Nine AI builds.
All in production.

From communication intelligence and workforce analytics to legal AI and fleet management — real systems built by Vanitech Labs, deployed for real enterprises, delivering measurable outcomes.

09
AI Case Studies
150+
Projects Delivered
100%
Client Satisfaction
>90%
Client Retention
ABSA
Banking · SA
Capitec
Banking · SA
Nedbank
Banking · SA
Adumo
Payments · SA
Diesta
InsurTech · UK
ONGC
Oil & Gas · India
Case Study 02 ML · Predictive Analytics

AI Employee Productivity Dashboard

Real-time insights, intelligent automation, and AI-driven performance analytics turning raw employee activity data into actionable workforce intelligence.

PythonTensorFlowPyTorchFastAPIJira APISlack API
Live
KPIs & OKRs
Predictive
Planning
Sentiment
Detection
Personal
Coaching
View Case Study
Case Study 03 RegTech · LLM · A2P

AI Website Verification for 10DLC Compliance

LLM + chatbot system that automatically vets client websites against 60–70 critical 10DLC A2P messaging compliance checks — under 5 minutes with 95%+ accuracy.

OpenAI GPT-4LangChainPlaywrightFastAPIReact
<5 min
Per Site
95%+
Accuracy
60–70
Auto Checks
Dozens
In Parallel
View Case Study
Case Study 04 Sales AI · CRM Integration

AI-Powered Lead Follow-Up System

A virtual sales assistant with 360° memory across email, chat, SMS, calls, and meetings — scaling personalised follow-up to thousands of leads with zero missed touches.

OpenAIAnthropic ClaudeLangChainPineconeTwilio
+35%
Conversion
−50%
Sales Cycle
−60%
Manual Effort
0 missed
Follow-Ups
View Case Study
Case Study 05 CRM Automation · LLM

AI Chatbot for Go High Level CRM

An AI assistant letting non-technical staff run GHL CRM through natural-language chat — from Slack, WhatsApp, or directly inside GHL — eliminating all menu navigation.

OpenAINode.jsSpring BootGHL APIWhatsApp API
Chat-first
Zero UI
100%
Action Logged
Any team
No Training
Real-time
Sync
View Case Study
Case Study 06 Computer Vision · MLOps

AI Crowd Detection & Tracking System

Real-time computer-vision system detecting, counting, and tracking crowds across CCTV, drone, and aerial cameras — adaptive scene intelligence, zero manual tuning.

YOLOv8ByteTrackPyTorchOpenCVNVIDIA GPU
+30–50%
Count Accuracy
Real-time
Inference
Multi-cam
Compatible
Zero
Manual Tuning
View Case Study
Case Study 08 Automation · Browser AI

Visa Appointment Booking Automation System

Enterprise-grade browser automation monitoring visa slots in real time, solving CAPTCHAs, and booking appointments across multiple centres in parallel — 24×7 production deployment.

Node.jsPuppeteerPM2SupabaseN8N
24/7
Slot Monitor
Parallel
Multi-Session
Auto
CAPTCHA
Live
Production
View Case Study
Case Study 09 Logistics SaaS · IoT

Fleet Management & Cargo Tracking Platform

White-label multi-tenant SaaS combining hardware GPS tracking, electronic cargo seals, customs compliance, and native mobile apps for transport and logistics companies.

Node.jsReact NativeMQTTWebSocketsAWS
Multi-tenant
Architecture
5+ Protocols
Hardware GPS
E-Seal
Customs Ready
Native
iOS + Android
View Case Study
In Their Words

What our clients say about working with us.

We are extremely pleased with the quality of testing delivered. The team demonstrated strong domain expertise, ensured comprehensive coverage, and significantly improved the stability of our application. Their proactive approach and commitment helped us achieve a smooth and timely release.

The automation strategy implemented by the team has been a game-changer for us. Test execution cycles have been reduced considerably, enabling faster releases without compromising on quality. Their focus on scalable automation frameworks is truly commendable.

The team showcased excellent collaboration throughout the engagement. They worked as an extended part of our team, taking complete ownership and consistently aligning with our expectations. Thanks to the thorough testing approach, we observed a significant reduction in production defects.

Beyond execution, the team brought valuable insights and suggestions that enhanced our overall QA strategy. From requirement analysis to final release validation, they demonstrated strong end-to-end testing capabilities. Their early involvement helped identify gaps sooner and improved overall delivery confidence.

The eQE team played a critical role in elevating our overall product quality. Their early involvement in the lifecycle, combined with a strong focus on risk-based testing, enabled us to significantly reduce production defects. The structured approach to test automation has delivered tangible business value.

Your project next

Ready to build the next case study?

