AI for Bid Generation in B2B Marketplace

Introduction to Joe: AI Agent for Bid Creation in B2B Marketplace

Joe is an artificial intelligence agent designed to assist suppliers in a B2B marketplace (PDA), optimizing the generation of bids in response to buyer quotation requests. Its primary goal is to accelerate the response process through automated analyses, ensuring precision and alignment with supplier policies.

Key Features:

Request Monitoring:

Using a cronjob, Joe detects requests in the PENDING state, authorizes them, categorizes them, and changes their status to OFFERING, notifying relevant suppliers.

Triple-Match Analysis:

Categories: Verifies if requested items match the categories specified by the supplier in provider_category.

Territory: Validates that the buyer’s location falls within the supplier’s configured regions (country_register, city_id).

Semantics & Documents: Uses NLP to analyze descriptions and OCR/computer vision to extract information from attached images/certificates, seeking matches with the supplier’s catalog (tienda_fi_productos).

Inventory & Pricing Management:

Converts units of measurement (e.g., liters to gallons) and adjusts quantities based on available stock.

Applies pricing policies, discounts, and delivery costs defined in company_policies, considering geographic location (geo_latitude, geo_longitude).

Bid Creation:

Generates entries in global_bit (header) and individual_bit (items), attaches relevant documents (documentos_varios), and calculates final prices.

Operates in two modes:

Automated Seller: Directly submits bids and notifies the buyer.

Assistant: Prepares drafts in temporary tables for human review before submission.

Continuous Learning:

Analyzes purchase orders (global_purchasing_order) to understand why a bid succeeded or failed.

Adjusts strategies based on human corrections and historical data.

Security & Integration:

Ensures confidentiality between suppliers through data isolation and JWT authentication.

Integrates via secure API (HTTPS, rate limiting) and logs actions in AL_response for auditing.

Technologies Involved:

NLP, OCR, computer vision.

Relational database management (MySQL).

Event-driven architecture or cronjobs for triggers.

This introduction summarizes Joe’s scope, highlighting its ability to combine automation, advanced analysis, and adaptability. For full technical details (table structures, specific workflows, or implementation policies), refer to the attached document.