SME•BIZ

Inspiration

My friend Promise owns a small computer shop — typing, printing, registrations. We ride to evening events together on his bike, and almost every trip he vents about the same thing: "I still have to go meet my boy at the shop to close up and check if my account is balanced."

I'm an analyst, so I asked the obvious question: why not get software built for that? He laughed. "Do you know how much that costs? And the yearly fees? Let me just continue with my manual method o abeg."

I looked into it anyway — partly to prove him wrong. Turns out he was right. Custom software ran into hundreds of thousands of naira and months of waiting, and the "affordable" alternatives were either too rigid or too complicated for a shop like his.

So I built him something myself. Streamlit, because that's what I knew. A month later he had a working prototype, and he loved it enough to brag to two of his friends — who then asked for their own.

Streamlit couldn't hold up at scale, so I rebuilt on a leaner stack with an AI coding agent doing the heavy lifting. What took a month the first time took under seven days the second — without me missing a day of my own job. That's when it clicked: this wasn't just a favor for a friend anymore, it was a business model. Low cost, minimal infrastructure — a laptop, a phone, decent internet — the exact resources most SMEs around me already have.

That's how SME Biz was born: out of one friend's frustration on the back of a bike, built to answer a bigger question — what if AI-native development could make custom software actually reachable for the SMEs that need it most?

Small and medium-sized enterprises (SMEs) are the backbone of local economies, yet they are trapped in software purgatory.

On one side, off-the-shelf SaaS applications force business owners into rigid, one-size-fits-all workflows that never quite match how their shop actually runs. On the other side, traditional custom software development requires tens of thousands of dollars and months of delays—an investment far beyond the reach of most independent merchants, repair shops, and distributors.

As a result, business owners spend hours every evening manually reconciling fragmented spreadsheets, chasing overdue debtors on WhatsApp, and calculating margins on sticky notes.

We were inspired by a fundamental question: What if an AI-native coding workflow could reduce the cost and delivery time of custom enterprise software by 90%? What if a single operator—paired with an AI agent—could print tailored operational command centers for SMEs in days rather than months?

That is the vision behind SME•BIZ.


What it does

SME•BIZ is an AI-native app factory and operational command center designed to give SME owners 100% relief from operational chaos. Instead of generic templates, we deploy private, custom-fitted portals tailored explicitly to each business’s unique transaction rules, staff roles, and inventory workflows.

Key features & pricing structure include:

  • Centralized Operational Ledgers: Record every sale, purchase, and repair job in seconds with mistake-free transactional entries.
  • Real-Time Financial Analytics: Live tracking of gross margins, daily cash flow trends, and customer debtor ledgers.
  • Plain-English Financial Assistant: Business owners can ask natural, plain-English questions—such as "Who owes us money right now?" or "What was our highest-margin item last week?"—and get instant answers strictly grounded in their own verified database records.
  • Automated Nightly Digests: End-of-day summary reports delivered automatically via PDF invoices and secure Telegram digests.
  • Private Vault Data Security: Enterprise-grade data isolation powered by Row-Level Security (RLS) ensuring each tenant's records are strictly isolated.
  • Straightforward SME Pricing:
    • Starter: ₦50,000 setup (+ predictable monthly retainer) for single-location sales, ledgers & AI Q&A.
    • Growth: ₦150,000 setup (+ predictable monthly retainer) for multi-staff roles, debtor ledgers & automated PDF invoicing.
    • Enterprise: Custom Quote for multi-location command centers & custom integrations.

How we built it

We built SME•BIZ by pairing a Human Product Manager / QA Lead with an AI Coding Agent (Antigravity IDE powered by Gemini 3.5 models).

