AI-DLC Dictionary of Terms
A Practitioner's Reference to AI-Driven and Agentic Software Development Vocabulary
AI-DLC (AI-Driven Development Life Cycle) is AWS's specific, named methodology for AI-driven software delivery, comprising three phases — Inception, Construction, and Operations — and is distinct from related but different industry terms including Agentic Software Development (ASD), Agentic Development Security (ADS), and the general Agentic Development Lifecycle (ADLC).
This reference is independently published by Vīkṣya. Vīkṣya is not affiliated with, endorsed by, sponsored by, or otherwise associated with AWS, Forrester, Cycode, or any other third-party organisation mentioned in this document. All company names, product names, methodologies, analyst terms, trademarks, and copyrighted terms referenced belong to their respective owners and are used for descriptive and educational purposes only.
How to use this document: Section 2 resolves the single biggest point of confusion — the competing umbrella terms — first. Section 3 defines AWS's AI-DLC vocabulary in full. Section 4 maps AI-DLC concepts to their nearest Agile/SAFe/Waterfall equivalents. Section 5 is an A–Z master glossary for quick lookup.
Why This Dictionary Exists
Software development terminology is being rewritten faster than it can be standardised. In the space of roughly a year, four different organisations have introduced four overlapping — and easily confused — names for the same underlying shift: AI systems doing real development work, not just assisting with it. AWS calls its methodology AI-DLC. Forrester calls the emerging category of tools Agentic Software Development (ASD), and separately named a security category Agentic Development Security (ADS). Security vendors including Cycode use the term Agentic Development Lifecycle — also abbreviated ADLC — as a general description of the lifecycle these tools secure.
None of this is standardised yet. No analyst firm has published a settled maturity model. Vendors are actively shaping the vocabulary in their own favour. This dictionary exists to give practitioners — CTOs, VPs of Engineering, transformation leads, and the consultancies advising them — a neutral, accurate reference point before picking a side.
The Competing Umbrella Terms — Resolved
These five terms are the ones most often confused with each other. They are not interchangeable. Each comes from a different organisation, describes a different scope, and implies a different vendor relationship.
AI-DLC (AWS's methodology) and ADLC (the general "agentic development lifecycle" descriptor used by security vendors) are near-homophones describing different things — one is a specific, branded way of working; the other is a generic noun phrase. Confirm which one is meant before using either term in a client-facing document.
| Term | Origin | What It Actually Means |
|---|---|---|
| AI-DLC (AI-Driven Development Life Cycle) | AWS | A specific, named methodology with three phases (Inception, Construction, Operations), defined rituals (Mob Elaboration, Mob Construction), and its own vocabulary (Bolts, Units of Work). Openly documented and partly open-sourced by AWS. |
| ASD (Agentic Software Development) | Forrester (analyst category) | An analyst-defined market category, not a methodology. Describes any tool where AI agents plan, generate, modify, test, and explain software across multiple SDLC stages with a degree of autonomy. Used to evaluate and compare vendors, not to prescribe a way of working. |
| ADS (Agentic Development Security) | Forrester (analyst category) | A distinct, newer analyst category (2026) for tools that secure agentic development itself — governing AI agents, coding tools, MCP servers, and AI-generated code. Adjacent to ASD, not a synonym for it. |
| ADLC (Agentic Development Lifecycle) | Security vendors incl. Cycode | A general descriptive term for the end-to-end lifecycle in which AI agents participate in development — used mainly by security vendors to describe what their tools protect. Not a formal methodology in its own right, and easily confused with AWS's similarly-abbreviated AI-DLC despite describing a different thing. |
| AI-Native SDLC | Vendor-neutral / industry usage | An umbrella descriptive phrase — not owned by any single vendor or analyst firm — for any software development lifecycle redesigned around AI agent participation rather than AI as an add-on tool. Useful as a category label when you deliberately want to avoid endorsing one vendor's proprietary terminology. |
AI-DLC Core Vocabulary (AWS)
AWS's AI-DLC — created by Raja SP, AWS Principal Solutions Architect — is the most fully documented methodology in this space, with a published Method Definition Paper and an open-sourced workflow repository (awslabs/aidlc-workflows). The terms below are specific to this methodology and are listed in the order a project encounters them.
