Industrias Qatar
The salesperson that never sleeps
For Industrias Qatar, a Colombian manufacturer of tow hitches and bike racks, we built a WhatsApp sales assistant that answers every customer in seconds, at any hour. It understands natural Colombian Spanish, voice notes and photos, searches a visual catalog organized by vehicle, sends the exact product photo back, captures leads and books installations — and hands the chat to a human when it matters.
Duration
Ongoing · since March 2026
Started
March 11, 2026
Text, voice notes and customer photos, with Colombian holidays and business hours built in
Reply generation, lead extraction and four independent escalation detectors
Documents, catalog images, PDFs and videos on every inbound message
Mar 11 to Apr 14, 2026, with continuous improvements since
Industrias Qatar makes and installs tow hitches and bike racks, with a catalog where almost every vehicle make and model has its own product, and two locations in Bogotá and Chía.
Sales happened on WhatsApp, by hand: for every enquiry an advisor had to identify the customer’s vehicle, find the right reference, send the photo, quote, handle objections and book the installation — one message at a time, during office hours.
- — Volume and repetition: answering “does it fit my 2020 Hilux?” with the right photo and price is repetitive but demands precision
- — After hours: messages arrive at any time, and every late reply is a lead that cools down or goes to a competitor
- — Consistency: the sales pitch — lifetime warranty, Colombian manufacturing, installation included — depended on each advisor
- — Lost information: vehicle, product, value and appointment stayed buried in the chat and never fed a sales pipeline
Five systems delivered as one operating platform at ai.
industriasqatar.com, live five weeks after kickoff and improving continuously since.
- + WhatsApp AI agent · answers 24/7 in natural Colombian Spanish, understands text, transcribed voice notes and the photos customers send, and respects business hours and Colombian holidays
- + Visual RAG knowledge base · documents and a catalog of product images embedded in pgvector; the bot sends the exact photo, PDF or video for the customer’s vehicle automatically
- + Automatic lead extraction · name, vehicle, product, estimated value, location and appointment flow into a four-stage pipeline that only moves forward, with email alerts and appointment confirmations
- + Human escalation through four independent signals · frustration or an explicit request, the model’s own flag, operational promises the bot must not make, and payment proofs — sent to the team by WhatsApp and email
- + Operations panel · conversations in a WhatsApp Web–style view, lead board, self-service training with a phone simulator, analytics, and editable personality, prompt, hours and branding
Inside the build.
Brand, interface and system — shown as they shipped.
One message, sixteen steps.
This is not a website with a chatbot. A single WhatsApp webhook orchestrates a 16-step pipeline on every inbound message — deduplication, transcription, vision, four parallel vector searches, generation, media resolution, logging, analytics and four escalation detectors. Roughly 60% of the work lives in the backend.
A knowledge base that does not expire
Files uploaded to a hosted model API expire after 48 hours. We moved to our own embeddings in pgvector, searchable indefinitely and refreshed whenever the catalog changes.
The bot sends the photo, not a promise
The model answers with media tags tied to real catalog IDs; the pipeline resolves them into WhatsApp images, PDFs and videos, and strips any tag that points to an ID that does not exist.
No promises it cannot keep
A dedicated detector catches any installation or delivery date the bot should not commit to and alerts the team, so the AI handles volume and people handle commitments.
Built for serverless and Meta retries
A full turn can take 15–30 seconds. Extended execution time and a three-state deduplication make sure Meta’s retries never produce a duplicate reply, or no reply at all.
Never silent
If the model hits a rate limit, a safety filter or a timeout, the customer gets a polite fallback that is stored and auditable. A monthly token ceiling cuts off gracefully.
Hybrid AI Coding
Senior WebMinds engineering owns the architecture and every decision with the client, while AI agents (Claude Code) implement with QA on every change. That is how the platform went live five weeks after kickoff.
Technologies & Tools
The tech stack powering this solution
Beyond the numbers, we delivered a system that transformed how this client operates.
Sales without office hours
The assistant answers, quotes, shows the exact product and books installations around the clock in natural Spanish — including nights, weekends and Colombian holidays.
A complex catalog, mastered
A vehicle-by-vehicle catalog lives in a permanent vector knowledge base, and the bot sends the right photo of the hitch or rack for the customer’s car — work that used to mean an advisor searching by hand.
Nothing lost, humans when it counts
Every conversation is logged, leads land in a sales pipeline automatically, and four AI detectors bring in a person for frustration, explicit requests, payment proofs or sensitive commitments.
- Mar 11, 2026Kickoff: technical spec and first presentation
- Mar 17, 2026RAG with pgvector, lead extraction, email and branding
- Mar 27, 2026Visual catalog, image replies on WhatsApp and the phone simulator
- Apr 14, 2026Live operation, signed delivery
- Apr–May 2026Human escalation, WhatsApp profile, web widget, Click-to-WhatsApp campaign attribution and QR links
- OngoingContinuous improvements and catalog expansion
Ready to achieve similar results?
Let's discuss how we can transform your business with a tailored digital solution.