Case Study: From Slide Deck to Production LegalTech SaaS in 6 Weeks
The Situation
The founder behind Bonus Iuri had a vision and a deadline. The vision: anyone uploading a contract — a freelancer reviewing an NDA, a tenant checking a lease, a startup closing a commercial deal — should get an instant, reliable risk assessment backed by real Spanish legislation. Not a ChatGPT wrapper. A production SaaS with proper data handling, payment processing, and legal reasoning that cites actual articles of law.
The deadline: a legal industry conference in six weeks. No extensions.
The founder had domain expertise, market validation, and a slide deck. What they didn't have was an engineering team, a codebase, or any of the infrastructure needed to turn the idea into a product that could accept payments from real users.
What Needed to Be Built
This wasn't a landing page with a waitlist. The scope for a production-ready launch included:
An AI contract analysis engine covering nine Spanish contract types — employment, commercial, rental, purchase-sale, services, NDA, corporate, franchise, and agency agreements. Each type required its own legal reasoning pipeline tuned to specific legislation, risk patterns, and checklist criteria.
A Spanish legislation knowledge base with seven consolidated laws indexed from the BOE (Boletín Oficial del Estado), searchable and citable by the AI engine, with daily synchronization to capture legislative updates.
A full SaaS platform — React frontend with contract upload, risk visualization, and animated analysis flow. FastAPI backend with 31 API endpoints. JWT authentication with email verification. Stripe integration for one-time report purchases and annual subscriptions.
EU regulatory compliance baked into the architecture — GDPR data isolation, EU AI Act transparency obligations, LOPDGDD requirements, and CCBE ethics disclaimers. Not bolted on at the end. Built as foundational constraints.
Production infrastructure — HTTPS, CDN, monitoring, CI/CD pipeline, admin panel, analytics.
Six weeks. From nothing to all of that.
How We Delivered It
The Team: Two People, CTO-Led
We deployed a two-person senior engineering team. Not a project manager and a developer — a CTO who has built multiple AI products and a senior full-stack engineer. The CTO architected the system and built the core AI reasoning pipelines. The second engineer handled the frontend, payments, deployment, and platform infrastructure.
Two people. No handoffs between teams. No integration gaps between AI, backend, frontend, and infrastructure. Every decision maker was also a builder.
The Accelerator: In-House Agentic Framework
The single biggest factor in the timeline was not working harder — it was not rebuilding infrastructure that already existed.
Conectia maintains an in-house agentic AI framework — a production-tested LLM orchestration engine with tool registry, task routing, streaming responses, and error recovery. It's been deployed across multiple AI projects. For Bonus Iuri, we adapted this framework for legal reasoning rather than building LLM infrastructure from scratch.
That adaptation saved three to four weeks. Instead of spending a month on prompt engineering, retrieval-augmented generation (RAG) setup, model integration, and error handling — problems we'd already solved — we invested that time in the legal domain logic: the specific checklist criteria, risk scoring algorithms, and legislation mappings that make a legal AI tool actually useful.
The Timeline: Week by Week
Week 1 — Architecture and compliance.
Before writing a line of product code, we mapped the data flows and made compliance decisions. GDPR data minimization dictated the storage architecture: zero retention by default, per-user isolated S3 prefixes, right-to-erasure triggering full cleanup of documents and derived vectors. EU AI Act risk classification determined the transparency requirements: every analysis would display an AI Transparency Badge disclosing the models used, their limitations, and the guarantee that user data is never used for model training.
Infrastructure was provisioned in parallel: EC2, S3, PostgreSQL, Amazon Bedrock Knowledge Bases for the legislation RAG system.
Week 2 — Core AI engine.
The document processing pipeline went live: PDF extraction (with OCR fallback via AWS Textract), Word document parsing, image handling, and audio transcription for dictated contracts. The BOE legislation ingestion pipeline was built — section-aware legal chunking that detects article, section, and recital boundaries, vectorized through Bedrock Knowledge Bases, deduplicated via SHA256 content hashing.
The agentic loop was adapted for legal reasoning: nine contract-type-specific analysis prompts, each mapping to the relevant Spanish legislation and producing structured risk assessments.
Week 3 — SaaS platform.