- Intent AI-DLC · AWS
- The starting point of an AI-DLC engagement — a plain-language statement of what the team wants to build, replacing a pre-written backlog item as the trigger for work.
- Inception (Phase 1) AI-DLC · AWS
- The phase in which AI converts business intent into detailed requirements, user stories, and Units of Work, validated by the team through Mob Elaboration.
- Mob Elaboration AI-DLC · AWS
- The defining team ritual of Inception: the entire cross-functional team — product, engineering, QA, operations, security — assembles in real time to validate AI-generated questions, requirements, and proposals together, rather than routing them through sequential reviews.
- Unit of Work AI-DLC · AWS
- AI-DLC's replacement for the Agile epic — a decomposed, independently deliverable slice of the overall intent, with its own dependencies, acceptance criteria, and parallelisation boundaries.
- Construction (Phase 2) AI-DLC · AWS
- The phase in which AI proposes logical architecture, domain models, code, and tests, using the context validated during Inception, and the team clarifies technical decisions in real time through Mob Construction.
- Mob Construction AI-DLC · AWS
- The defining team ritual of Construction: the same cross-functional group works together in real time — often co-located — to clarify architectural and technical decisions as AI generates implementation, rather than reviewing finished code after the fact.
- Bolt AI-DLC · AWS
- AI-DLC's replacement for the Agile sprint — a short, intense execution cycle measured in hours or days rather than weeks, scoped to a single Unit of Work or a slice of one.
- Domain Design AI-DLC · AWS
- The step within Construction where AI models core business logic using domain-driven design principles, independent of infrastructure or deployment concerns.
- Logical Design AI-DLC · AWS
- The step that extends Domain Design into architectural pattern recommendations — for example, event-driven processing or CQRS — documented for human review.
- Architecture Decision Record (ADR) AI-DLC · AWS
- A structured, persisted record of an architectural decision and its rationale, generated during Construction for human review and approval, and retained as part of the project's audit trail.
- Deployment Unit AI-DLC · AWS
- The packaged output of Construction — secured, instrumented, de-risked, tested, and ready for production — that moves into the Operations phase.
- Operations (Phase 3) AI-DLC · AWS
- The phase in which AI applies accumulated context from Inception and Construction to manage infrastructure-as-code and deployment, with the team maintaining oversight. The least mature and least documented of the three AI-DLC phases as of mid-2026.
- Verification Gate AI-DLC · AWS
- A mandatory human checkpoint within a Bolt at which the AI cannot proceed without explicit approval. AWS's own documentation describes 10–26 such gates per Bolt, depending on complexity.
- Persistent Context AI-DLC · AWS
- The accumulated record of plans, requirements, decisions, and design artefacts that AI-DLC stores back to the project repository at each phase, so that context carries forward across sessions and phases instead of being re-established each time.
- Adaptive Workflow (Principle 10) AI-DLC · AWS
- AWS's design principle that AI-DLC should not force every project through the same fixed sequence of stages — the depth and steps applied should scale to the complexity and nature of the specific piece of work, skipping ceremonies that add no value.
- Process Atrophy AI-DLC · AWS
- The risk condition — not an AWS term, but a necessary one for any adopting organisation — in which governance gates and verification checkpoints exist on paper but are silently bypassed as delivery velocity pressure increases.
AI-DLC to Agile / SAFe / Waterfall — Cross-Mapping
This table maps AI-DLC concepts to their nearest equivalents in established methodologies. Treat these as orientation points, not exact substitutions — in most cases the AI-DLC concept changes the underlying mechanics, not just the label.
