How do you handle AI reply accuracy and avoid wrong responses?

We implemented a custom confidence scoring engine. Every AI response gets scored; if confidence is below a business client's configured threshold, the message is routed to a human review queue instead of sending an automated reply.

How is multi-tenant security guaranteed?

Tenant isolation is enforced at three distinct layers: JWT token claims at the API layer, tenant scope validation at the application service layer, and mandatory tenant-id filtering in database middleware.

What is the typical cost range for building a platform of this scope?

Developing a custom SaaS platform of this complexity (webhook gateway, AI classification, interactive commerce, multi-tenant UI, auth) typically ranges between ₹5–15 lakhs depending on scope depth.

PortfolioWavo.digital

Visit Live Platform

SaaS · AI Automation

Wavo.digital

Building an AI WhatsApp Marketing & Conversational Commerce SaaS from 0 to Enterprise Production.

Request Similar System All Case Studies

Client

Oaksol Technologies

Industry

SaaS · AI Automation

Role & Scope

Lead Full-Stack & API Architect (Sole Developer)

Duration

14 Weeks

0 → Production 14 Weeks Solo

Build Velocity

99.9% Zero Incidents

System Uptime

100% Zero Leaks

Data Security

Executive Summary

A funded Indian startup needed an enterprise-grade platform built from scratch in 14 weeks. Wavo.digital receives thousands of real-time Meta webhooks, processes them through an AI reply engine, executes in-chat commerce flows, and surfaces analytics in a multi-tenant React dashboard. Engineered solo by TheUnipick with 99.9% uptime since production release.

Commercial Background

About Oaksol Technologies

Oaksol Technologies is an Indian technology startup focused on conversational commerce — using messaging platforms, particularly WhatsApp, to automate customer communication and drive sales for businesses. With over 500 million active WhatsApp users in India, SMBs and e-commerce brands respond to hundreds of customer messages daily manually. Founder Dhruv needed a high-performance system to automate routine inquiries, present product catalogs, and close sales directly inside WhatsApp.

Problem Statement

The Core Challenge

Building an enterprise WhatsApp platform requires processing unpredictable real-time webhook bursts, executing intelligent conversational replies, managing in-chat catalog purchasing, and isolating client data with multi-tenant security.

Key Pain Points & Technical Constraints:

High-volume, real-time message processing via Facebook Graph & WhatsApp Cloud API webhooks

Contextual AI reply generation balancing instant answers with human fallback thresholds

In-chat product catalog browsing and checkout state management inside WhatsApp conversations

Strict multi-tenant security guaranteeing data isolation across client businesses

Architectural Solutions:

Facebook Graph Webhook Gateway

Event-driven Node.js gateway acknowledging Meta webhooks sub-second while queuing async payload processing.

AI Automation & Intent Engine

Context-aware reply engine matching FAQs, classifying customer purchase intent, and enforcing confidence thresholds.

In-Chat Interactive Commerce Pipeline

Interactive product catalogs and checkout state machines executing direct sales within WhatsApp.

Multi-Tenant React Dashboard

Clean, role-based dashboard for non-technical users to build campaigns and monitor live chat streams.

Discovery & Research

The Discovery Process

We spent two weeks mapping the end-to-end journey for three distinct actors: business clients, end customers on WhatsApp, and platform administrators. We identified early that the Facebook Webhook Gateway was the critical path. Everything else — AI engine, dashboard, commerce flows — depended on a stable, high-throughput gateway for receiving and verifying messages. We prioritized stabilizing the webhook ingestion pipeline before layering UI and intelligence.

Architectural Choices

Engineering Strategy Rationale

Node.js was chosen over Python for the core gateway because incoming webhooks (500+ events/sec under peak load) represent an I/O-bound workload, where Node's event-driven non-blocking architecture excels. Python was isolated strictly for CPU-bound AI context processing. Multi-tenancy was enforced at both the database schema level (tenant_id) and database middleware layer to prevent cross-tenant data leakage by design rather than relying solely on developer query discipline.

Architecture & Data Pipeline

Technical Solution Flow Diagram

Interactive Pipeline • Click node to inspect

Stage 01 Detail Breakdown

React Dashboard

Client Automation Setup

Client configures automated AI reply rules, product catalogs, and escalation thresholds in the React dashboard.