We take on projects where the outcome matters and the brief is ambitious. Tell us what you're building — we'll tell you honestly if we're the right team for it.

Book a Discovery Call Send your project brief
Case Study 01  ·  Enterprise Productivity · NLP · Generative AI
AI Communication Priority Dashboard
A unified inbox that uses NLP, generative AI, and predictive analytics to filter, summarise, and rank communication across Email, Slack, Teams, and WhatsApp — so employees act on what matters first.
OpenAIAnthropic ClaudeLangChainPythonFastAPIPostgreSQLRedisReactSlack APIMS Graph

The Problem

  • Employees lose 40–60% of their week to fragmented inboxes and notifications
  • Critical client escalations buried under team chatter and FYIs
  • Context-switching destroys deep work; response times slip silently
  • No visibility into communication overload at team or org level

Our Solution

  • Unified inbox across Email, Slack, Teams, WhatsApp, and internal chat
  • AI ranks every message HIGH / MEDIUM / LOW by urgency + sender role + project
  • Generative AI condenses long threads into 2–3 line action summaries
  • Auto-extracts tasks and pushes them into Jira, Asana, or Trello

Key Capabilities

  • Smart Prioritization — Ranks messages by urgency, sender, and project relevance
  • AI Summaries — Long threads compressed into 2–3 line digests
  • Action Extraction — Tasks and deadlines auto-pushed to ticketing systems
  • Focus Mode — Suppresses low-priority chatter during deep work blocks
  • Burnout Detection — Flags after-hours load and communication overload risk
C-SuiteEngineeringHR LeadersKnowledge Workers
40–60%
Time Reclaimed
3 Tiers
Auto-Priority
2–3 lines
Thread Digests
0 noise
Focus Mode
Case Study 02  ·  Workforce Analytics · ML · Predictive Analytics
AI Employee Productivity Dashboard
Real-time insights, intelligent automation, and AI-driven performance analytics that turn raw employee activity data into actionable workforce intelligence — for executives, HR, managers, and individuals.
PythonTensorFlowPyTorchscikit-learnFastAPIPostgreSQLRedisReactJira APISlack API

The Problem

  • Productivity data scattered across Jira, Slack, calendars, and HRIS
  • Managers see lagging KPIs, not real-time bottlenecks or burnout risk
  • No personalisation — same dashboard for executives and individual contributors
  • Skill gaps and learning paths handled manually, if at all

Our Solution

  • Real-time monitoring of activities, task completion, and engagement
  • AI detects bottlenecks, workload imbalances, and stress signals early
  • Sentiment analysis across communication channels flags disengagement
  • Personalised learning suggestions linked to performance and skill gaps

Key Capabilities

  • Productivity Insights — Bottlenecks, imbalances, and efficiency gaps surfaced live
  • Smart Task Management — AI prioritises tasks by urgency, deadlines, and impact
  • Well-being Tracking — Burnout, stress, and disengagement signals
  • Predictive Planning — Workload forecasts, hiring needs, shift planning
  • Learning Paths — Skill-gap detection + Coursera / Udemy / LinkedIn integration
C-SuiteHRTeam LeadsEmployees
Live
KPIs & OKRs
Predictive
Workforce Planning
Sentiment
Detection
Personalised
Coaching
Case Study 03  ·  RegTech · LLM · A2P Messaging Compliance
AI-Powered Website Verification for 10DLC Compliance
An LLM + chatbot system that automatically vets client websites against 60–70 critical 10DLC A2P messaging compliance checks — reducing manual review from hours to under five minutes with 95%+ accuracy.
OpenAI GPT-4LangChainPythonFastAPIPlaywrightBeautifulSoupPostgreSQLReact

The Problem

  • Manual 10DLC compliance review takes hours per website and is error-prone
  • Inconsistent scoring across reviewers delays carrier approval
  • Onboarding bottleneck — clients wait days to begin SMS campaigns
  • No audit trail for why a site passed or failed which criterion

Our Solution

  • AI chatbot evaluates each site against 60–70 compliance questions automatically
  • LLM analyses content, meta-data, opt-in flows, and policy pages in context
  • Each criterion graded pass/fail with reasoning — full transparent report
  • Bulk verification — dozens of client sites processed in parallel

