@vs_labs WhatsApp
Book a consultation
Home›Cases
⟩ 23 PROJECTS · REAL NUMBERS

Our cases

No "improved customer experience" without numbers. We show what we built, how it works, and what actually moved in the business — from UAE fintech to fashion-tech startups.

CASE 01 · FINTECH · AML / KYC

AML / KYC automation for UAE fintech

Client: UAE-licensed fintech

What it does
  • Screens every new client and every transaction against sanctions lists (OFAC, UN, EU, UAE Central Bank), PEP, adverse media
  • Builds client risk profile and operation risk score
  • Maintains KYC dossiers with auto-reminders for re-certification
  • Drafts STR/SAR for the compliance officer
  • Prepares filings for goAML (FIU UAE)
Results
  • 95% clients clear KYC automatically
  • ~5 min KYC time vs 30–60 manual
  • 4× client growth without compliance hires
PythonFastAPIAWS TextractComplyAdvantagegoAMLPostgreSQL
AML / KYC automation for UAE fintech
CASE 02 · REAL ESTATE · PROPERTIX & AI AGENT

Deep real-estate analytics + AI Real Estate Agent

Client: real-estate investment firm (UAE, Europe)

What it does
  • Parses 10+ data sources (classifieds, developers, government feeds, brokers)
  • Cleaning, dedup, enrichment (coordinates, infra, transit, POIs)
  • Computes 50+ metrics: price trends, liquidity, occupancy, yield
  • ML models for price prediction and undervalued-asset detection
  • Investor dashboards and alerts on matching inventory
  • AI Real Estate Agent — qualifies buyers/renters in chat 24/7, asks budget/area/timeline, suggests matching listings, books viewings
  • Auto-syncs new matches with each client and pings the broker only when the deal is hot
Results
  • hours vs 2–3 weeks for district analytics
  • 30 min data refresh interval
  • + alpha client sees deals first
ScrapyFastAPIClickHouseXGBoostKubernetesMetabase
Deep real-estate analytics + AI Real Estate Agent
CASE 03 · MARKETING · CREATIVE FORGE

Marketing automation — AI marketer

Ad-account analytics, campaign management, creative pipeline

What it does
  • Ad-account analytics (Meta, Google, Yandex)
  • Auto-management of campaigns, budgets, bids
  • Creative pipeline at scale (Creative Forge)
  • A/B testing, kills underperformers, audience analysis
Results
  • −15–25% cost per lead
  • 10–20 hrs back per week for marketer
Meta APIGoogle AdsStable DiffusionFLUXOpenAIn8n
Marketing automation — AI marketer
CASE 04 · ANALYTICS · CRM

AI CRM / ERP analyst

Stop digging through dashboards — just ask AI

What it does
  • Analyzes CRM, ERP, and other sources, surfaces dropped leads
  • Audits client handling, chats, calls, Zooms
  • Answers business questions in seconds — voice or text
Results
  • +10–25% funnel conversion
  • seconds per ad-hoc report
HubSpotSalesforcePipedriveamoCRMWhisperOpenAIClickHouse
AI CRM / ERP analyst
CASE 05 · SALES · AI TEAM

AI sales team — voice + chat

AI rep that handles Telegram and WhatsApp 24/7 + voice AI on phones

What it does
  • Handles Telegram and WhatsApp 24/7
  • Voice AI on inbound and outbound calls
  • Qualifies leads, books Zoom meetings, sends links
  • Writes call summaries to CRM, moves leads through the pipeline
Results
  • 10× faster lead handling
  • +20–40% lead → meeting conversion
  • 1–2 FTE saved on routine
OpenAI RealtimeWhisperTwilioTelegramWhatsAppHubSpot
AI sales team — voice + chat
CASE 06 · B2C · PERSONAL AI

Personal AI agent 24/7

Personal assistant in Telegram / WhatsApp that takes over your routine

What it does
  • Telegram and WhatsApp 24/7 (voice + text)
  • Manages email, calendar, meetings, contacts
  • Years of memory across people, projects, topics
  • Joins Zoom/Meet — writes minutes, captures decisions
  • Posts content, drafts messages, prepares pre-meeting briefs
  • Spawns and orchestrates sub-agents for complex tasks
Results
  • 15–20 hrs saved per week
  • 0 subscriptions to us
  • you own the server and the agent
LangGraphGPT-5Claude15+ toolsVPS Hetzner
Personal AI agent 24/7
CASE 07 · ENTERPRISE · ON-PREMISE LLM

AI on-premise — model on the client's server

“Worried about data leaks? AI on your local server. Not a byte goes outside.”