+---------------------+       +-----------------------+       +------------------------+
| 1. Client Intake    | ----> | 2. AI Spec Adaptation | ----> | 3. Phased Execution    |
| (client_req.md)     |       | (todos.md)            |       | (Antigravity + Gemini) |
+---------------------+       +-----------------------+       +------------------------+
                                                                          |
                                                                          v
+---------------------+       +-----------------------+       +------------------------+
| 5. Production Tenant| <---- | 5-Phase QA Checkpoint | <---- | 4. Human Verification  |
| (Live Handover)     |       | (Human-in-the-Loop)   |       | (Automated & Manual)   |
+---------------------+       +-----------------------+       +------------------------+

Technical Stack & Architecture

  • Frontend: React + TypeScript + Vite + Vanilla CSS / Tailwind CSS designed with glassmorphism and modern dark/light mode UI components.
  • Backend API: FastAPI (Python 3.12, PydanticAI, SQLModel, Alembic).
  • Multi-Tenant Foundation: Hosted Supabase (PostgreSQL) enforcing strict Row-Level Security with business_id tenant isolation.
  • Automated Operations: Nightly Cloud Scheduler triggers generating PDF reports and Telegram operational digests.

The Math Behind Our Efficiency (fixed for Devpost)

Traditional agencies scale linearly: total build time = hours × billing rate, summed across every developer on the project. More scope always means more hours, means more cost.

We break that curve. By letting an AI agent handle the repetitive work — schema migrations, API routes, UI components — while a human operator acts purely as QA gatekeeper, we get roughly an 8.5x velocity multiplier:

Agent build time ≈ (Traditional build time ÷ 8.5) + Human verification time

That's what makes ₦50,000 (Starter) and ₦150,000 (Growth) setup pricing possible — a live production portal in 4–10 days instead of months.


Challenges we ran into

  1. Eliminating AI Hallucinations in Financial Queries: When business owners ask plain-English questions about cash flow or debtors, inaccurate answers are unacceptable. We solved this by implementing bounded PydanticAI tools that enforce SQL execution constraints and validate database outputs before presenting them to the user.

  2. Ensuring Strict Multi-Tenant Data Isolation: Because multiple client portals share architectural foundations, preventing cross-tenant data leakage was a top priority. We resolved this by embedding strict PostgreSQL Row-Level Security policies (USING (business_id = current_setting('app.current_business_id'))) verified by automated multi-tenant isolation tests.

  3. Balancing AI Speed with Human Quality Control: Allowing an AI agent to write full applications unguided often leads to architecture drift. We engineered a 5-Phase Human-in-the-Loop Workflow (.agents/skills/build-client-app) where the agent pauses at defined gates (Database Schema, API Ledgers, AI Assistant, Frontend UI, Deployment) for explicit human sign-off.


Accomplishments that we're proud of

  • Deployed Real Client Portals: Built and onboarded real business command centers, including our flagship tech retail client Promise Computers (01_promise_tech) and manufacturing client Plastic Biz Demo (02_mobile_plastic).
  • Instant Plain-English Analytics: Successfully connected a Gemini-powered agent to live SQL databases with 0% hallucination rates on financial metrics.
  • 90% Reduction in Build Time: Reduced average client onboarding time from weeks to just 4–7 business days.
  • High-Converting SME Landing Page: Built a responsive, outcome-oriented website with interactive lightbox showcases and clear ₦50,000 / ₦150,000 / Custom Quote pricing tiers.

What we learned

  • Non-Technical Business Owners Value Relief, Not Tech Jargon: SME owners don't care about "multi-tenant schemas" or "LLM tool calling". They care about saving 2 hours every night, knowing who owes them money, and having clean records.
  • Human-in-the-Loop is Essential for Production Code: AI coding agents are extraordinarily fast, but human oversight at key architectural gates is what separates an experimental script from a production-grade enterprise app.

What's next for SME Biz

  • Vertical-Specific App Templates: Expanding our starter template library to support specialized verticals such as auto repair shops, pharmacies, and logistics providers.
  • Automated WhatsApp Business Integration: Enabling owners and customers to send transaction receipts and query debtors directly via WhatsApp.
  • Agency Partner Network: Licensing our AI-native app factory workflow to independent software consultants so they can build high-margin software agencies for local businesses in their cities.

Built With

  • ai-studio
  • antigravity-ide
  • gemini
Share this project:

Updates