Key Capabilities

  • Automated Q&A — AI checks every site against the full compliance checklist
  • Regulatory Alignment — Opt-in flows, privacy, terms, CTA, intent transparency
  • Scoring Engine — Pass/fail per criterion with reasoning
  • Bulk Processing — Parallel verification across dozens of client sites
  • Audit Report — Carrier-ready transparent compliance document
A2P CarriersSMS AggregatorsMartechCompliance Ops
<5 min
Per Site
95%+
Accuracy
60–70
Checks Automated
Dozens
In Parallel
Case Study 04  ·  Sales AI · Multi-Channel · CRM Integration
AI-Powered Lead Follow-Up System
A virtual sales assistant that remembers every interaction across email, chat, SMS, calls, and meetings — and follows up in a human, context-aware way that scales to thousands of leads without losing personalisation.
OpenAIAnthropic ClaudeLangChainPineconePythonFastAPIPostgreSQLTwilioHubSpot APISalesforce API

The Problem

  • Sales reps manually piece together history across 5+ channels per lead
  • Leads go cold while reps juggle outreach across pipelines
  • Repeated questions and inconsistent messaging hurt trust
  • Humans can only run personalised outreach to a few dozen leads per day

Our Solution

  • AI agent with 360° memory across Email, Chat, SMS, WhatsApp, Calls, Calendar
  • Generates replies grounded in the full prior conversation — no repetition
  • Predicts the optimal follow-up time using engagement scoring
  • Logs every AI interaction back to HubSpot / Salesforce / Zoho / custom CRM

Key Capabilities

  • Multi-Channel Memory — Full 360° view of the lead journey
  • Contextual Replies — Tailored to relationship stage and prior history
  • Smart Scheduling — Engagement-score-driven follow-up timing
  • CRM Sync — HubSpot, Salesforce, Zoho, or custom integrations
  • Parallel Personalisation — Thousands of leads, all feel 1:1
Sales LeadersRevOpsB2B SaaSOutbound Teams
+35%
Conversion Rate
−50%
Sales Cycle
−60%
Manual Effort
0 missed
Follow-Ups
Case Study 05  ·  CRM Automation · LLM · Intent Mapping
AI Chatbot for Go High Level CRM
An AI assistant that lets non-technical staff run Go High Level CRM through natural-language chat — from Slack, WhatsApp, or directly inside GHL — eliminating menu navigation and manual updates entirely.
OpenAINode.jsSpring BootGHL APISlack APIWhatsApp Business APIPostgreSQLRedis

The Problem

  • GHL is powerful but UI-heavy — staff waste time clicking through menus
  • Non-technical team members avoid the CRM, leading to stale data
  • Repetitive tasks (tagging, follow-ups, status updates) clog the day
  • Slack/WhatsApp-first teams resent context-switching back into GHL

Our Solution

  • Chat-driven CRM — "Tag John as VIP" instead of 6 menu clicks
  • LLM maps natural language to GHL API actions in real time
  • Embeds in Slack, MS Teams, WhatsApp, or as a GHL sidebar widget
  • Confirms every action and updates the pipeline transparently

Key Capabilities

  • Contact Management — Add, update, merge, deduplicate via chat
  • Tagging & Scoring — Auto-tag by intent, maintain lead scores
  • Campaign Triggers — Run SMS/email campaigns by command
  • Pipeline Actions — Move leads, update status, fetch by stage
  • Reports On-Demand — "Top 5 leads this week" returns instantly
Marketing AgenciesGHL UsersSMB SalesRevOps
Chat-first
Zero UI
100%
Action Logged
Any Team
No Training
Real-time
Sync
Case Study 06  ·  Computer Vision · Public Safety AI · MLOps
AI-Powered Crowd Detection & Tracking System
A real-time computer-vision system that detects, counts, and tracks crowds across CCTV, drone, and aerial cameras — with adaptive scene intelligence that adjusts models and thresholds automatically per camera angle.
YOLOv8VisDrone-YOLOByteTrackOpenCVPyTorchPythonFastAPIDockerNVIDIA GPU

The Problem

  • Traditional CCTV analytics fail in dense or aerial scenes
  • False positives from split-body detection inflate counts
  • ID switching during occlusion breaks individual tracking
  • Manual tuning required for every new camera — doesn't scale

Our Solution

  • Smart Scene Selector detects camera geometry and density automatically
  • Switches between YOLOv8 and VisDrone-YOLO based on view type
  • ByteTrack maintains stable IDs through occlusion and surges
  • Vertical merge logic eliminates split-body double-counting

Key Capabilities

  • Accurate Counting — Handles dense, overlapping crowds reliably
  • Multi-Camera — CCTV, drone, and aerial feeds in one pipeline
  • Adaptive AI — Auto-tunes model + threshold + tracker per scene
  • Stable Tracking — Consistent IDs through occlusion and movement
  • Real-Time Output — Live count, density, and visualisation overlays
Smart CitiesTransport HubsEventsSurveillance
+30–50%
Count Accuracy
Real-time
Inference
Multi-cam
Compatible
Zero
Manual Tuning
Case Study 07  ·  Legal AI · Semantic Search · Indian Law
LegalDraft-AI · Legal Document Intelligence
An end-to-end AI platform for legal document understanding, semantic case retrieval, and judgment outcome prediction — purpose-built for Indian legal data using InLegalBERT and FAISS vector search.
InLegalBERTPyTorchTransformersFAISSFastAPIStreamlitpdfplumberDockerPostgreSQL

