# Case Study: IDS Astra (CSI)

Integrated Dealer Systems (IDS) is a Constellation Software Inc. (CSI) portfolio company — Operating Group: Perseus Software

**IDS Astra (CSI) partners with raia to accelerate modernization of its legacy G2 application through an AI agent–assisted modernization sprint—using agents to ingest and understand legacy code + SQL, extract embedded business rules, and rapidly prototype modernized screens via prompt-driven code generation.** ### Snapshot (planned targets) | Metric | Before raia | After raia (target) | | --------------------------- | ---------------------------------------------------------------------- | -------------------------------------------------------------------------------------- | | Modernization velocity | Two devs in two years produced one modernized screen | Tangible progress in 3 weeks: agent stack + insights + prototype screen(s) | | Legacy system understanding | Business rules embedded in stored procedures and tightly coupled logic | Data dictionary + architecture/gap analysis + business rule catalog surfaced by agents | | Screen replication risk | High UX retraining risk if UI changes too early | Near pixel-for-pixel replication first, with a path to future UX refactor | | Engineering leverage | Slow manual discovery and redevelopment | Reusable AI-augmented dev workflow, prompt packs, runbooks, and documentation | | Executive readiness | Hard to show progress quickly | Executive demo assets packaged by end of November (prototype + reports + demo video) | > **Note:** This case study describes the **planned solution and target outcomes** of a three-week sprint. It is intended to validate feasibility and produce decision-ready artifacts—not deliver full production modernization in three weeks. ### 1 | Company & Challenge IDS Astra’s G2 application, a core part of its business, was built on a decades-old legacy architecture. Critical business logic was tightly coupled to the data layer and embedded in a maze of stored procedures, resulting in a highly coupled legacy architecture that made modernization extremely difficult and slow. Progress through traditional approaches was stalling: two developers had spent two years producing just one modernized screen. IDS needed a time-boxed way to validate whether an AI-first approach could accelerate discovery and delivery while preserving business logic and minimizing user retraining risk. ### 2 | Why IDS Astra chose raia IDS Astra selected raia to run a fast, controlled modernization sprint designed to produce practical outputs: * **Agent-assisted discovery:** ingest GitHub + SQL assets to enable high-quality Q\&A, cross-repo search, and dependency understanding. * **Business-rule visibility:** extract hidden rules from stored procedures and legacy code into human-readable documentation. * **Prompt-coded prototyping:** rapidly replicate a legacy screen (“Sales Quote”) in a modern stack for apples-to-apples validation. * **Reusable workflow:** deliver prompt packs, runbooks, and an AI-augmented development process that IDS can reuse beyond the sprint. ### 3 | Solution Design The engagement is a three-week, AI agent–assisted modernization sprint with four working components. | Component | Role | Key Integrations | | ----------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ | | **Codebase + Database Intelligence Agent** | Ingests IDS GitHub repos + SQL Server assets into a vector store; generates data dictionary; enables cross-repo code Q\&A and architecture/gap analysis. | GitHub, SQL Server (Olympus schema + backup), raia platform | | **Prompt-Coded Screen Replication (Sales Quote Prototype)** | Builds a modern, working version of the legacy Sales Quote screen with near pixel/flow parity; connects to SQL; deploys a demo environment. | Prompt-driven code generation tooling (e.g., IDE copilots), GitHub, SQL Server | | **Hidden Business Rule Discovery Agent** | Parses stored procedures and key code paths; extracts rules/validations; creates a versioned Business Rules Catalog + parity test plan. | raia platform, GitHub, SQL Server | | **AI-Augmented Dev Workflow & Orchestration** | Sets up operational foundations: ingestion pipelines, prompt templates/packs, governance, enablement runbook and training. | raia Launch Pad/Copilot/Academy, GitHub ETL hooks, SQL Server | #### Guardrails (by design) * **API/asset access controls:** least-privilege GitHub + SQL access; UAT/sample data where possible. * **Scope guardrails:** prioritize modules tied to Sales Quote flow; incremental ingestion for very large repos. * **Quality controls:** SME validation loop; parity checklist and test plan to confirm prototype behavior matches legacy rules. * **Governance & observability:** audit logs and versioned embeddings to reflect evolving codebases. ### 4 | Implementation Timeline (3-week sprint) A rough sprint timeline was defined as: * **Week 1:** Access + ETL setup + initial ingestion; generate data dictionary; Gap/Architecture Report v1. * **Week 2:** Prompt pack creation; Sales Quote prototype generation; initial business rule extraction. * **Week 3:** Prototype refinements; parity validation; Business Rules Catalog v1; enablement sessions; demo packaging. Estimated effort: **\~90–110 hours** (baseline 100 hours over 3 weeks). ### 5 | Deliverables By sprint end, planned deliverables include: * Running Intelligence Agent connected to GitHub/SQL vector store * **Data Dictionary v1** + **Gap/Architecture Report v1** * **Sales Quote prototype** with SQL connectivity; code committed to GitHub; short demo video * **Business Rules Catalog v1** with parity test plan * AI Development Runbook, prompt templates, and enablement session recordings ### 6 | Expected Impact (targets; measured results to follow) IDS Astra’s target outcomes for the sprint include: * **Accelerate discovery and reduce uncertainty** by surfacing data definitions, dependencies, and hidden rules quickly. * **Demonstrate feasibility of AI-first modernization** with a working, comparable prototype screen. * **Create decision-ready modernization artifacts** that inform a go-forward plan and executive demo by end of November. * **Establish a repeatable AI-augmented dev workflow** IDS can reuse across additional screens and modules. ### 7 | Lessons for Prospective Clients 1. **Start with intelligence, not refactors.** Vectorizing code + schema and extracting rules reduces modernization risk before rebuilding. 2. **Replicate before you redesign.** Pixel/flow parity provides a safe benchmark and lowers retraining risk. 3. **Treat business rules as first-class artifacts.** A versioned catalog + parity tests prevents silent behavior drift. 4. **Time-box to validate.** A three-week sprint can prove feasibility and produce the artifacts needed for a confident roadmap. ### 8 | What’s Next Following the sprint, IDS Astra can extend the same workflow to additional G2 screens and modernization waves—using the gap report, rules catalog, and prototype learnings to prioritize modules and sequence delivery toward a broader modernization program. ### 9 | Key Takeaway IDS Astra is using raia to run a three-week, AI agent–assisted modernization sprint that turns a hard-to-understand legacy system into actionable intelligence (data dictionary + architecture insights + business rules) and a working modern prototype (Sales Quote)—creating an executive-ready demo and a repeatable path to modernize G2 faster while preserving core business logic.