| AI-DLC Concept | Nearest Equivalent | Key Difference |
|---|---|---|
| Bolt | Sprint | Fixed two-week cycle vs. an AI-recommended cycle of hours to days, scoped dynamically to the work rather than a calendar. |
| Unit of Work | Epic | Similar in scope but explicitly designed with parallelisation and AI-decomposition in mind from the outset, not broken down manually over time. |
| Mob Elaboration | Backlog Grooming / Refinement | Real-time, whole-team, synchronous validation session (hours) replacing asynchronous, iterative grooming (days to weeks). |
| Mob Construction | Sprint Execution + Code Review | Collapses build and review into a single real-time collaborative session rather than a write-then-review sequence. |
| Verification Gate | Definition of Done / Acceptance Criteria | An explicit, per-checkpoint human approval embedded in the AI workflow itself, rather than a criteria checklist assessed at the end of a sprint. |
| Persistent Context | Documentation / Knowledge Base | Actively maintained and consumed by the AI system itself at every phase, not a passive reference artefact maintained separately from delivery. |
| Velocity (AI-DLC sense) | Velocity (Scrum sense) | Measured in throughput of validated human decisions and completed Units of Work, not story points burned per sprint. |
| Adaptive Workflow | Tailoring the Framework (SAFe) | AI dynamically selects which stages a specific piece of work needs, versus a human team agreeing on a tailored process in advance. |
| Program Increment (PI) Planning | No direct AI-DLC equivalent | AI-DLC has no prescribed multi-team, calendar-fixed planning event; cross-team dependency coordination in AI-DLC is comparatively immature and unaddressed as of mid-2026. |
A–Z Master Glossary
A single alphabetised reference spanning all terms in this dictionary, for quick lookup. Each entry is a standalone, single-sentence definition.
- Adaptive Workflow
- AI-DLC design principle: process depth scales to the complexity of the specific task rather than following a fixed sequence for every project.
- Agentic Development Lifecycle (ADLC)
- General descriptive term, used mainly by security vendors, for the end-to-end lifecycle in which AI agents participate in development.
- Agentic Development Security (ADS)
- Forrester's analyst category for tools that secure agentic software development — the AI agents, tooling, and AI-generated code itself.
- Agentic Software Development (ASD)
- Forrester's analyst category for AI agents that plan, generate, modify, test, and explain software across multiple SDLC stages with a degree of autonomy.
- AI-DLC (AI-Driven Development Life Cycle)
- AWS's specific, named methodology comprising Inception, Construction, and Operations phases.
- AI-Native SDLC
- Vendor-neutral umbrella phrase for a development lifecycle redesigned around AI agent participation.
- Architecture Decision Record (ADR)
- A persisted record of an architectural decision and its rationale, generated for human review during Construction.
- Bolt
- AI-DLC's short execution cycle (hours to days), replacing the fixed-length sprint.
- Construction
- AI-DLC Phase 2 — architecture, domain modelling, code, and test generation, validated through Mob Construction.
- Deployment Unit
- The packaged, tested, production-ready output of Construction.
- Domain Design
- Business-logic modelling step within Construction, using domain-driven design principles.
- Inception
- AI-DLC Phase 1 — converting business intent into requirements and Units of Work, validated through Mob Elaboration.
- Intent
- The plain-language statement of what to build that triggers an AI-DLC engagement.
- Logical Design
- Architecture pattern recommendations extending Domain Design, documented via ADRs.
- Mob Construction
- Real-time, whole-team ritual for validating technical and architectural decisions during Construction.
- Mob Elaboration
- Real-time, whole-team ritual for validating requirements and AI proposals during Inception.
- Operations
- AI-DLC Phase 3 — AI-managed infrastructure and deployment under team oversight.
- Persistent Context
- The accumulated record of decisions and artefacts carried forward across AI-DLC phases and sessions.
- Process Atrophy
- The risk that governance gates exist on paper but are bypassed under delivery velocity pressure.
- TuringBot
- Forrester's term for an AI agent performing software development tasks.
- Unit of Work
- AI-DLC's decomposed, independently deliverable slice of intent, replacing the epic.
- Verification Gate
- A mandatory human approval checkpoint within a Bolt; AWS documents 10–26 per Bolt depending on complexity.
Citation
This dictionary paraphrases and synthesises publicly available material. No text is reproduced verbatim from any source. Primary sources are listed first by publisher, in the order referenced through this document.
