S

ADLC — how the system articulates phase by phase

Agentic Development Lifecycle · for each step: what triggers it · who runs it · what it produces · what feeds next

Mantra: every phase has a trigger, an agent fleet, humans in the loop, artifacts produced, and an explicit handoff. No bookshelf outputs. The loop closes from Operate back to Discover — production reality is the strongest signal for the next round of work.

Tier: Built & live MUST — to engineer Nice to have
Show:
Discover

Sense what to do. Surface signals worth acting on.

Built 7MUST 8Nice 0
Briefing surfaces an anomaly
LOB lead raises a need
AOP intake question opens
PM UX LOB lead CS
Chorus Pendo Salesforce Atlassian (CNQ) Memory federation
Project Brief Research synthesis Daily briefing
evidence conversation
→ Feeds Plan · brief + signals + open questions drive PRD authoring
Plan

Decide what to build and why. PRD · architecture · epics.

Built 14MUST 5Nice 0
Discover artifact lands
Capacity opens in a LOB
Strategic objective set by VP
VP Product VP Tech PM Architect LOB lead
Atlassian (JPD) Notion Bedrock / Claude Memory federation
BRD PRD Architecture doc Initiative plan Workflow template
decision evidence
→ Feeds Build · PRD + architecture + epic decomposition land in build queue
Build

Implement the spec. Agents draft code → open a PR on Azure DevOps.

Built 10MUST 8Nice 0
PRD approved
Epics decomposed
Sprint kicks off
Engineer Architect UX Designer Tech Writer
Authoring Canvas Create-Agent wizard Launcher Build Board
Azure DevOps (PR target) GitHub Bedrock / Claude Devin (autonomous · opens PR) azdo-pr-deploy
Story spec AzDO Pull Request Unit tests Docs
decision code-review note pr-link
→ Feeds Test · the PR opens in AzDO with the agent as author. Pipelines run automatically; human reviewer (Architect / Engineer) gates the merge.
Test

PR review · AzDO Pipelines · eval gates. Nothing merges without sign-off.

Built 10MUST 5Nice 0
PR opened on AzDO
Pipelines complete
Pre-publish eval requested
QA Architect Engineer
Azure DevOps (PR + pipelines) Eval framework Datadog Browser / Computer Use Bedrock / Claude
PR review + approval Eval report QA checklist Epic-IQ score
eval-run lesson pr-review-notes
→ Feeds Ship · PR merges (or back to Build for rework). Merge triggers AzDO release pipeline.
Ship

PR merges → AzDO pipeline deploys. Launch readiness in parallel.

Built 11MUST 9Nice 1
PR merged on AzDO
Release pipeline succeeds
Launch window opens
VP Product VP Tech Exec Sales Marketing
Deploy Hub Org Rollup Connections SPA
Azure DevOps (release pipelines) AWS CloudFront AWS Lambda AWS ECS AWS S3 Slack (announce) DocuSign
Release notes Launch narrative Sales battle cards Live deployment
decision (launch) status
→ Feeds Operate · running production system that's actually used
Operate

Sense · learn · iterate. The loop closes here.

Built 13MUST 7Nice 0
Production incident
Usage drop or anomaly
FinOps cap warning
Customer feedback arrives
Memory Curator VP Tech CS LOB lead
Datadog PagerDuty Kubernetes AWS S3 (logs) FinOps caps
Incident report Lesson entry Metric digest Prompt-edit draft
lesson status decision
↻ Feeds Discover · production reality is the strongest signal for the next round

The loop closes — every cycle

Discover
Plan
Build
Test
Ship
Operate
back to Discover

Every phase outputs something a downstream phase consumes — that's the no-bookshelf rule, enforced by the Workflow Output Validator. In most cases the Build phase terminates in an Azure DevOps Pull Request, which is where a human reviewer (Architect, Engineer, or QA) gates the merge. The PR is the explicit handshake between AI agents and the human review process — no agent merges without sign-off. Once merged, AzDO release pipelines take over to actually deploy. The Operate signals then close the loop: incidents and usage data flow back through the Briefing into the next round of Discover. Memory accumulates across cycles — decisions, evidence, lessons — so each pass is informed by the last, and the Conductor agent helps anyone navigate where they are in the cycle.

Home