Platform Engineering in the age of Generative AI - NDC Copenhagen 2026

Establishing a platform that serves the needs of your developers is a challenge for teams of all sizes. But with Generative AI tools like GitHub Copilot, surely this is now all easy, and all we need is a few prompts, right?

Well, no, not really. Your onboarding processes are probably poorly documented, the tools your developers use on a day-to-day basis vary wildly between teams, and the way your teams deploy to production vary from dated bash scripts to crossed fingers over beers on Friday. Hardly the solid context needed for generative AI to actually be helpful.

In this talk, we’ll explore how platform engineering and generative AI can work together to improve developer experience, if you get the foundations right. We’ll look at common organisational and cultural challenges that limit productivity, and how to start addressing them with simple, practical steps. From there, we’ll show how “everything-as-code” patterns and AI-assisted tooling can accelerate platform maturity and reduce friction across teams.

Whether you’re a platform engineer, developer, or tech leader, you’ll leave with a clearer view of how to build a platform that AI can actually help with, and how to make your developers happier in the process.

Transcript

1. Will Velida, Senior Software Engineer, Microsoft - @willvelida PLATFORM ENGINEERING IN THE AGE OF GEN AI
2. Senior Software Engineer at Microsoft, working in the Industry Solutions Engineering team. Previously a Microsoft Azure MVP focusing on Kubernetes and Open Source. Former rugby player (now I get my injuries through on- call, being old, and screaming at my IDE/terminal). Conference speaker and YouTuber: https://www.youtube.com/@willvelida https://bsky.app/profile/willvelida.com About Me!
3. Please don’t tell my boss! :D WARNING: THIS TALK MAY CONTAIN MY OWN OPINIONS…..
4. PLATFORM ENGINEERING IS A PRACTICE BUILT UP FROM DEVOPS PRINCIPLES THAT SEEKS TO IMPROVE EACH DEVELOPMENT TEAM’S SECURITY, COMPLIANCE, COSTS, AND TIME-TO- BUSINESS VALUE THROUGH IMPROVED DEVELOPER EXPERIENCES AND SELF-SERVICE WITHIN A SECURE, GOVERNED FRAMEWORK. IT'S BOTH PRODUCT-BASED MINDSET SHIFT AND A SET OF TOOLS AND SYSTEMS TO SUPPORT IT. https://learn.microsoft.com/en-us/platform-engineering/what-is-platform-engineering
5. Platform Engineering Tools Help devs be self-sufficient Assist with common tasks Encapsulate common practices and patterns into reusable blocks Provide early feedback on risks Simplify operations by managing underlying infra and tools
6. Focuses on company’s internal development practices. Developers are primary customer of IDP Tools provide sufficient automation that ensure stakeholder requirements are met Centralize and scale knowledge across you DevOps lifecycle to reduce cognitive load. Internal Developer Platforms
7. The developer experience is transforming
8. Developers are super fast! AI writes all the code for us! Ship features overnight Poorly documented processes Inconsistent tooling across teams Cross-fingers solutions (--yolo) THE REALITY THE PROMISE Just use AI Bro…
9. • Cui et al. 2024 (4,867 devs): +26% PRs/week; no significant gain for senior devs: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4945566 • METR July 2025: devs forecasted 24% increase in speed, but experienced 19% slowdown: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/ • CodeRabbit (Dec 2025): AI Code has ~2.74x more security vulnerabilities: https://www.coderabbit.ai/blog/state-of-ai-vs-human-code-generation-report • Stack Overflow 2025: “AI Solutions almost right, but not yet” – 66%: https://survey.stackoverflow.co/2025/ai#2-ai-tool-frustrations What the data actually shows
10. Vibe coders…..
11. BURN ALL THE SUPPLY CHAINS DOWN!
12. Just a little clean up…..
13. VIBE CODERS HYPER VELOCITY ENGINEERING
14. Mike Lanzetta, DS Discipline Lead, North Asia, Microsoft ISE TO ME, I SEE THE DISTINCTION BETWEEN HYPERSPEED AND HYPERVELOCITY TO BE QUITE IMPORTANT. VELOCITY IS A VECTOR – IMPLYING BOTH SPEED AND DIRECTION. IT’S NOT ABOUT DEVELOPING CODE FAST, IT’S ABOUT DEVELOPING QUALITY CODE WITH SOLID ENGINEERING MORE QUICKLY. THIS ISN’T VIBE CODING, IT’S A TEAM WORKING TOGETHER WITH AI HELP TO EXECUTE MUCH MORE EFFECTIVELY.
15. Finding Data Use AI at work 90% Report Productivity Gains >80% Little/no trust in AI Output ~30% DORA 2025: “AI is an amplifier, not a silver bullet”
