Skip to main content

Canonical Definitions

This page is the single source of truth for ADG terminology. Other pages should link here rather than restating definitions.

Note: These definitions explain WHAT terms mean in ADG. They are not requirements or controls (those belong in /docs/standards/).

Agentic Coding​

Agentic coding is a development workflow where a human directs an AI agent to make code changes (often across multiple files and layers) and the agent executes those changes autonomously.

Key characteristics:

  • the human provides goals/constraints, not line-by-line edits
  • the agent may modify code, tests, configs, docs, and build scripts
  • the output may be “usable” even if its internal structure is poor
  • review is frequently partial due to volume and speed

Agentic Programming​

Agentic programming is a broader term covering agent-driven implementation of program behavior, not limited to “coding”:

  • refactors
  • dependency changes
  • test generation
  • API design
  • performance improvements
  • security-related changes (sometimes unintentionally)

Agentic programming emphasizes the agent’s role in shaping program behavior and architecture, not just writing syntax.

Agentic Development​

Agentic development is end-to-end software development performed with AI agents as active participants across the SDLC, including:

  • feature development
  • schema and migration work
  • CI/CD pipeline changes
  • infrastructure provisioning
  • incident response automations
  • documentation updates
  • release management tasks

Agentic development often involves multiple agents operating in parallel and handing work off between each other.

Agentic DevOps​

Agentic DevOps is the application of agentic development to operational domains:

  • deployment automation
  • environment management
  • secrets and config handling
  • monitoring/alerting changes
  • runbook execution
  • rollback orchestration

Agentic DevOps increases speed but also increases risk because errors can be systemic and fast.

Drift​

Drift is the gradual (or sudden) divergence between:

  • architectural intent and implemented reality
  • documentation and behavior
  • schema and code assumptions
  • environments (dev/staging/prod)
  • security policy and actual access paths

In agentic systems, drift can be accelerated by high-velocity, high-volume change.

Agentic Development Governance (ADG)​

Agentic Development Governance (ADG) is the discipline and implementation layer that ensures software created or modified by agents remains:

  • correct
  • coherent
  • secure
  • compliant
  • documented
  • maintainable
  • auditable

ADG exists because traditional safeguards (manual review, ad hoc QA, “we’ll catch it later”) do not scale to agentic change velocity.