User authentication with JWT tokens and email verification (OTP codes). Stripe payment integration — webhook-driven subscription lifecycle for both one-time report purchases at €14.90 and annual professional plans at €490.90. The React frontend took shape: contract upload with drag-and-drop, a five-step animated analysis progress indicator, and the Semàfor de Risc (traffic-light risk visualization).
Week 4 — Legal reasoning depth.
Twelve-point checklists per contract type, each point with status indicators (OK, Warning, Missing), textual extracts from the uploaded contract, and detailed legal analysis citing specific articles from the relevant legislation. Risk matrix with severity scoring. Citation verification against BOE references — enforcing the "no source, no claim" rule where every legal assertion must include a traceable reference.
Week 5 — Polish and compliance documentation.
Branded PDF report generation. SEO landing pages per contract type. Google Analytics 4 and Tag Manager integration with funnel event tracking from registration through upload, analysis, and purchase. Cookie consent implementation, privacy policy, and the admin panel for subscription management.
Week 6 — Launch.
Production deployment with HTTPS (Let's Encrypt), CloudFront CDN, monitoring, and the full QA pass. Token cost tracking dashboard for the founder to monitor per-analysis economics.
The platform launched on schedule. Live users at the conference. You can see the result at bonusiuri.pro.
The Numbers
| Metric | Result |
|---|---|
| Time from kickoff to production | 6 weeks (42 days) |
| Team size | 2 engineers (CTO + senior full-stack) |
| Contract analysis time | 60 seconds vs. 2 hours manual review (97% reduction) |
| Contract types supported | 9 specialized legal workflows |
| Laws indexed | 7 consolidated Spanish legislations with daily sync |
| API endpoints delivered | 31 |
| File formats supported | 12 (PDF, Word, images, audio, and more) |
| Regulatory frameworks addressed | 4 (GDPR, EU AI Act, LOPDGDD, CCBE Ethics) |
| Average cost per free analysis | Under $0.01 |
| Average cost per premium analysis | $0.03–$0.05 |
Why It Worked in Six Weeks
Three factors made this timeline possible. Remove any one of them and the project would have taken three to four months.
Factor 1: Reusable AI infrastructure. Building an LLM orchestration layer, RAG pipeline, and multi-model routing system from scratch takes four to six weeks for a senior team. Our in-house framework eliminated that entire phase. The engineering time went into domain-specific logic — the legal checklists, legislation mappings, and risk scoring that define the product's value — not into infrastructure plumbing.
Factor 2: CTO-led execution. The person making architectural decisions was also writing code. No translation layer between "what should we build" and "what are we building." When the CTO decided that compliance would be an architectural constraint rather than an afterthought, that decision was immediately reflected in the data model, the storage layer, and the API design — because the same person designed and implemented all three.
Factor 3: Full-stack ownership by one team. A two-person team building everything sounds like a constraint. In practice, it was an advantage. No API contract negotiations between frontend and backend teams. No deployment coordination between infrastructure and application teams. No compliance review by a separate team that hasn't seen the codebase. One team, full context, zero handoff overhead.
What the Founder Said
The founder's feedback after launch was direct: they went from a slide deck to a production SaaS with nine contract types, Stripe payments, and EU AI Act compliance in six weeks. The engineering team didn't just implement specifications — they contributed legal domain expertise that made the AI analysis genuinely useful for practitioners.
That last point matters. A generic AI engineering team could build the infrastructure. The difference was engineers who understood why a rental contract analysis needs to check minimum duration against articles 9–10 of the LAU, or why an agency agreement must flag the mandatory client goodwill compensation under article 28 of Ley 12/1992.
What This Means for You
Six-week delivery isn't magic — it's the result of three specific conditions: reusable AI infrastructure, CTO-level engineering, and zero handoff overhead. Not every project meets these conditions, and we're transparent about that in discovery calls.
But if your project involves AI-powered analysis, SaaS delivery, and a tight timeline — particularly in regulated domains like legal, healthcare, or finance — the pattern we used for Bonus Iuri is directly applicable. The framework is built. The compliance patterns are tested. The question is how we adapt them to your domain.
Building an AI-powered product with a deadline that feels impossible? Talk to a CTO about what's realistic — and what we'd need to make it happen.
