Agentic SDLCs Challenge Agile Manifesto Foundations

Steve Jones, Executive VP at Capgemini, ignited industry debate by asserting that AI-powered agentic SDLCs fundamentally conflict with the Agile Manifesto's core tenets. Unlike traditional Agile workflows designed for human collaboration cycles, agentic systems leverage AI agents to execute development tasks with unprecedented speed and scale, creating three critical incompatibilities:

Tool Dependency Conflicts

Agentic SDLCs inherently prioritize toolchains over human interaction—a direct contradiction to the Manifesto's "individuals and interactions over processes and tools" principle. Implementation specifics drastically impact outcomes: systems built on Replit's collaborative IDE exhibit different behavioral characteristics than those using Anthropic's Claude Code agents. When integrating multiple agentic systems, toolchain interoperability becomes critical infrastructure requiring explicit design decisions.

Temporal Incompatibility

Agentic SDLCs collapse development timelines from weeks to hours. Jones documents cases where functional applications are built during transcontinental flights—rendering two-week sprints architecturally irrelevant. This velocity necessitates real-time monitoring systems capable of tracking agent activities at sub-minute resolution. Performance benchmarks show Claude Code completing standard CRUD implementations in 3.7±0.8 minutes versus human teams averaging 14.2±3.1 hours.

Technical Debt Amplification

The "working software over documentation" principle becomes hazardous when AI generates superficially functional code. Without comprehensive documentation and architectural guardrails, agent-produced solutions accumulate technical debt at exponential rates. AWS's 2026 architectural guidance prescribes "intent design" patterns where humans define:

• Role-based agent permissions
• Fallback protocols for non-deterministic outputs
• Validation checkpoints at 15-minute intervals

Industry Responses

Kent Beck advocates "augmented coding" maintaining TDD rigor while AI handles implementation. His B+ Tree library experiment achieved 98.6% test coverage with AI-generated Rust/Python code under human supervision. Casey West's draft Agentic Manifesto shifts focus from verification (instruction compliance) to validation (outcome correctness).

Contrasting data emerges from Forrester's 2025 State of Agile report: 95% of organizations maintain Agile relevance, with 49% integrating generative AI into existing workflows. Hybrid approaches show promise—Sandvik uses AI for boilerplate generation while retaining Agile ceremonies for requirement refinement.

Deployment Considerations

Organizations implementing agentic SDLCs report these operational requirements:

Component	Specification	Implementation Risk
Agent Orchestration	Minimum 10Gbps inter-agent mesh	Latency-induced race conditions
Validation Layer	Semantic diff tools + anomaly detection	False positives in complex systems
Knowledge Graph	Real-time updated dependency map	Schema drift inhibition

While the Manifesto's adaptability principles endure, its human-centric workflows require reengineering for AI collaboration. The emerging consensus suggests not obsolescence but evolution—retaining Agile's philosophical core while rebuilding its operational scaffolding for agentic velocity.

About the Author

Steef-Jan Wiggers is a Domain Architect at VGZ and InfoQ's Cloud Queue Lead Editor. His work focuses on Azure DevOps implementations and AI-integrated cloud architectures. A Microsoft Azure MVP for 16 years, he regularly speaks at international conferences on infrastructure modernization.