Bid Intelligence Workbench
Public-Procurement Bid Intelligence for Healthcare Distribution
Co-developed with a Philippine healthcare distributor’s team trained on BMAD. We built a governed pipeline from PHILGEPS evidence through Neo4j graph intelligence — with a Deep Agent workbench that keeps humans in the approval loop.
Bid Windows Move Faster Than Manual Review
Public procurement in the Philippines publishes invitations to bid and supporting packs across PHILGEPS and related channels. Healthcare distributors must triage large, uneven document sets — often scanned PDFs — match line items to catalog SKUs, and decide where to compete, under tight deadlines.
The client needed more than a chatbot: a production-minded system that captures evidence, structures content, extracts line items, matches SKUs, and projects relationships into a graph — with enterprise SSO and clear human oversight for agent-driven steps.
BMAD + Graph + Deep Agents
ITQ co-developed the solution alongside the client team, who were trained on the BMAD method so delivery stayed collaborative and accountable. The stack combines BMAD delivery discipline with Neo4j for graph projection, LangGraph for governed orchestration, and Deep Agents for the Bid Intelligence Workbench.
The core evidence-to-graph pipeline is in place. A Deep Agent workbench lets agents propose and run analysis while keeping humans in the loop for consequential steps — with UX polish and operational hardening still ongoing alongside the client team.
What the System Delivers
End-to-end bid intelligence — from public evidence to approved agent workflows.
PHILGEPS / ITB Evidence Capture
Systematic capture of invitation-to-bid and related procurement evidence from public sources, so bid teams work from a complete, current record.
OCR → Markdown
Document conversion pipelines turn scanned and PDF bid packs into structured Markdown suitable for downstream extraction and review.
LLM Line-Item Extraction
Large language models extract line items from ITBs and related documents with human-visible outputs ready for validation.
SKU Matching
Extracted line items are matched against the distributor catalog to surface coverage, gaps, and competitive positioning.
Neo4j Graph Projection
Bids, products, suppliers, and evidence links are projected into a knowledge graph for queryable bid intelligence.
Deep Agent Workbench
A Bid Intelligence Workbench where Deep Agents propose analysis steps under human approval — operators stay in control of consequential actions.
Entra SSO
Enterprise identity via Microsoft Entra SSO, so access aligns with the client’s existing security and directory policies.
How We Built It
BMAD
Structured delivery method with client-trained team
Neo4j
Graph projection of bids, SKUs & evidence
LangGraph
Governed agent orchestration
Deep Agents
Bid Intelligence Workbench with human approval
Where Things Stand
Core pipeline in place; Deep Agent workbench advancing with the client team.
Core pipeline delivered: evidence capture, OCR→Markdown, line-item extraction, SKU matching, and Neo4j graph projection.
Agents can propose and run analysis with human approval for consequential steps. UX polish and operational hardening continue with the client team.
Building Bid Intelligence for Your Team?
Let's discuss governed agentic systems for procurement, distribution, and complex document workflows.
Start a Conversation