WebisoftDeal Execution Platform
A development proposal for Good Morning Capital — feasibility, recommended approach, timeline, cost, and IP ownership, addressed directly.
Prepared by Phil Therien, Partner, Webisoft Technologie Inc.
Executive Summary
Webisoft has reviewed the GMC Deal Execution Platform specification (V1, April 2026) and the discovery session notes in full. This proposal addresses the four areas you requested: feasibility, recommended approach, timeline, and cost. It also covers IP ownership directly.
Our position is straightforward: this is a well-specified, buildable platform. The specification reflects genuine domain knowledge, not a generic PropTech wishlist. The underwriting logic is deterministic, the data model is coherent, and the workflow maps cleanly to a standard modern stack. There are no architectural surprises.
One question your brief raises is whether Odoo could serve as a foundation. We address this directly in Section 02. Our recommendation is to build from scratch on a purpose-fit stack — the reasons are practical, not ideological.
V1 Launch
Months 1–4
V1 Cost Range
$105K–$146K
Full Platform (V1–V3)
~$280K–$381K
IP Ownership
100% GMC
Recommended Approach
On the Odoo Question
Odoo is a capable ERP for standard business operations: accounting, inventory, HR, CRM. It is not a fit for this platform, and the reason is architectural rather than a matter of preference.
GMC's platform has three highly specialized components that Odoo cannot accommodate without becoming an obstacle:
Deterministic underwriting engine
Must produce identical outputs from identical inputs, with full formula traceability and no black-box behavior.
Document ingestion pipeline
PDF extraction, classification, confidence scoring, and rent roll validation — none of which Odoo has any native support for.
Server-side lender package generator
Renders watermarked, version-controlled institutional PDFs from live financial data with no copy-paste.
Forcing these into Odoo means building three custom Python modules that fight the framework at every step, while still paying Odoo licensing fees and inheriting its data model. You'd own neither the architecture nor the ergonomics.
Recommended Stack
The spec was written by someone who understood what the platform needs. We recommend building on exactly the stack it calls for.
| Layer | Technology | Why |
|---|---|---|
| Frontend | React + TypeScript | SPA with real-time underwriting recalc client-side |
| Backend | Node.js + Express (V1), GraphQL (V2) | Stateless, clean REST API, job queue support |
| Underwriting Engine | Pure TypeScript functions (shared) | Deterministic, unit-tested, same logic client + server |
| Document Ingestion | Python / FastAPI microservice | pdfplumber, pandas — best ecosystem for this job |
| PDF Generation | Node.js + Puppeteer | Server-side HTML-to-PDF, version-controlled, S3 stored |
| Database | PostgreSQL + Redis | Relational integrity for financials, Redis for job queue |
| AI Layer | Anthropic Claude API | Chase Agent drafts, future narrative drafting (V2) |
| Auth | Supabase Auth or Auth0 | Magic link support for lender portal, RBAC |
| Storage | AWS S3 / Cloudflare R2 | Signed URLs for documents, versioned PDF packages |
Feasibility Assessment
Module-by-module assessment. All four core modules rate High feasibility — the spec is detailed enough to scope and build directly.
Underwriting Engine
The spec is clear — pure functions, deterministic output, full formula trace, unit-tested against GMC's existing Excel models. This is standard engineering done carefully. There is one parked item (CMHC vs conventional normalization benchmarks) that requires a working session with the GMC advisory team before the CMHC engine is finalized. We would schedule this in Month 1 and use conventional-first sequencing to avoid blocking progress. The underwriting engine is the highest-scrutiny module in the build and receives exhaustive testing before any live deal touches it.
Document Ingestion
Digital PDFs and Excel files extract cleanly using pdfplumber and pandas. Scanned documents require manual data entry in V1 — this is explicitly noted in the spec and is the right call. The classifier is rule-based in V1, with AI confidence-boost for borderline cases. This is conservative and correct. Attempting ML classification in V1 without a training dataset would introduce unreliability into a process that must be reliable. Scanned document handling and ML classification are V2 upgrades that become tractable once real data accumulates.
User Roles and Permissions
Seven roles with clearly defined visibility layers map directly to a standard RBAC model. The permission matrix in the spec is complete and unambiguous. Magic-link lender access requires no account management overhead. The external advisor tier (V2) adds complexity around licence verification and revenue share logic, but it's well-scoped. No technical concerns here.
Lender Package Generator
Server-side Puppeteer rendering from HTML templates is a proven approach for institutional document generation. Data flows directly from the underwriting model at generation time. The dual-template design (CMHC and conventional) is architecturally clean — same data layer, two rendering templates. Watermarking, version numbering, and S3 storage are standard. Narrative sections require advisor input in V1; AI drafting is a V2 enhancement. This is the right sequencing — get the data layer right before adding AI generation.
Timeline
Three phases over 18 months. V1 is the critical path — external advisors and developers should not be onboarded until the internal platform is proven on live deals.
Months 1–4
Key deliverables
Deal workspace · document intake + chase · underwriting engine (conventional + CMHC) · red flag engine · lender package generator (both templates) · org chart builder · sponsor/borrower module · mandate generator · lender portal · fee tracking · AI Chase Agent · principal dashboard · lender CRM
Months 5–9
Key deliverables
External advisor onboarding + tier system · revenue share automation · developer portal (portfolio, renewal tracker, document vault, messaging) · market intelligence layer · Ontario expansion (FSRA) · AI Package Reviewer · AI Narrative Drafter
Months 10–18
Key deliverables
Institutional tier (white-label, multi-seat, API) · lender appetite profiles + AI Lender Matcher · AI Market Intelligence Agent · automated data feeds · national expansion · developer performance scoring
We recommend a 4-to-6-week internal pilot with GMC advisors on real files before V2 onboarding begins.
