Spot2 – IA Automations

AI-Powered Automation

In a marketplace like Spot2, a large part of critical tasks were operated manually: lead acquisition, requirement profiling, space uploading, data validation, and coordination between parties. This generated operational bottlenecks, missed opportunities, and a high commercial workload.

To scale operations without multiplying human resources, I led the construction of an AI-based infrastructure capable of automating key tasks and creating a more intelligent, fast, and efficient system.

Strategic Objective

To design and implement an AI-based automation infrastructure, natively integrated into Spot2’s business flow, with the aim of:

Accelerating high-impact repetitive processes
Standardizing data collection and quality
Decoupling commercial tasks from manual effort
Preparing Spot2 for a distributed, always-on operation based on triggers and data

Components of the Automation System

Each module was designed and developed with AI tools + backend logic to function autonomously or integrated:

1. Lead Profiling (AI Chatbot on WhatsApp)

Natural conversation that collects sector, zone, budget, size, and comments
Automatically creates a qualified project and assigns it to an agent
Designed with LangChain + OpenAI and deployed via Twilio API

2. Intelligent Matching between Spaces and Requirements

Recommendation algorithm created in collaboration with Data
Uses embeddings, location vectors, and sector categorization
Automates the loading of suggested spaces into the project, without the need for manual search

3. Automated Visit Scheduling

Intermediate bot that contacts the owner via WhatsApp to validate availability
Coordinates and schedules visits directly between tenant and owner
Reduces human friction and improves confirmation rate

4. Automatic Space Upload (Brokers)

Reception of PDFs, Excels, or text messages with spaces
Automatic parsing to detect property name, address, size, price, and upload it to the system
Validates if the space already exists and updates or creates a new one

5. Intelligent Space Updating

Cron job + AI that detects spaces without movement in 30+ days
Automatically contacts brokers to validate availability and conditions
If the space is still active, marks it as updated. If not, puts it in review

Success Metrics (AI Infrastructure)

North Star Metric

% of core operations performed by the system without human intervention

Operational Metrics:

Initial response time to lead: < 2 min
% of automatically created projects: > 65%
Confirmation of visits with owner without manual intervention: > 50%
Space upload via AI/parsing vs. manual: +35%
Decrease in outdated spaces: -40%

AI Planning

Architecture

Orchestration with LangChain for agents that process language and trigger actions
Supabase for persistence of states and triggers per client, space, or session
Asynchronous workflows (cron jobs + trigger functions) to verify spaces without movement
Modular system to add new agents without breaking existing logic

Internal Operational Interface

Administrative panel in Next.js for:
- Viewing bot conversation history
- Monitoring status of spaces pending verification
- Editing prompts and conversational flows
- Visualizing which tasks were executed automatically

Results (to date)

+70% of conversations that arrive at Spot2 via WhatsApp are automatically converted into structured projects
Reduction of operational load on the sales team by >40%
Increased precision in space-client matching with AI vector logic
Decrease in time between incoming lead and informed agent: from hours to minutes

Personal Role

I designed the vision, architecture, and modular implementation of the system
I programmed agents in LangChain for language processing, data extraction, and triggers
I integrated OpenAI + WhatsApp via Twilio as a universal interface for clients and brokers
I validated with internal teams and commercial agents to calibrate the output and improve system accuracy
I orchestrated the progressive rollout of each module, measuring impact and refining flows