Automating Agentic AI Success Using This SECRET Workflow

2026-06-10 · Source: Modern Software Engineering · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning · Depth: Intermediate, long

Summary

Dave Farley's "Modern Software Engineering Channel" explores how Acceptance Test-Driven Development (ATDD) and Behavior-Driven Development (BDD) can automate agentic AI programming success. Guests Stefan Ellerstorfer from smarter software and Christian Gessel from Rohde & Schwarz discuss their experiences. Ellerstorfer highlights ATDD's role in mitigating AI's non-determinism by establishing precise, verifiable outcomes via executable specifications, allowing humans to focus on problem exploration. Gessel reinforces that ATDD fosters the specificity required to effectively instruct AI agents, viewing good prompts as akin to good specifications. The approach treats high-level behavioral specifications as a "fitness function" for AI, enabling agents to iterate towards solutions and potentially generate better tests and code, thereby raising the programming abstraction level.

Key takeaway

For AI Engineers developing with agentic systems, adopting Acceptance Test-Driven Development (ATDD) is crucial for managing AI's non-deterministic nature. You should define clear, executable behavioral specifications as a "fitness function" for your AI agents. This approach ensures verifiable outcomes, allowing your team to focus on higher-level problem definition while the AI iterates on implementation details, ultimately improving development speed and quality.

Key insights

ATDD provides a structured, deterministic framework for guiding non-deterministic agentic AI development through executable specifications.

Principles

• AI development benefits from precise, verifiable outcomes.
• Human natural language can be a structured programming language.
• Specificity in requirements is crucial for effective AI instruction.

Method

Use ATDD to define executable behavioral acceptance criteria. These specifications act as a "fitness function" for AI agents, allowing them to iterate and self-verify against desired outcomes, raising the programming abstraction level.

In practice

• Define acceptance criteria as executable specifications.
• Use BDD-style tests for security and performance.
• Provide agents with self-verification capabilities.

Topics

• Agentic AI
• Acceptance Test-Driven Development
• Behavior-Driven Development
• Software Engineering
• Continuous Delivery
• AI-Assisted Development

Best for: AI Architect, Software Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• Article: From Prompts to Production: A Playbook for Agentic Development — Agentic AI development requires a specialized lifecycle and testing to manage nondeterministic behavior and ensure production readiness.
• Agents should interview you — AI agents can significantly accelerate development by interviewing users for requirements and automating code generation and testing.
• New course! Spec-Driven Development — Spec-driven development uses detailed specifications to guide AI agents, improving code quality and managing developer cognitive load.
• I Applied Andrej Karpathy’s Auto-research to Software Development — Integrating LLM agents with code-defined invariants enhances autonomous software development iteration.
• SE Radio 711: Scott Hanselman on AI-Assisted Development Tools — Specificity and foundational knowledge are crucial for effectively steering AI-assisted development tools and managing inherent ambiguity.
• Why (Senior) Engineers Struggle to Build AI Agents — Philipp Schmid, Google DeepMind — Building AI agents demands a paradigm shift from deterministic software, embracing non-determinism, semantic understanding, and continuous evaluation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Modern Software Engineering.