Cost
Preliminary ranges based on the specification. Firm pricing follows a scoping session where we walk through the spec module by module and lock the functional requirements.
V1 — GMC Internal Platform
Months 1–4
| Module | Low | High |
|---|---|---|
| Deal workspace + stage management | $8,000 | $11,000 |
| Document intake, classification + chase management | $12,000 | $16,000 |
| Underwriting engine (conventional + CMHC) | $18,000 | $25,000 |
| Red flag + normalization engine | $8,000 | $11,000 |
| Lender package generator (both PDF templates) | $14,000 | $19,000 |
| Org chart builder | $5,000 | $7,000 |
| Sponsor + borrower module | $5,000 | $7,000 |
| Mandate generator (legal review external) | $4,000 | $6,000 |
| Lender portal (magic link, read-only, Q&A) | $5,000 | $7,000 |
| Fee tracking + invoice automation | $4,000 | $6,000 |
| AI Chase Agent | $5,000 | $7,000 |
| Principal reporting dashboard | $5,000 | $7,000 |
| Lender relationship management | $4,000 | $6,000 |
| Auth, RBAC, infrastructure, DevOps | $8,000 | $11,000 |
| V1 Total | $105,000 | $146,000 |
Note
The mandate generator requires external legal review by a Quebec lawyer to validate Civil Code compliance before launch. This is not included in the development budget and is GMC's responsibility to procure.
V2 + V3 Estimates
External advisors + developer portal
$90,000 – $120,000
Intelligence layer + national scale
$85,000 – $115,000
Full Platform Total
V1 + V2 + V3 combined
~$280K – ~$381K
V2 and V3 estimates will be refined at the end of each prior phase based on actual build learnings.
IP Ownership
Webisoft's standard terms for custom development engagements assign full IP ownership to the client on payment of each milestone. This means:
All code is yours on delivery
No license fees, no vendor lock-in, no conditions.
All data is yours
Deals, underwriting models, cap rate database, lender appetite map — all in your codebase, on your infrastructure.
All proprietary logic is yours
Underwriting formulas, normalization rules, red flag engine — exists in your codebase and nowhere else.
The platform is resaleable
License it, white-label it, or spin it out independently. Webisoft's involvement does not constrain any of those options.
Captured in our standard MSA. We are happy to work from a client-form agreement if you have one — send it over and we'll mark it up.
Re-Sale Potential
The platform as designed has genuine re-sale potential. The spec already contemplates this — external advisor tier, white-label institutional tier, and SaaS pricing model are built in from V1.
What makes this realistic rather than aspirational:
Defined, addressable market
Underwriting logic specific to Canadian commercial mortgage brokerage — no dominant purpose-built software player.
CMHC expertise is a barrier
Most generic platforms don't handle CMHC program logic, normalization benchmarks, or MLI Select scoring.
Quebec Civil Code compliance
The mandate generator's Civil Code handling is a differentiator in the Quebec market specifically.
Data moat compounds
A platform with 500 Quebec CMHC deals processed carries lender appetite data and cap rate benchmarks a new entrant cannot replicate.
Practical recommendation: build V1 for GMC internal use first. Prove the platform on real deals. The re-sale case becomes credible once there is transaction history behind it. Webisoft will architect for multi-tenancy and white-labeling from the outset so the foundation supports that path when you're ready.
Engagement Model
Milestone-based, with bi-weekly deliverable reviews. No long gaps between visibility windows.
| Milestone | Deliverable | Timing |
|---|---|---|
| M1Kickoff + Scoping | Finalized functional spec, stack confirmed, CMHC working session completed, dev environment live | Week 2 |
| M2Core Infrastructure | Auth, RBAC, deal workspace, document upload + classification, database schema | Week 6 |
| M3Underwriting + Red Flag | Conventional underwriting engine, red flag engine, normalization rules — tested against GMC Excel models | Week 10 |
| M4Package Generator | Lender package PDF generation (both templates), org chart, sponsor module, mandate generator | Week 14 |
| M5Operations Layer | Chase Agent, fee tracking, lender portal, principal dashboard, lender CRM | Week 17 |
| M6V1 Launch | Internal pilot with GMC advisors on live deals, bug fixes, performance hardening | Weeks 18–20 |
Team for V1
- 2senior full-stack engineers
- 1Python/data engineer (ingestion + AI layer)
- 1UX/UI designer (first 6 weeks focused on design system)
- 1Phil — account lead & technical oversight throughout
Next Steps
Review this proposal
Raise any questions — available for a call this week or next.
Schedule a 90-minute scoping session
Andy and Phil walk through the spec module by module, confirm priorities, and issue a fixed-price SOW for V1.
Legal: MSA structure
Share your preferred MSA or we send Webisoft's standard agreement for your review.
Identify the Quebec lawyer
For mandate generator review — on the critical path for Stage 3 gating.
We're not proposing a paid discovery phase. The specification is detailed enough to scope immediately and issue a fixed-price SOW for V1 off the back of one working session.
Direct contact: phil@webisoft.com
Webisoft · Montreal, QC · webisoft.com
Confidential — Prepared exclusively for Good Morning Capital