The Problem

  • Keyword search misses semantically similar legal precedents
  • Massive judgment volumes — manual analysis is unscalable
  • Generic NLP models miss Indian legal terminology and structure
  • No predictive insight into likely judgment outcomes

Our Solution

  • InLegalBERT — a transformer trained specifically on Indian legal text
  • FAISS vector search returns top-k semantically similar cases in milliseconds
  • Layout-aware PDF parsing preserves legal context and citations
  • Judgment outcome scoring grounded in historical patterns

Key Capabilities

  • Legal Doc Understanding — Long-form judgments parsed and embedded
  • Semantic Retrieval — Meaning-based, not keyword-based, search
  • Outcome Prediction — Historical patterns inform research
  • Domain-Specific Model — InLegalBERT vs. generic LLMs
  • Interactive UI — Streamlit for lawyers, FastAPI for integrations
Law FirmsLegal ResearchComplianceLegal Tech
−60–70%
Research Time
Semantic
Not Keyword
Millions
Docs Scalable
India-law
Specific Model
Case Study 08  ·  Automation · Browser Orchestration · Multi-Tenancy
Visa Appointment Booking Automation System
An enterprise-grade browser automation platform that monitors visa slots in real time, solves CAPTCHAs, and books appointments across multiple centres in parallel — production-deployed with multi-tenant agent limits.
Node.jsPuppeteerExpressSupabasePostgreSQLPM2Bottleneckp-limitN8N

The Problem

  • Visa slots open and disappear within seconds — humans can't react fast enough
  • Multi-step auth flows with security questions block simple automation
  • CAPTCHAs gate every action; manual solving breaks scale
  • Immigration consultants manage hundreds of clients across centres

Our Solution

  • Puppeteer with stealth plugin avoids bot detection
  • Auto CAPTCHA solving + proxy rotation for reliability
  • Parallel concurrent sessions with rate limiting per agent
  • Webhook alerts to Slack / N8N the moment a slot is booked

Key Capabilities

  • Stealth Automation — Puppeteer + stealth plugin, proxy rotation
  • Multi-Step Auth — Handles security questions and complex login flows
  • CAPTCHA Solving — Integrated auto-resolution
  • Multi-Location — Books across multiple visa centres simultaneously
  • Agent Limits — Multi-tenancy with configurable per-agent quotas
Immigration ConsultantsTravel AgenciesVisa Services
24/7
Slot Monitor
Parallel
Multi-Session
Auto
CAPTCHA
Live
Production
Case Study 09  ·  Logistics SaaS · IoT · Multi-Tenant · White-Label
Fleet Management & Cargo Tracking Platform
A white-label multi-tenant SaaS for transport and logistics companies — combining hardware-based GPS tracking, electronic cargo seals (E-Seals), customs compliance, and native mobile apps for fleet owners and drivers.
Node.jsLaravelReactReact NativePostgreSQLMQTTWebSocketsGoogle MapsAWSDocker

The Problem

  • Fleet operators juggle multiple tools for GPS, cargo, customs, and maintenance
  • Customs cargo monitoring (E-Seals) requires tamper-proof event history
  • Multi-tenant white-label SaaS for transport companies didn't exist off the shelf
  • Mobile apps for drivers and clients need to be fast, native, and branded

Our Solution

  • Multi-tenant SaaS — Super Admin + per-company dashboards with role-based access
  • Hardware GPS protocol support: Teltonika, Concox, Sinotrack, Coban, Queclink
  • E-Seal module — multi-vendor seal registration, tamper alerts, audit trails
  • Native iOS + Android apps with live map, geofence, alerts, and shipment status

Key Capabilities

  • Real-Time GPS — Live map, route replay, speed, geofence, deviation alerts
  • Cargo & Customs — Shipment registration, border checkpoints, compliance reports
  • E-Seal Management — Multi-vendor seals, tamper alerts, tamper-proof event history
  • Maintenance — Scheduling, fault tickets, insurance/license reminders
  • White-Label — Custom logo, colours, domain per client company
Transport & LogisticsCustoms AuthoritiesFleet Owners
Multi-tenant
Architecture
5+ Protocols
Hardware GPS
E-Seal Ready
Customs Compliant
Native
iOS + Android