Before committing budget or a team, leadership needed what most modernization proposals never provide: a quantitative, item-level breakdown of exactly what existed in a legacy public-sector case-management platform, and a defensible estimate of how hard it would be to rebuild.
At a Glance
| Metric | Result |
|---|---|
| Functional units assessed | ~421 |
| Complexity points (aggregate) | 966 |
| Baseline estimate (2-dev parallel) | 3.5–4.5 weeks |
| Actual delivery | 1 calendar week |
| Speed vs. baseline | ~5–8× faster |
Key Transformations
- Rigorous, auditable sizing: Every controller, model, typed action, view, and reusable UI fragment in the legacy CodeIgniter 3 / HMVC codebase was scored against a 4-tier, non-linear complexity rubric (Trivial / Simple / Medium / Complex). The result was a single comparable number instead of a vendor's round-figure guess.
- Hidden mass surfaced early: A 96-entry typed-action registry, easy to mistake for "just configuration," accounted for roughly two-thirds of the entire backend point budget. Finding it up front prevented an undersized estimate.
- Self-verifiable methodology: The report shipped with a checklist so the client could independently re-derive the numbers (counting files, methods, and fragments directly from source) before committing budget.
- Staffing modeled as a dial, not a guess: Because total effort stayed roughly flat (~6 to 9 man-weeks) regardless of headcount, the client could choose delivery speed against a known cost curve instead of an opaque quote.
- AI-directed delivery: The actual rebuild (React + Vite + TypeScript on the frontend, Node.js + Express + Prisma + Zod + JWT on the backend) was generated using AI-directed code generation (Claude Code) with human developer review, under a recommended 2-developer parallel model.
- Risks flagged, not hidden: Items the point count couldn't fully capture (a divergent code fork, complex-script PDF generation, an opaque legacy ERP connector, character-set conversion risk) were called out separately, not folded into the headline number.
Key Benefits
| Outcome | Result |
|---|---|
| Defensible go/no-go input | 966 points, decomposed to file/method level |
| Cost driver identified | Medium-tier items = 45% of count, 62% of points |
| Delivery speed | 1 week actual vs. 3.5–4.5 week conservative estimate |
| Effective throughput | ~95–100 points/dev-day vs. 25–35 baseline (~3× faster) |
Why It Matters
Modernization decisions don't have to be made on gut feel. By measuring the real cost drivers before writing a line of code, and being transparent about what the estimate does and doesn't capture, EWIS AEGIS gave leadership a conservative, evidence-based upper bound. Execution validated that AI-assisted delivery can materially outperform that baseline, turning a high-risk budget decision into a measured one.
Reference material. Prepared by EWIS AEGIS.