, network, file, cpu, gpu, memory, … • Monitor/profile system activity • Monitor/advise (network policy, seccomp, …) github.com/inspektor-gadget/ig-mcp-server github.com/Azure/aks-mcp Wide range of tools for managing, diagnosing, and optimizing AKS • Cluster management • Networking & Compute • Monitoring, Control Plane, Fleet • Azure Advisor & detectors • Select Inspektor Gadget tools
39. Demo
41. Thank you! Andrew Randall linkedin.com/in/andrewrandall ahrkrak.bsky.social Pal Lakatos-Toth linkedin.com/in/lakatostothpal pallakatos.bsky.social