| name | ax2012-d365fo-migration |
| description | Use when analyzing or migrating AX 2009/2012 XPO exports, repositories, model metadata, SQL, integrations, data migration, or project plans to D365 F&O. Produces evidence-led inventory, Standard/Extend/Rebuild/Remove decisions, risks, target patterns, tests, and review artifacts. |
AX 2012 → D365 F&O Migration
Treat migration as evidence-led modernization, not code conversion. Preserve source evidence, distinguish facts from recommendations, and never claim a generated extension or test has been compiled in a D365 environment unless that environment was actually used.
Intake and safety
Accept a repository, XPO export, model-store extract, SQL scripts, or a combination. Before analysis, establish the AX version, legal entities/company scope, source snapshot date, available model/ISV metadata, and whether any extract contains credentials or production data. Do not copy credentials or personal data into generated reports.
If the input is a model store or SQL database that cannot be safely inspected directly, request an approved metadata export. Use read-only access for SQL discovery. State gaps explicitly: an XPO/source-only analysis cannot prove security assignments, SQL usage from external applications, or run-time execution frequency.
Inventory workflow
- Create an immutable input manifest: path, size, timestamp, SHA-256, and extraction method.
- Run the bundled decision compiler for the complete first-pass deliverable set:
python scripts/ax_migration_compiler.py <input-path> --out <output-folder> --generate-proposals --review-room --usp-suite --executive-suite --process-suite
- Inspect the generated
inventory.json and migration-backlog.md; validate suspicious classifications against the actual artifact and cite the source path/line where possible.
- Correlate with AX metadata or SQL extracts when available: model/layer, AOT object ID, configuration keys, security privileges/duties/roles, and integration endpoints.
- Reclassify each candidate using the decision rules below. Keep the original heuristic result and a reviewer override when they differ.
Classification rules
Use exactly one primary classification per object, with a confidence level and rationale:
| Classification | Use when | Typical D365 F&O target |
|---|
STANDARD | Microsoft functionality meets the business requirement after configuration/data migration. | Configuration, parameters, feature management. |
EXTEND | The requirement remains valid and can be implemented through supported extensibility. | Extension model, event handler, Chain of Command, delegate, data entity extension. |
REBUILD | The business capability remains but the implementation/architecture is incompatible or materially different. | New extension/service/workflow/report/integration, with a design decision. |
REMOVE | Obsolete, duplicate, unused, or unsupported without a retained business need. | Retirement plan, archive/data retention decision. |
Do not use object type alone as a classification. Assess business ownership, usage, standard D365 fit, extension points, integrations, data impact, and testability. REBUILD requires a named business owner and an explicit target architecture recommendation.
Required discovery coverage
Inventory Tables, Maps, Views, Queries, Base Enums, EDTs, Classes, Forms, Menu Items, Reports, Batch classes/jobs, AIF/services, SSRS/report artifacts, Security roles/duties/privileges, workflow, data entities, SQL scripts, external integrations, and ISV objects.
Flag these patterns at minimum:
- Overlayering/layer conflicts and modified Microsoft objects.
- Direct SQL:
Connection, Statement, createStatement, executeQuery, DDL/DML literals and external .sql files.
- Fragile X++ patterns:
runAs, WinApi, COM/CLR interop, file-system paths, client/server assumptions, ttsBegin around external calls, and reflection/dynamic calls.
- Integration migration candidates: AIF documents/services, Business Connector, .NET interop, file drops, SOAP, and custom endpoints.
- Reporting migration candidates: MorphX reports and direct report execution.
- Security gaps: code-only role checks, missing privilege/duty decomposition, and data access outside XDS/role design.
- Data risks: table extensions, surrogate key assumptions, cross-company usage, deleted actions, and direct database writes.
- ISV provenance and a vendor/support decision for every detected non-Microsoft model or namespace.
Risk and effort
Score risk from 1 (low) to 5 (critical) with evidence-backed factors: overlayering, data contract impact, integration coupling, security impact, unsupported APIs, code complexity, external dependency, and missing tests. Estimate effort as a range and state assumptions; do not present heuristic estimates as commitments.
Use this default effort rubric per object, then adjust for dependencies and validation scope: XS (≤1 day), S (1–3 days), M (3–8 days), L (8–20 days), XL (>20 days/design spike). A portfolio estimate must include contingency and non-build work: environment setup, data migration, regression testing, cutover, training, and ISV/vendor coordination.
