> Markdown version of [/videos/100035-developers-become-orchestrators-from-human-in-the-loop-to-spec-in-the-loop?t=1099](https://www.wearedevelopers.com/videos/100035-developers-become-orchestrators-from-human-in-the-loop-to-spec-in-the-loop?t=1099). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# Developers become Orchestrators: From Human-in-the-Loop to Spec-in-the-Loop

Treating AI as a basic copilot creates a cognitive bottleneck. Stop hand-holding algorithms. Transition to spec-in-the-loop workflows where developers orchestrate architecture and agents execute code.

- **Speakers:** [Bastian Heilemann](https://www.wearedevelopers.com/@bastian-heilemann), [Stefan Bley](https://www.wearedevelopers.com/@stefan-bley)
- **Event:** World Congress 2026 Europe
- **Published:** July 9, 2026
- **Duration:** 30:01
- **URL:** https://www.wearedevelopers.com/videos/100035-developers-become-orchestrators-from-human-in-the-loop-to-spec-in-the-loop

## Summary

Software development teams frequently hit a plateau when treating AI merely as an advanced copilot, discovering that it accelerates coding but simultaneously increases cognitive load, risk, and architectural inconsistency. Overcoming this bottleneck requires shifting from a 'human-in-the-loop' model to 'spec-in-the-loop' agentic workflows. By treating specifications as concrete execution boundaries rather than static documentation, developers can transition into orchestrators. They design the system constraints, architecture decision records (ADRs), and safety guarantees, while delegating the automated generation of implementation, CI/CD pipelines, and tests to AI.

Safely pushing low- and medium-criticality systems into fully automated delivery demands rigorous preparatory work. Agentic AI acts as an amplifier, meaning teams must establish strong guardrails—such as everything-as-code paradigms, strict linting, templates, and continuous static code verification—before delegating execution. In a practical experiment modifying a specific SLO within a Backstage-bootstrapped Java service, AI autonomously drafted the code, created given-when-then test scenarios, and prepared a deployable pull request. However, practical trials underscore the vital necessity of human approval gates for critical changes, particularly to prevent runaway models from compromising underlying Kubernetes infrastructure or altering production databases without constraint.

Adopting spec-driven development ultimately reallocates engineering time away from mundane coding toward robust prototyping, business alignment, and deep architectural problem-solving. As models mature, integrating overarching contextual memory via MCP servers and repository-wide knowledge graphs will become the industry standard for mapping out complex inter-service vulnerabilities. Eventually, the capacity to automatically orchestrate entire microservice rewrites—translating legacy stacks between languages like Java to Go without conducting manual point-by-point migrations—will shift from a theoretical exercise to a routine platform engineering strategy.

**Keywords:** spec-driven development, agentic AI workflows, copilot cognitive load, shift-left AI delegation, architecture decision records, model context protocol servers, repository knowledge graphs, infrastructure as code pipelines, static code analysis verification, automated microservice implementation, backstage developer portals, service-level objective automation, everything-as-code guardrails, LLM execution boundaries, automated software delivery

