Skip to content

Modernization Operating Model

The continuous SDLC operating model is covered in depth across Spec Sprint, Implementation Sprint, Team, and Enablement Partnership. It is sized for live applications with evolving scope: two staggered sprint cadences, four ontology custodians, a layered team structure, an enablement partnership across years. That operating model does not fit a modernization project, where the work is bounded and the target is well-defined.

This page covers the operating model that does fit modernization. It is the human-side discipline that wraps the modernization agent fleet across the five-stage pipeline. Where the continuous SDLC operating model is sprint-cadenced and custodian-owned, the modernization operating model is stage-sequenced and custodian-governed.

How the Two Operating Models Differ

The two operating models share principles (agents constrained by structure, humans validating rather than implementing every step, named owners for each artifact) but differ on six dimensions that derive from the use case shape.

DimensionContinuous SDLC operating modelModernization operating model
CadenceTwo staggered sprint cycles (Spec Sprint ahead of Implementation Sprint) running continuouslyFive sequential stages running once across the modernization project
Knowledge anchoringContinuous custodianship: the four custodians articulate intent into their ontology layers on a sprint cadenceStage-sequenced custodianship: the same four custodians govern the modernization ontologies; the Product Owner and Architect lead the annotation discipline (Retain, Modify, Replace, Retire) at module scoping, adjust during functional gate reviews, and validate at UAT
IterationPer-change SDLC flow runs once per change requestPer-module migration-validation iteration loop runs until the four validation gates pass
Human roleDeveloper in the loop on per-change decisions; PR Validation Agent gates each mergeModernization Expert validates per-iteration target code output; Product Owner validates the parity contract at UAT
Engagement lengthYears of continuous operationQuarters per modernization estate
Team compositionThe four custodians plus the implementation team plus the enablement layerThe Modernization Expert plus the Product Owner plus the Architect plus the engagement’s Chief Architect, sized to the modernization scope

A client who runs both engagement types uses the modernization operating model for the bounded migration project and then transitions to the continuous SDLC operating model once the modernized system is in live operation. The transition point is the hand-over at the end of the maintenance stage.

The Five-Stage Delivery

Modernization unfolds across five sequential stages. The first two build the modernization ontologies; the next two execute the migration and validate it; the fifth maintains the modernized system through hand-over. This section covers each stage from an operating-model perspective: who participates, what they decide, what gets produced, and how the SME tuning loop runs across stages.

Stage 1: Discovery and Analysis

ElementDetail
ParticipantsModernization Expert, Product Owner, Architect, engagement’s Chief Architect, client’s CTO or VP Engineering as sponsor
DurationTwo to four weeks
Decisions madeThe modernization scope at module level. The target stack (typically chosen by the client before the engagement starts and confirmed here). The MVP module that will pilot the migration. The pace and shape of the SME engagement across the project.
OutputsComprehensive Analysis Report on the legacy estate. Migration recommendations and detailed migration plan with target architecture and designs. MVP module identification.
Where the agents runThe ingestion and enrichment agents run on the legacy estate to produce a preliminary Source-state ontology that supports the analysis. The target blueprint is selected from the engagement’s options or composed bespoke.

Stage 2: Ontology Generation and Enrichment

ElementDetail
ParticipantsModernization Expert, engagement’s Chief Architect, Product Owner for periodic review
DurationOne to three weeks
Decisions madeThe aperture of the Source-state ontology: which legacy modules are decomposed at full statement depth, which are captured at module-summary depth (typically the modules slated for Retire treatment).
OutputsThe full Source-state ontology in the graph database, with AI-driven enrichment for dependencies, descriptions, and metrics. The Target-state ontology from the blueprint ingestion agent.
Where the agents runThe ingestion and enrichment agents complete their full pass. The specification extraction agent produces the first specification document and the human-readable extractions (business requirements document, BDD function tests, end-to-end test scripts, code chatbot).

Stage 3: MVP Migration (One Identified Module)

ElementDetail
ParticipantsModernization Expert, Product Owner for specification annotation and functional gate review, Architect for standards and architecture gate review, engagement’s Chief Architect for validation oversight
DurationSix to twelve weeks
Decisions madeThe SME’s annotation across the MVP module (Retain, Modify, Replace, Retire per Source-state node). The customization and configuration of the migration agents (frontend, backend, services) per migration requirements. The autonomy level the migration and validation agents will operate at for this engagement.
OutputsThe MVP module migrated end-to-end. Engagement-specific tuning of the agents based on Architect and Product Owner feedback. A documented per-module workflow that the Scaled Migration stage will run against.
Where the agents runThe migration agents produce target code candidates per module sub-unit. The four validation gates check each candidate. The iteration loop runs until the gates pass. The SME tuning loop runs in parallel: feedback from the SME on the MVP output drives agent prompt and configuration tuning for the subsequent modules.

Stage 4: Scaled Migration

ElementDetail
ParticipantsModernization Expert, Product Owner for annotation on remaining modules and functional gate review on each, Architect and Chief Architect for ongoing validation oversight
DurationQuarters per module group, paced by the engagement’s release cadence
Decisions madeThe order in which remaining modules are migrated (typically guided by the dependency map and risk classification from Stage 1). The deployment plan for each module group.
OutputsThe remaining modules migrated end-to-end. Deployment and testing of each module group. Ongoing agent tuning based on feedback across the engagement’s life.
Where the agents runThe Stage 3 workflow runs at scale across the remaining modules, with the agents operating at the engagement-tuned steady-state autonomy level.