Deliverables
Produce these under a dated output folder:
input-manifest.json — source provenance and known limitations.inventory.json — every discovered object, dependencies, signals, risk, classification, confidence, and evidence.migration-backlog.md — grouped portfolio counts plus one row per non-standard decision.target-architecture.md — supported D365 F&O pattern recommendation for each retained capability.migration-evidence.md — decision log mapping source evidence to target proposal, test strategy, owner, and open questions.test-strategy.md — unit, component/integration, security, data migration, and regression coverage.decision-compiler.json — a reviewable decision card per object: classification, target pattern, effort range, open questions, evidence, and confidence.d365-standard-fit.md — explicit standard-fit hypothesis and configuration-versus-code decision.extension-coach.md — supported extensibility recommendation with a Chain-of-Command/event-handler choice and rejection criteria.blast-radius.md — object/dependency graph plus data and integration cutover risks.proposed/ — review-ready X++ extension/data-entity/test skeletons and a generated proposal index when --generate-proposals is used.review-room/ — an evidence-led migration review queue, integration cutover scenarios, standard-fit challenges, negotiated test acceptance, ISV exit plan, and an auditable decision-update template when --review-room is used.usp-suite/ — ten AI-native capability views: evidence-to-decision, standard-fit, overlayering, X++ intent, integration contract, data semantic mapping, security delta, report decision, cutover failure, and ISV exit intelligence when --usp-suite is used.executive-suite/ — ten CEO/CIO/CFO/programme-lead decision views, an evidence-labelled metric snapshot, and a board-ready brief when --executive-suite is used.process-suite/ — ten Process Owner/Manager Standard-First capability views, a design-authority queue, UAT contracts, and a process-owner decision brief when --process-suite is used.
When the user asks to generate code, generate only review-ready D365 F&O artifacts in a separate proposed/ tree: extension, Chain-of-Command/event handler choice and justification, data entity extension when warranted, and SysTest skeletons. Do not overwrite an existing D365 package or claim deployment readiness without compilation, best-practice checks, and execution evidence.
Ten product capabilities
Run and report all five capabilities as one connected workflow:
- Migration Decision Compiler — produce a decision card from each source object and its evidence; decisions are reviewable, not autonomous approvals.
- D365 Standard-Fit Navigator — identify where a configuration/standard-fit assessment must happen before custom code. A standard-fit hypothesis is not proof of product capability; validate it against the target version and licensed modules.
- Overlayering-to-Extension Refactoring Coach — recommend supported extension patterns. Prefer configuration, then extension metadata, event/delegate handlers, and Chain of Command only where an extensibility seam exists; otherwise recommend redesign.
- Migration Test Evidence Factory — create traceability from decision to unit/component/security/data/regression tests. Generated tests are skeletons until assertions and fixtures are implemented and run.
- Data & Integration Blast-Radius Graph — show dependencies, direct SQL, AIF/services, reports, and security signals as cutover risks. It is a static analysis graph, not a run-time dependency proof.
- Migration Review Room — arrange all evidence into a priority queue that requires a human action (
ACCEPT, REJECT, REQUEST_EVIDENCE, DEFER, or REMOVE_CANDIDATE). Include a counter-hypothesis, named owner requirement, cutover scenario, test acceptance contract, and ISV verification path per item.
- X++ Intent Reconstruction — express static code evidence as a business-behavior hypothesis, explicitly labelled as inferred and awaiting owner confirmation.
- Data Semantic Successor — provide reviewable source-to-target data-semantic hypotheses for tables, EDTs, enums, and data paths; do not invent D365 metadata.
- Security Delta Navigator — turn AX security artifacts and access signals into persona, privilege, duty, SoD, and access-test questions.
- Report-to-Decision Transformer — assess reports by the business decision they support, then test standard inquiry/workspace/SSRS alternatives before rebuilding layout.
Output language and final summary
Use the stakeholder's language. Lead with the portfolio totals in this form:
<total> AX objects analysed
<standard> Microsoft standard
<remove> obsolete/removal candidates
<extend> extension candidates
<rebuild> manual redesign required
Then state source limitations, top five migration risks, the proposed target patterns, and the next validation gate. Separate observed, inferred, and decision required items.