16. Ensuring consistent behavior across developers in your organization Different developers may use different coding assistants. My agents work on my machine!
17. Announced Dec 9, 2025 as part of the Linux Foundation Co-founded by Anthropic, Block, OpenAI Supported by Microsoft, AWS, Google, Cloudflare, Bloomberg Agentic AI Foundation
18. Model Context Protocol AGENTS.MD files Agent Skills The AI-Ready Codebase Stack
19. The “USB-C for AI Agents” Model Context Protocol (MCP) is an open protocol that standardizes how apps provide context to LLMs. MCP provides a standardized way to connect AI models to different data sources and tools. This standardization provides us with numerous ways to provide context to LLMs Some examples include enhancing chat interfaces, AI-powered IDEs, or custom AI workflows. MCP (Model Context Protocol)
20. The README for Agents! Contains extra context coding agents need, such as build steps, tests, and conventions. Precise, agent-focused guidance that complements existing README and docs https://agents.md AGENTS.md
21. Agents are getting more capable, but lack context to do the work reliably. Agent Skills are folders of instructions, scripts, and resources that agents can discover and use to do things more accurately and efficiently. Skills solve this by giving agents access to procedural knowledge and company context. Agents with access to a set of skills can extend their capabilities based on the task https://agentskills.io/home Agent Skills
22. Domain expertise New capabilities Repeatable workflows Interoperability What can Agent Skills enable?
23. Tool Config Files Notes GitHub Copilot .github/copilot-instructions.md + .github/instructions/*.instructio ns.md + *.prompt.md Reads AGENTS.md natively; ‘/skills’ command Claude Code CLAUDE.md + .claude/skills + .claude/agents/ + hooks + plugins Agent Teams Feb 2026; no native AGENTS.md Cursor `.cursor/rules/*.mdc` (Always / Auto-Attached / Agent- Requested / Manual) `.cursorrules` **deprecated**; native Skills + AGENTS.md Windsurf `.windsurf/rules/rules.md` + Memories AGENTS.md native Codex `AGENTS.md` + `.agents/skills/` Native Skills Dec 2025 Copilot Custom Instructions
24. SPEC-DRIVEN DEVELOPMENT
25. Spec-Driven Development “SPEC-DRIVEN DEVELOPMENT FLIPS THE SCRIPT ON TRADITIONAL SOFTWARE DEVELOPMENT. FOR DECADES, CODE HAS BEEN KING — SPECIFICATIONS WERE JUST SCAFFOLDING WE BUILT AND DISCARDED ONCE THE "REAL WORK" OF CODING BEGAN. SPEC-DRIVEN DEVELOPMENT CHANGES THIS: SPECIFICATIONS BECOME EXECUTABLE, DIRECTLY GENERATING WORKING IMPLEMENTATIONS RATHER THAN JUST GUIDING THEM.”
26. Intent-driven development where specifications define the "what" before the "how" Rich specification creation using guardrails and organizational principles Multi-step refinement rather than one-shot code generation from prompts Heavy reliance on advanced AI model capabilities for specification interpretation github/spec-kit: Toolkit to help you get started with Spec-Driven Development GitHub SpecKit
28. Research the topic Plan the feature Implement Review RPI (Research Plan Implement) Pattern
29. Don’t commit markdown files unless they add value (README, CONTRIBUTING, SECURITY are fine) If you use AI to create specs/documents, think about where you want to share them Use MCP Servers or CLI commands in agents to create developer documentation where it matters. For example, Architecture Decision Records to Azure DevOps wikis Create specs/documents/artifacts that are consistent with your organization.
30. • Agent = Model + Harness: https://martinfowler.com/articles/harness-engineering.html • Two Control Types • Computational – Deterministic and fast, run by the CPU. Includes tests, linters, type checkers, structural analysis. Takes milliseconds to seconds to run, and results are reliable. • Inferential – Semantic Analysis, AI Code review, “LLM-as-judge”. Typically run by a GPU or NPU. Slower and more expensive, results are more non-deterministic. Agentic Harness Engineering
33. Direction Computational or Inferential Example implementations Coding Conventions Feed Forward Inferential AGENTS.md, Skills Instructions (how to bootstrap a new project) Feed Forward Both Skill with instructions and a bootstrap script Code mods Feed Forward Computational A tool with access to OpenRewrite Recipes Structural tests Feedback Computational A pre-commit hook running unit tests Instructions on how to review Feedback Inferential Skills Examples
39. AI IS AN AMPLIFIER, NOT A SILVER BULLET HIGH-QUALITY PLATFORMS AMPLIFY AI BENEFITS; WEAK ONES AMPLIFY DYSFUNCTION.
40. THANK YOU