ADLC · Slide 2 · Co-development

Agent and Human
Co-Development

Agent Execute, Human Steer. Agents own reversible execution work. Humans own judgement, architecture, risk, and approval.

Explore Ownership Model Back to ADLC Agenda
Agent and Human Co-Development ADLC slide
Slide 2 · Agent and Human Co-Development
The core split

Who owns what and why it matters

The division is not random. ADLC delegates work based on reversibility: agents handle tasks that can be verified and corrected, while humans own decisions where errors compound.

💡

Agent Owns

Reversible · Verifiable · Executable
Code GenerationWriting implementation from spec; output is inspectable, testable, and replaceable.
Test AuthoringCreating unit, integration, and regression tests against defined acceptance criteria.
RefactoringRestructuring code for maintainability without changing observable behavior.
Code DocumentationGenerating inline docs, READMEs, and API references from existing code.
Simple DebuggingDiagnosing and fixing well-scoped, reproducible failures with clear signals.
🔁 Reversible mistakes can be caught and corrected
VS
⚙️

Human Owns

Irreversible · Judgement-driven · Strategic
Problem DefinitionDefining scope, users, constraints, and the real business problem.
ArchitectureChoosing system structure, boundaries, data flow, and tradeoffs that become costly to reverse.
Risk IdentificationIdentifying technical, operational, compliance, and business risks.
Risk AcceptanceDetermining which risks can be accepted, mitigated, or should block progress.
Stage ApprovalSigning off at checkpoints before the next stage begins.
🔒 Irreversible judgement cannot be outsourced
Guiding principle

Reversibility is the axis of delegation

The more reversible and verifiable a task is, the more autonomy an agent can safely receive. The more irreversible and judgement-heavy a decision is, the more human ownership it requires.

Agent territory �Shared zone �Human territory

ADLC does not ask “can the agent do it?â€?It asks “what happens if the agent is wrong?â€?/p>

�Reversible executionReview + gatesIrreversible judgement �/span>
AgentWrite a sorting function
HumanChoose a database engine
AgentGenerate API documentation
HumanDefine service boundaries
AgentRefactor duplicate logic
HumanAccept a security risk tradeoff
AgentWrite regression tests
HumanApprove production release

The agent is the most efficient and hardworking doer in the room, while the human is responsible for the final outcome.

Prashant Dhingra, Agentic Development Lifecycle Framework
What this means in practice

Implications for how teams build

Co-development changes planning, review, accountability, and speed. It is not simply “use AI moreâ€? it is a new operating model for safe delegation.

🗂ï¸?/div>

Task decomposition changes

Separate agent-suitable tasks from human-owned decisions before sprint execution begins.

�/div>

Review is lighter, not gone

Human review shifts from writing every line to validating that outputs satisfy criteria.

🚦

Gate approval stays human

Every stage gate needs explicit human approval, even when execution is highly autonomous.

📋

Accountability is traceable

The person who authorizes a gate owns the result. The model does not absorb accountability.

📐

Architecture debt is human debt

Agents amplify the structure humans define, so architecture quality matters even more.

📈

Speed increases where safe

Teams move faster by assigning reversible work to agents and reserving judgement for humans.