Core Technology Component: React Dashboard

Engineering Resilience

Challenges & Architectural Solutions

Technical Challenge

Facebook Webhook Retries causing duplicate automated message processing during high traffic bursts

Engineered Solution

Implemented idempotency keys fingerprinting incoming Meta payloads before processing; duplicate events are acknowledged immediately and safely discarded

Business Outcome

Eliminated duplicate customer replies and prevented invalid bot behavior under heavy traffic

Technical Challenge

Meta API strict acknowledgment timeout (sub-second response required)

Engineered Solution

Decoupled webhook acknowledgment from AI processing using an asynchronous event-queue processing pattern

Business Outcome

Guaranteed 100% webhook receipt compliance without Meta gateway drops

Technical Challenge

AI overconfidence leading to potential inaccurate responses to customer inquiries

Engineered Solution

Engineered configurable AI confidence thresholds with a human-agent escalation queue

Business Outcome

Allowed clients to balance automated speed vs response quality based on risk tolerance

Technical Challenge

Multi-tenant query degradation as database volume scaled rapidly

Engineered Solution

Added compound indexes (tenant_id, conversation_id, timestamp) and cursor-based pagination

Business Outcome

Maintained sub-50ms dashboard query speeds at 10× initial test volume

Execution Timeline

Structured Phase Execution

01 4 weeks

API Architecture & Gateway

Built and stress-tested the Facebook Graph webhook ingestion engine with signature verification and event queues.

02 4 weeks

AI Engine & Commerce Flow

Implemented contextual reply logic, confidence scoring, and WhatsApp interactive message catalog templates.

03 4 weeks

Multi-Tenant React Dashboard

Designed and constructed the client UI for campaign creation, live conversation takeover, and analytics.

04 2 weeks

Production Ops & Security Audit

Enforced DB middleware tenant isolation, JWT authentication, and automated zero-downtime deployment.

Measured Impact

Delivered Outcomes

Concrete business and performance results achieved upon production deployment:

99.9% system uptime maintained continuously across first 90 days post-launch

Zero cross-tenant data isolation incidents in testing or production

Full production system delivered solo in 14 weeks from blank repo to investor launch

Direct investor confidence validation cited by the founding team during funding rounds

Visual Showcase & Interface Highlights

Click image to expand

Expand Preview

Expand Preview

Commercial Asset

Long-Term Client Value

Oaksol Technologies received a clean, modular technical foundation. The decoupled architecture ensures new payment gateways or AI models can be integrated without touching the core gateway. Written documentation and strict linting enabled seamless developer onboarding without technical debt.

Engineering Insights

Key Architecture Takeaways

Webhook-heavy systems require decoupling receipt acknowledgment from heavy processing to avoid retry storms.

Security in multi-tenant SaaS must be enforced at both application code and DB middleware layers.

Building confidence-calibrated AI fallbacks is essential for preserving client trust in B2B automation.

System Evolution

Future Platform Roadmap

Phase 02

Voice AI message transcription & automated voice note responses

Phase 03

Shopify & WooCommerce real-time product catalog auto-sync

Phase 04

A/B testing for automated conversational conversion paths

Phase 05

Sentiment analysis triggering pre-churn human escalation alerts

Client Knowledge Base

Frequently Asked Questions

Technical and strategic answers regarding the execution and architecture of Wavo.digital.

Wavo.digital was delivered in 14 weeks by a single full-stack developer. For custom enterprise SaaS platforms with AI automation, in-chat commerce, and multi-tenant dashboards, a 12–16 week timeline is realistic.

"TheUnipick took our complex requirements and delivered a production-ready system that impressed our investors. The architecture is rock-solid."

Dhruv Verified Client

Founder, Oaksol Technologies

Stack & Infrastructure Used

ReactNode.jsMongoDBMeta Graph APIWhatsApp Cloud APIJWTVercel

Next Case Study

Dear Valentine

Customizable Valentine's Story & Daily Task Gifting SaaS — Global reach with viral referral mechanics.

Read Case Study

Ready to Scope Your Next Digital System?

We design and build production-grade web platforms, SaaS applications, and scalable architectures.

Book Technical Scoping Call