## Chapters

1. **Overview of Zeiss and the shift to agentic workflows** (00:02) — Moving software from human-in-the-loop to spec-in-the-loop requires understanding the organizational context.
1. **Digital customer experience platforms and current development processes** (01:44) — Existing platform engineering foundations like backstage templates provide the necessary context for introducing AI-driven development.
1. **Mapping ai maturity levels in software development** (03:39) — Transitioning from unstructured code completion to agentic workflows shifts the developer role towards defining intent and execution boundaries.
1. **Establishing guardrails and verification for ai agents** (05:00) — Reliable agentic execution requires structured artifacts like coding guidelines, architectural decisions, and strict verification steps.
1. **Targeting low-criticality software components for full automation** (06:26) — Focusing AI delegation on low-criticality services frees up developer time for high-value feature delivery.
1. **Structuring software specifications as a strict execution boundary** (08:19) — A well-defined specification must include intent, technical stack context, execution boundaries, and explicit success criteria.
1. **Evaluating an existing microservice for full ai readiness** (09:48) — Assessing a data translation webhook service reveals how much existing platform engineering and documentation is needed for AI delegation.
1. **Customizing ai agent skills and deploying server hooks** (12:04) — Utilizing tools like Google agent skills and MCP servers allows managing context windows and parsing repository graphs.
1. **Automating service level objective implementation with agents** (13:24) — An AI agent successfully translates a service-level requirement into implementation, tests, and pull requests without manual intervention.
1. **Key learnings and infrastructure challenges with agentic automation** (18:19) — Adopting agent skill frameworks improves code base understanding but requires careful infrastructure guardrails to prevent catastrophic errors.
1. **Balancing human system intent with automated ai execution** (22:04) — While AI handles the tactical execution of code generation and pipelines, humans must own systemic intent and non-functional requirements.
1. **Exploring complete codebase regeneration and programming language migrations** (24:24) — Regenerating entire services or migrating programming languages using AI proves feasible but requires robust connections to specific coding guidelines.
1. **Enhancing agent repository context using engineering knowledge graphs** (26:40) — Combining code search indices with platform engineering databases provides agents with the required context to navigate complex organizational structures.

## Related Moments

- [Transitioning software engineering teams to AI-native development workflows](https://www.wearedevelopers.com/videos/100087-ai-ready-what-enterprise-transformation-actually-takes) (from "AI-Ready? What Enterprise Transformation Actually Takes")
- [Shifting developer workloads and realistic AI productivity gains](https://www.wearedevelopers.com/videos/1830-wearedevelopers-live-speculaitions) (from "WeAreDevelopers LIVE - SpeculAItions")
- [Designing complex software architecture in the era of AI](https://www.wearedevelopers.com/videos/1365-wearedevelopers-live-the-weekly-developer-show-with-chris-heilmann-and-daniel-cranney) (from " WeAreDevelopers LIVE - the weekly developer show with Chris Heilmann and Daniel Cranney")
- [The limitations of spec-driven development with AI](https://www.wearedevelopers.com/videos/100210-why-optimizing-for-system-comprehension-is-key-to-implementing-ai-for-software-development) (from "Why optimizing for system comprehension is key to implementing AI for software development")
- [Designing task boundaries between human developers and AI tools](https://www.wearedevelopers.com/videos/1950-the-scrum-master-as-an-orchestrator-guiding-human-ai-collaboration-in-modern-teams) (from "The Scrum Master as an Orchestrator: Guiding Human–AI Collaboration in Modern Teams")
- [Navigating developer bottlenecks and human accountability](https://www.wearedevelopers.com/videos/100265-fireside-chat-in-conversation-with-werner-vogels-cto-of-amazon-com) (from "Fireside Chat - In conversation with Werner Vogels, CTO of Amazon.com")

## Related Articles

- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)
- [Transforming Software Development: The Role of AI and Developer Tools](https://www.wearedevelopers.com/magazine/527-transforming-software-development-the-role-of-ai-and-developer-tools)
- [Exploring AI: Opportunities and Risks for Developers](https://www.wearedevelopers.com/magazine/522-exploring-ai-opportunities-and-risks-for-developers)
- [What is Software Engineering in the Age of AI?](https://www.wearedevelopers.com/magazine/640-what-is-software-engineering-in-the-age-of-ai)

## Related Jobs

- [Staff Software Engineer, Copilot Experiences](https://www.wearedevelopers.com/jobs/ext/164361-staff-software-engineer-copilot-experiences) at **GitHub**
- [Head of AI Applications](https://www.wearedevelopers.com/jobs/ext/1456210-head-of-ai-applications) at **ZEISS Group**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Head of AI Applications](https://www.wearedevelopers.com/jobs/ext/1231536-head-of-ai-applications) at **ZEISS Group**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**