What it does
  • Open-source models (Llama, Qwen, Mistral, DeepSeek) deployed on client hardware
  • All processing inside the corporate perimeter — docs, mail, DBs never leave
  • API integration with internal systems (1C, AD, mail, CRM)
  • Fine-tuning on the client's domain data
Results
  • 0 bytes leave the perimeter
  • 100% regulator compliance (DFSA, HIPAA, PDPL)
LlamaQwenMistralvLLMOllamaDockerKubernetes
AI on-premise — model on the client's server
CASE 08 · SMM · CONTENT FACTORY

SMM automation

Content factory: text, images, carousels, video — months ahead

What it does
  • Auto-generates content per playbook, plan, taxonomy
  • Multiple formats: text, images, carousels, video
  • Edit, approve, schedule months ahead
  • Posts to Telegram, Instagram, YouTube, LinkedIn
Results
  • 4–5 platforms instead of 1–2
  • 3 months of content in one day
OpenAIStable DiffusionSoraInstagramYouTubeLinkedInn8n
SMM automation
CASE 09 · ANALYTICS · DASHBOARDS

Dashboards + AI analytics

Full digital visibility: every system in one window with AI recommendations

What it does
  • Connects to CRM, ERP, messengers, email, Slack
  • Pulls data from every system the business runs on
  • Real-time KPIs, role-based views
  • AI analytics: instant recommendations, no time spent reading numbers
Results
  • 1 window instead of 10 systems
  • real-time metrics across all stacks
ClickHouseMetabaseApache SupersetAirbyten8nOpenAI
Dashboards + AI analytics
CASE 10 · REAL ESTATE · RENTAL OPS

Real-estate / short-term rental management platform

Multi-platform booking management + smart locks + AI concierge

What it does
  • Booking management across Airbnb, Booking, etc.
  • Smart-lock control
  • AI concierge for check-in / check-out
  • Cleaning and maintenance automation
  • Dynamic pricing based on demand and competitors
Results
  • 5–10× more units per manager
  • +15–25% occupancy via dynamic pricing
Channel ManagerSmart LockTelegramWhatsAppOpenAIStripe
Real-estate / short-term rental management platform
CASE 11 · FINANCE · AI ACCOUNTANT

AI accountant

Captures source documents, posts to the accounting system, prepares filings

What it does
  • Captures source docs (statements, invoices, acts, bills) from email, Telegram, EDI, WhatsApp
  • Validates fields, reconciles amounts: contract → invoice → act → payment
  • Posts entries to the accounting system, runs vendor reconciliations
  • Prepares accounting reports and tax filings
Results
  • minutes per posting vs hours
  • zero lost documents
  • auto tax filings prepared
OCR1C APIEDI APIsTelegramWhatsAppOpenAIPostgreSQL
AI accountant
CASE 12 · OPS · DOCUMENT FLOW

AI document workflow

Full automation of inbound document handling

What it does
  • Collects documents from email or Telegram channels
  • Sorts and files into the right folders
  • Extracts key fields (OCR + NLP)
  • Routes for approval, tracks deadlines
Results
  • zero lost documents
  • massive time saved on routine
OCROpenAIGoogle DriveTelegramn8n
AI document workflow
CASE 13 · GROWTH · OUTREACH

Cold outreach engine

Competitor parsing, personalization, sending, reply tracking

What it does
  • Parses competitors, collects emails, stores in DB
  • Generates personalized pitches per recipient
  • Sends sequences, tracks delivery and replies
  • All replies routed into the CRM
Results
  • hundreds of personalized emails per day
  • auto domain warmup, bounce / spam control
ApolloHunterInstantlySmartleadOpenAIPostgreSQL
Cold outreach engine
CASE 14 · LEGAL · AI LAWYER

AI lawyer

Reads contracts, checks against the corporate playbook, summarizes risk

What it does
  • Reads inbound contracts and proposals
  • Checks compliance with the corporate playbook
  • Drafts summaries with risk flags 🔴🟡🟢
  • Maintains a registry with reminders for renewals and deadlines
  • Generates contracts from templates and instructions
Results
  • minutes per contract review vs hours
  • 0 missed renewals
OpenAIClaudeOCRE-signaturePostgreSQLFastAPI
AI lawyer
CASE 15 · OPEN-SOURCE · CLAW BOT