Stage 5: UAT, Deployment, and Maintenance

ElementDetail
ParticipantsModernization Expert, Product Owner for UAT, client’s UAT team, deployment engineering
DurationPer release wave, then ongoing through hand-over
Decisions madeThe UAT acceptance per module. The deployment sequencing. The post-deployment monitoring window. The pace and shape of the hand-over to the client (or to a continuous SDLC engagement on the modern stack).
OutputsDeployed modern system. UAT fixes based on feedback. Technical documentation for the migrated application code. The knowledge graph in a state suitable for ongoing operation.
Where the agents runThe maintenance discipline operates the modernized system. Modernization Experts remain engaged at a reduced cadence to support change-impact analysis until hand-over to the client or to a continuous SDLC engagement on the modern stack.

The Two Iteration Loops

Stages 3 and 4 both run two distinct iteration loops in parallel. The loops are different in kind: one is automated and runs on a fast cadence; the other is human-driven and runs on a slow cadence.

LoopCadenceWhat it drives
Migration-Validation iteration loopPer migration candidate (typically minutes to hours)Re-generation of the target code candidate until the four validation gates pass
SME tuning loopPer module (typically weeks)Adjustments to agent prompts, agent configurations, and the SME’s annotations based on review of the gate-passed output

The SME tuning loop is what makes the modernization operating model accumulate engagement-specific knowledge. By the end of the MVP stage, the agents are operating at the autonomy level the engagement supports. By the end of the Scaled Migration stage, the per-module workflow is well calibrated and the SME’s annotation patterns are captured in the engagement-specific configuration.

The Expert Review Pattern

The Modernization Expert is the named human owner across the modernization project. The role pairs with the engagement’s Chief Architect on the validation side and with the Product Owner on the specification and parity-contract side.

Where the Modernization Expert reviewsWhat they validate
At the end of every migration-validation iteration where the four gates passThat the target code output is consistent with the parity contract from the SME’s annotations and the target blueprint. That the gate evidence is sufficient to advance the module to deployment.
At the end of every Stage 3 and Stage 4 module migrationThat the module is ready for deployment. That the engagement-specific agent tuning is consistent with the gate outcomes. That any Product Owner or Architect feedback has been incorporated into the configuration.
At the hand-over at the end of Stage 5That the modern system is operating as expected. That the knowledge graph is in a state that can transfer to a continuous SDLC engagement or to the client’s internal team for ongoing maintenance.

The Expert Review pattern is what makes the modernization model’s “validation, not per-step approval” stance operationally credible. The agents run autonomously through migration and validation; the Expert reviews the bounded outputs at defined gates rather than approving each agent action.

Team Composition Across the Engagement

RoleWhere they engage
Client CTO or VP EngineeringEngagement sponsor. Stage 1 scope decisions, mid-engagement reviews, hand-over sign-off.
Product Owner (one or more, per domain)Specification annotation in Stage 3. Functional gate review per module across Stages 3 and 4. UAT in Stage 5. The named human owner of the parity contract.
Architect (typically the client’s lead architect or senior engineer)Standards gate and architecture gate review per module. Validates that the modern output is consistent with the target architecture and the engagement’s standards. Pairs with the engagement’s Chief Architect.
Modernization Expert (from the implementing partner)Named human owner of the agents in the engagement. Tuning loop driver. Per-iteration Expert Review.
Engagement’s Chief Architect (from the implementing partner)Validation oversight across the four gates. Calibrates the gates’ thresholds to the engagement’s standards. Reviews migration-validation iteration outcomes.
Engineering Team for deployment (client or implementing partner)Stage 5 deployment, post-deployment monitoring.

The team is sized to the modernization scope. A 500K LOC modernization typically engages one Modernization Expert, one Chief Architect, two to three Product Owners (one per major domain), and one Architect. A 3M LOC modernization typically engages two Modernization Experts, one Chief Architect with one supporting senior architect, four to six Product Owners, and two Architects.

How the Modernization Operating Model Connects to the Continuous SDLC Operating Model

The two operating models do not overlap during their respective engagements, but they connect at the hand-over.

When the modernization completes and the modernized system is in live operation, the knowledge graph that the modernization built is fully populated against the modern stack. At this point the client has options.

PathWhat follows the modernization
Client takes over operationallyThe Source-state ontology (now the running state of the modern system) is handed to the client’s internal team. The client operates the modernized system on their own. The implementing partner may provide bounded support agreements but the operating-model engagement ends.
Client engages a continuous SDLC operating modelThe Source-state ontology becomes the seed for the four-layer Code Ontology in the SDLC instantiation. The same four custodians who owned the modernization ontologies continue as the custodians of the four-layer ontology. They articulate the ongoing intent into the Functional, Design, and Architecture ontologies. The continuous SDLC operating model resumes from the seeded graph.
Client retires the engagement entirelyThe modernized system is in production and the client has no ongoing engagement. The knowledge graph remains the client’s; the offboarding doctrine (see Enablement Partnership) governs the transition.

The three paths are not exclusive. A common engagement shape is: the modernization is delivered, then a continuous SDLC engagement runs for a defined initial period (typically eighteen months), then the client takes over operationally with a residual support agreement.

How Accion Labs operationalizes this

The ASIMOV platform implements the five-stage pipeline (Discover, Document, Migrate, Validate, Maintain) and the four validation gates that the operating model wraps. The five engagement modes (Documentation Only through Maintain / Operate) are described on the ASIMOV engagement model section. The Breeze.AI platform implements the continuous SDLC operating model that the modernization engagement can transition to at hand-over.


The Team covers the operating model that delivers continuous SDLC. Spec Sprint and Implementation Sprint are the two cadences that operating model runs on. The Enablement Partnership is the engagement frame for the implementing partner’s role across both operating models in their respective use cases.