AWS — AI-DLC (AI-Driven Development Life Cycle)
- Raja, S.P. (2025). "AI-Driven Development Life Cycle: Reimagining Software Engineering." AWS DevOps & Developer Productivity Blog, July 31, 2025. aws.amazon.com/blogs/devops/ai-driven-development-life-cycle
- AWS (2025). "Open-Sourcing Adaptive Workflows for AI-Driven Development Life Cycle (AI-DLC)." AWS DevOps & Developer Productivity Blog, November 30, 2025. aws.amazon.com/blogs/devops/open-sourcing-adaptive-workflows-for-ai-driven-development-life-cycle-ai-dlc
- AWS (2025). "Building with AI-DLC using Amazon Q Developer." AWS DevOps & Developer Productivity Blog, November 30, 2025. aws.amazon.com/blogs/devops/building-with-ai-dlc-using-amazon-q-developer
- AWS Labs (2025–2026). "aidlc-workflows: AI-Driven Life Cycle (AI-DLC) Adaptive Workflow Steering Rules for AI Coding Agents." GitHub repository, MIT-0 licence. github.com/awslabs/aidlc-workflows
Forrester — Agentic Software Development (ASD) and Agentic Development Security (ADS)
- Lo Giudice, D. (2026). "Agentic Software Development: Defining The Next Phase Of AI-Driven Engineering Tools." Forrester Blogs, March 2, 2026. forrester.com/blogs/agentic-software-development-defining-the-next-phase-of-ai-driven-engineering-tools
- Lo Giudice, D. (2026). "Agentic Software Development Takes The Lead: From Code Assistants To Orchestrated SDLC Agents." Forrester Blogs, June 8, 2026. forrester.com/blogs/agentic-software-development-takes-the-lead-from-code-assistants-to-orchestrated-sdlc-agents
- Worthington, J. (2026). "Agentic Development Security: Why AppSec Needs A New Operating Model." Forrester Blogs, April 3, 2026. forrester.com/blogs/agentic-development-security-why-appsec-needs-a-new-operating-model
- Forrester Research (2026). "The Top 10 Emerging Technologies In 2026." Press release, April 15, 2026. forrester.com/press-newsroom/forresters-top-10-emerging-technologies-for-2026
Security Vendors — Agentic Development Lifecycle (ADLC) Usage
- Cycode (2026). "Forrester Names Cycode in The Agentic Development Security Tools Landscape, Q2 2026." Cycode Blog, May 27, 2026. cycode.com/blog/forrester-names-cycode-in-the-agentic-development-security-tools-landscape-2026
Note: as of this edition, Forrester's Agentic Development Security Tools Landscape, Q2 2026 names 35 vendors using ADS and ADLC terminology, including Cycode, ReversingLabs, Apiiro, and Arnica. Cycode is cited as the representative primary source for the ADLC term specifically.
Vendor-Neutral / Industry Usage — AI-Native SDLC
- Augment Code (2026). "How AI Changes the SDLC: A Six-Stage Guide." Augment Code Guides. augmentcode.com/guides/how-ai-changes-the-sdlc
A note on currency: every source above was published or updated in the twelve months preceding this edition. Given the pace of change in this space, Viksya reviews and re-verifies these citations at each dictionary revision rather than carrying them forward unchecked.
Published free, reviewed as the vocabulary settles.
This dictionary is published free by Vīkṣya from information available in the public domain, without affiliation to any brand or entity named within it, and will be reviewed as the terminology in this space consolidates — expect meaningful revision within 12 months as analyst firms formalise maturity models and vendor naming settles.
Vīkṣya provides practitioner-grade tools for enterprise AI and digital transformation programs, including a growing set of tools for organisations adopting AI-driven development methodologies: an AI adoption readiness assessment, an ROI calculator, use case prioritization, AI project estimation, and governance and SAFe/Agile-to-AI-DLC transition planning. Vīkṣya also builds enterprise-grade professional systems for digital creators. Review all Viksya offerings at www.viksya.com.
Vīkṣya.com — Enterprise Tools for AI, Digital Transformation Programs and Creator Economy