Claw Bot setup

Open-source equivalent of the personal AI agent — on your infrastructure

What it does
  • Self-hosted assistant with full personal-agent capabilities
  • Telegram and WhatsApp 24/7
  • Years of memory across people, projects, topics
  • Plug in any tools and skills
  • Spawns and orchestrates sub-agents
Results
  • 100% client owns everything
  • 0 vendor lock-in
ClawLangGraphLocal modelsVPS / on-premiseTelegramWhatsApp
Claw Bot setup
CASE 17 · CONTENT · VIDEO PIPELINE

Content factory

Auto-generates scripts + videos + descriptions and posts to your socials

What it does
  • Auto-generates scripts and produces video
  • Writes descriptions and titles
  • Posts to your socials on schedule
  • Per-platform performance analytics
Results
  • daily video with no daily work
  • focus team on strategy, not ops
OpenAIElevenLabsHeyGenSoraFFmpegYouTubeTikTok
Content factory
CASE 18 · SALES · VOICE AI

AI call center — voice agent

Handles inbound and runs outbound across your phone base

What it does
  • Answers inbound calls
  • Runs outbound (just feed in a phone list)
  • Talks per playbook and knowledge base
  • Writes summaries to the CRM, moves leads through the pipeline
Results
  • 0 missed calls
  • orders of magnitude lower cost per call
  • 100% CRM logging coverage
OpenAI RealtimeElevenLabsTwilioSIPWhisperHubSpot
AI call center — voice agent
CASE 19 · ANALYTICS · CALLS & ZOOM

Call & Zoom analytics

100% of customer conversations under a microscope

What it does
  • Transcribes every call and Zoom
  • Analyzes by your rules — script compliance, escalations, etc.
  • Detects fraud (atypical promises, gray-area schemes)
  • Flags reference-grade calls for training
Results
  • 100% of calls reviewed (vs 2–5% manual)
  • +10–25% sales conversion
  • 50× lower QA cost
WhisperOpenAIClaudeZoomGoogle MeetClickHouse
Call & Zoom analytics
CASE 20 · SERVICE · CUSTOMER SUPPORT

AI Customer Support with RAG

Listens across all channels and answers from your knowledge base

What it does
  • Listens to inbound from email, WhatsApp, Telegram, Instagram
  • Answers from your knowledge base (RAG) and playbook
  • Drives users toward the target action — e.g. Zoom with leadership
  • Books meetings and reserves calendar slots
Results
  • 0 human-hours on first-line support
  • seconds average response time
OpenAIpgvectorPineconeLangChainTelegramWhatsAppCalendly
AI Customer Support with RAG
CASE 21 · B2B · MARKETPLACE

Industrial Market — AI categorization module

Client: B2B marketplace for industrial goods

What it does
  • Suppliers upload goods with their own naming — doesn't match marketplace taxonomy
  • Solution: dedicated microservice, gRPC + Kafka queue
  • BERT-family ML/NLP model with continuous fine-tuning on feedback
Results
  • 92% auto-categorization with 98%+ accuracy
  • 12× less manual work
  • minutes from upload to live catalog
PythonPyTorchTransformersgRPCKafkaKubernetes
Industrial Market — AI categorization module
CASE 22 · STARTUP · AI APP

Custom AI app — multi-model chat with agents

Client: startup, B2C AI product

What it does
  • Mode selector (quality / speed / creative / analysis) → different models
  • Library of pre-built AI agents
  • Image generation, analysis, editing
  • Voice (ASR + TTS), text, images
  • Semantic search over chat history
Results
  • on schedule shipped to prod
  • thousands of concurrent users, stable
  • next round of funding secured
Next.jsNestJSpgvectorRedisOpenAI / Claude / GeminiStripe
Custom AI app — multi-model chat with agents
CASE 23 · FASHION-TECH

Find Look — photo-to-product search

Client: fashion-tech startup

What it does
  • Computer vision: segmentation, normalization, feature extraction
  • ML search: embeddings + NN over 10M+ SKUs
  • Classification: category, brand, color, material with confidence
  • 20+ stores, dedup, card unification, cross-store price comparison
Results
  • 90%+ recognition accuracy
  • <3s photo → results
PythonPyTorchCLIP-familyFAISSClickHouseReact Native
Find Look — photo-to-product search
⟩ LET'S TALK

Your problem is more complex
(or simpler)? Even better.

Every project is unique. Tell us what you're dealing with — we'll sketch how we'd solve it, roughly what it costs, and the timeline.

OR
CHAT WITH US ON ⟩ TELEGRAM ⟩ WHATSAPP ⟩ EMAIL