Continuous AI and Agentic AI Workflows

AI Agents are transforming software development and boosting developer productivity. Today, we achieve this interactively using tools like GitHub Copilot via the IDE (VS Code Agent Mode) or Terminal (Copilot CLI). But what it they could work on your behalf continuously in the background without you having to ask?

​In this talk, we’ll look at work being done by our colleagues at GitHub Next and Microsoft Research on a concept called Continuous AI that is powered by AI Agentic Workflows. We’ll look at what kinds of tasks this is good for, show examples of these workflows in action, and give you an intuitive sense for how you can apply these tools and concepts to your own projects. The work is not a product - but rather, an experimental platform for learning and exploring the possibilities of agentic AI in automating and managing repository actions. Learn more about the project here.

Transcript

1. Continuous AI & AI Agentic Workflows Nitya Narasimhan, PhD Senior AI Advocate Microsoft AI NATIVE DEV NYC MEETUP Using Natural Language Programming for Repository Automation Actions
2. Single model Simple POCs Point-in-time ROI Reactive monitoring Human-driven deployment Multiple point-based security solutions Deterministic logic features Manual coding – Building – Testing Multi-model Automated agentic workflows Continuous improvement Next-gen AI PLATFORM Ongoing predictive observability Dynamic AI-driven orchestration End-to-end platform-based security approach AI-agentic DEVOPS Proactive SECURITY and governance Adaptive reasoning agents Writing prompts – Creating – Evaluation Shift The way we build AI apps and agents is changing From To
3. AI Development needs end-to-end support 1 AI app and agent orchestration 2 Frontier & OSS Models 3 Knowledge and Tools 4 Observability and agent controls 5 Fine-tuning and customization 6 Deployment to local and edge
4. Microsoft Foundry Simple Trusted Powerful
5. Microsoft Foundry Simple Trusted Powerful
6. 1 Models 2 Agents 3 Dev Tools 3 Core Capabilities for Empowering Developers
7. AI Agents – In Software Dev Lifecycle Code Verify Plan Test Operate Optimize Deploy Agent mode Coding agent Spark workbench Create issues Brainstorm Spec kit Code quality Code review Autofix Migration Spark runtime Deployment App modernization SRE agent Support Investigate Metrics Dev Tools 3
8. GitHub Next – Long Term Bets, Reliable Tooling GitHub Next investigates the future of software development. Projects in various stages – from napkin sketch .. to product!
9. GitHub Spark – Create & Share Micro-Apps Build a #30DaysOfMicrosoftFoundry website with a landing page that contains a hero panel describing the site followed by a responsive card gallery with 1 card per day. Each card should have a title, photo, description and label tags -- the gallery panel should support search, sort and filter options. Clicking the card should take you to a detailed blog post for that day. Use the attached image to identify the topics for the 30 days.
10. Continuous AI – Powered By Agentic Workflows Continuous Integration (CI/CD) was rules based. Continuous AI (CAI) expands to support natural language based automation. “Describe to build” Continuous Integration was driven by GitHub Actions. Continuous AI is driven by Agentic Workflows
11. GitHub Universe – Oct 2025 Talk & Demo • Continuous Quality • Continuous Translations • Continuous Accessibility • Continuous Expts. (Doodle Jump) • Continuous Coverage (Tests) • Continuous Efficiency (Swarms) • Continuous Triage (Bug fixes) • Continuous Evolution (Agent Q)
12. gh.io/gh-aw Starting point for all AI Agentic Workflows Jonathan ”Peli” De Halleux Slide Deck on AI Agentic Workflows
13. Join The Community Let’s Build Together!