Executive Summary
Always-current procurement data, without APIs
The client's procurement team depended on dozens of supplier portals that offered no API access, forcing manual copy-paste audits that were slow, inconsistent, and prone to missing critical expiry changes. I led the architecture, build, and governance program for a procurement intelligence pipeline that replaced that manual process with resilient, scheduled automation — giving the business always-current, auditable supplier data for the first time.
The Business Problem
The client's inventory and quality teams depended on product and expiry data locked inside dozens of supplier web portals with no API access. The only way to keep that data current was a manual copy-paste audit process, run inconsistently across suppliers and prone to human error. The business risk was not just the effort involved — it was missed expiry changes that led to inventory write-offs and quality incidents, and an audit trail too weak to satisfy internal controls.
A baseline audit at engagement kickoff quantified the exposure: procurement staff spent an estimated 22 hours per week on manual portal checks across the supplier base, missed or late-caught expiry changes drove an estimated $180K in annual inventory write-offs, and there was no consistent, timestamped record of when supplier data had last been verified. Leadership needed a way to keep supplier data current at scale without depending on APIs the suppliers would never build.
The Solution
I designed and delivered a governed procurement intelligence pipeline that navigates supplier portals through resilient browser automation, extracts product and expiry data through a combination of parsing and OCR, and normalizes everything into a consistent schema. Scheduled runs keep the data current automatically, and alerts fire the moment an item nears expiry or a supplier record changes — routed directly into the inventory and quality workflows the team already uses.
Every extraction run is logged with a timestamp and a reliability check that flags anomalies — a portal layout change, a failed extraction, an unexpected data gap — rather than silently propagating bad data downstream. That reliability discipline, not just the automation itself, was what made procurement and quality leadership trust the pipeline enough to retire the manual audit process entirely.
How It Works
Automation
Browser automation for dynamic portals with resilient navigation strategies that adapt to layout changes.
Extraction
HTML parsing plus OCR and text routines for consistent capture across inconsistent supplier layouts.
Outputs
Normalized CSV/XLSX outputs and expiry alerts designed to drop directly into existing procurement tools.
Architecture & Governance
The diagram below traces how a supplier portal's raw data becomes an alert-ready, normalized record, with an automated reliability check flagging anomalies before they ever reach inventory or quality teams.
System Architecture — Portal-to-Alert Flow
Five stages · one automated reliability checkpoint
Resilient browser automation across dozens of no-API supplier portals, on a fixed schedule
HTML parsing combined with OCR and text routines for consistent capture across every layout
Raw extracted content mapped into a consistent, queryable schema across every supplier
Layout changes, failed extractions, and data gaps are flagged rather than silently propagated
Normalized CSV/XLSX and expiry alerts delivered directly into inventory and quality workflows
Audit & run log runs beneath every stage — every extraction run is timestamped and traceable, closing the audit gap the manual process could never satisfy.
- Browser automation layer for complex navigation and dynamic content across dozens of distinct portal layouts.
- Extraction logic combining parsing, OCR, and field normalization into one consistent output schema.
- Concurrency controls to scale across suppliers in parallel while respecting each portal's rate limits.
- Scheduled runs and alert generation for items nearing expiry or records that have changed since the last check.
Business Impact & ROI
- Cut missed or late-caught expiry incidents by 81% by moving from periodic manual audits to continuously scheduled checks.
- Eliminated 90% of the 22 weekly hours procurement staff previously spent on manual portal copy-paste work.
- Reduced estimated annual inventory write-offs tied to missed expiry changes from $180K toward a small fraction of that exposure.
- Delivered a fully costed ROI model presented to the executive committee: total program investment recovered in approximately 5 weeks of production use, driven by labor savings and reduced write-offs, with a 5.6x return in year one.
- Gave quality and compliance teams a timestamped, auditable record of every supplier data check for the first time.
- Created a scalable extraction framework the client has since extended to additional supplier categories without incremental engineering effort.
My Role
I owned this engagement end-to-end as the client's AI/ML leader, operating at CTO scope: I built the business case and ROI model presented to the executive committee, set the reliability standards the pipeline had to meet before the manual process could be retired, and designed the extraction and normalization architecture that made supplier data trustworthy at scale.
I also aligned outputs and alerts directly to the inventory and quality workflows stakeholders already used, rather than asking them to adopt a new tool, and led the phased rollout across the supplier base that proved reliability before full retirement of the manual process. The result was an operating capability the client's leadership could trust and audit, not a fragile scraping script.