Innovation evidence adapters
Run the ten next-wave capabilities with:
python scripts/ax_migration_innovation.py <input-path> --out <output-folder>
Optional sanitized evidence inputs are --d365-metadata, --runtime-events, and --learning-log JSON. Never include credentials, production personal data, or unrestricted database extracts. The generated innovation-suite/ provides import contracts and evidence gates; it does not connect to or modify D365 F&O.
Project Lead OS
Run the evidence-led project operating pack with:
python scripts/ax_migration_project_lead.py <input-path> --out <output-folder>
It produces a charter, WBS, RAID, RACI, wave plan, change-control template, governance cadence, status report, cutover command center, and autonomy boundary. Optionally provide a sanitized --project-context JSON for sponsor, owners, capacity, calendar and approved planning assumptions. The agent prepares and maintains evidence-led artefacts; humans retain people, budget, contract, risk-acceptance and go-live authority.
Azure DevOps adapter
Create review-gated Boards, PR and build-evidence payloads with:
python scripts/ax_migration_azure_devops.py <input-path> --out <output-folder> --organization <org> --project <project>
The adapter never stores credentials or calls Azure DevOps. In an authenticated Codex session, use the available Azure DevOps connector tools to apply only reviewed payloads. Never auto-merge, auto-deploy, or create work items without a named owner and human approval.
Data Migration Factory
Generate the full review-gated data migration pack with:
python scripts/ax_data_migration_factory.py <input-path> --out <output-folder>
Optional sanitized --target-entities and --profile JSON improve mapping and profiling. The factory produces mapping, profiling, transformation, sequencing, staging/retry, reconciliation, privacy/retention, UAT, D365 import payload and cutover-gate artifacts. It never reads or writes production data, executes a D365 import, or approves data migration.
Commerce & Integration Factory
Generate all 20 shop, marketplace, dropship, CRM, order-management and fulfilment capabilities with:
python scripts/ax_commerce_integration_factory.py <input-path> --out <output-folder>
The suite creates 20 review-gated integration artifacts plus a canonical commerce contract. It performs no external connection or write; channel credentials, orders, prices, stock, refunds, replays and customer merges remain explicitly owner-approved connector actions.
Commerce Scale Unit
Generate the ten channel-scale operating capabilities with:
python scripts/ax_commerce_scale_unit.py <input-path> --out <output-folder>
It creates launch, country, economics, assortment, inventory, seller health, incident, partner, shock-simulation and operating-review artifacts. Human owners retain launch, commercial-policy, contract and external-write authority.
Enterprise Delivery Factory
Generate the 16 final-mile enterprise capabilities as one review-gated delivery pack:
python scripts/ax_enterprise_delivery_factory.py <input-path> --out <output-folder> --d365-metadata <sanitized-metadata.json> --runtime-events <sanitized-events.json> --build-evidence <sanitized-builds.json> --isv-register <sanitized-isv.json> --data-profile <sanitized-profile.json>
All optional evidence files are sanitized JSON and may be omitted. The pack covers target metadata, fit-to-standard, build/BP gate, automated testing, transformation workbench, process usage, process twin, security/SoD, cutover, hypercare, ISV, integration contracts, data quality, value, operating model and evidence ledger. It creates enterprise-delivery-suite/ in REVIEW_ONLY mode. It never connects to D365, stores credentials, reads/writes production data, compiles/deploys X++, approves a decision, or performs cutover/go-live actions.
For an approved official demo environment, use the guidance in DEMO_ENVIRONMENTS.md: F&O trials are suitable for fit-to-standard/UAT demonstrations; developer-focused sandboxes are required for metadata, Visual Studio build and test evidence.
Quality Control, Evidence Ledger and Connector Requests
After the decision compiler runs, validate all optional evidence inputs and create reproducible management/control artifacts:
python scripts/ax_migration_quality.py --decisions <output-folder>\decision-compiler.json --out <output-folder> --language de
The quality pack includes secret scanning, input-schema gates, versioned rule audit, input/run hashes, a full evidence ledger, benchmark results, an evidence-labelled executive dashboard and review-only D365/Azure DevOps/build/monitoring connector requests. See QUALITY_AND_CONNECTORS.md. Do not supply credentials, connection strings, personal data or unrestricted production extracts.
