How One E-Commerce Co Cut Processing Cost 84%

A PE-owned e-commerce ops team piped unstructured shopping lists into a multi-model pipeline on OpenAI, Gemini, and Claude — collapsing turnaround from 24 hours to 30 seconds.

84%

Cut in annual processing costs

4–8 weeks

Implementation Time

Chris Taylor

CEO & Co-Founder

Fractional AI

Connect ↓

The Challenge

A PE-owned e-commerce company was entirely dependent on a third-party BPO to perform a core back-office operation: matching inbound shopping lists — arriving as images, PDFs, spreadsheets, and other unstructured formats — to its internal standard product taxonomy. The offshore team had accumulated all the institutional knowledge for this process, creating deep vendor lock-in and fragility. Turnaround time exceeded 24 hours per batch, degrading the customer experience. Without change, the business would remain hostage to an expensive, slow, and opaque external dependency with no path to operational ownership or cost reduction.

What They Built

Fractional AI designed and deployed a custom generative AI pipeline that automated the end-to-end shopping list processing workflow — from ingesting unstructured inputs to outputting items mapped against the company’s standard product taxonomy. The system used OpenAI models for the majority of processing, Gemini for text extraction, and Claude for select intermediate steps. Critically, evaluation infrastructure was established early to objectively measure accuracy and guide iteration — and it proved the AI system was outperforming the BPO before go-live. A confidence-scoring layer flags low-certainty outputs for human review, and a feedback loop saves corrections back to the database, allowing the system to improve automatically over time. The unexpected outcome: the company discovered for the first time how inaccurate the BPO’s manual work had actually been.

Fractional AI began by establishing evaluation infrastructure before building the solution — a deliberate choice that allowed the team to benchmark accuracy objectively and demonstrate the AI's performance against the BPO baseline before go-live. The pipeline was built on AWS to ingest the full range of input formats the client received: shopping lists arriving as images, PDFs, and spreadsheets with no standardized structure. Google Gemini handled text extraction across these diverse formats; OpenAI models performed the core taxonomy classification work; Claude handled select intermediate processing steps. Each model was chosen for the specific subtask it performs most reliably. A confidence-scoring layer was built on top to flag outputs below a certainty threshold for human review, ensuring the system could scale without unacceptable error rates. A feedback loop saved human corrections back to the database, allowing the system to improve its accuracy automatically over time. The engagement ran over 2–4 months. A parallel QA layer was retained at the client's discretion. The unexpected finding from the accuracy benchmarking: the BPO had been performing worse than assumed — a fact that had been invisible without a measurement system in place.

AI Role

OpenAI models handle the primary shopping list processing — classifying unstructured inputs (images, PDFs, spreadsheets) and mapping them to the company's internal product taxonomy. Gemini handles text extraction, and Claude performs select intermediate processing steps. A confidence-scoring layer flags low-certainty outputs for human review, and a feedback loop saves corrections back to the database so the system improves automatically over time.

AI Model

OpenAI API

Infrastructure

AWS (hosting and orchestration environment) • Client's proprietary internal product taxonomy database

Integration Points

OpenAI API — primary taxonomy classification and product mapping • Google Gemini API — text extraction from unstructured inputs (images, PDFs, spreadsheets) • Claude API — intermediate processing steps • Confidence-scoring layer routing flagged outputs to human review queue • Feedback loop writing human corrections back to taxonomy database for continuous improvement

Impact

84% Cost Reduction in Year One

The AI system replaced the BPO as the primary processing engine, with roughly two-thirds of remaining cost attributed to an optional QA layer the client chose to maintain. That layer is expected to be phased out in year two, pushing savings toward near-100%.

24+ Hours to ~30 Seconds

What previously required routing batches to an offshore team for 24-plus hour turnaround now completes in approximately 30 seconds per list, enabling real-time operational use cases that were previously impossible.

AI Accuracy Exceeded Human Baseline

By deploying evaluation infrastructure early and running blind side-by-side comparisons, Fractional AI demonstrated the new system outperformed the legacy manual process on accuracy — giving leadership the confidence to decommission the BPO.

Technology Utilized

Implementation Complexity

Cost Reduction

Time Savings

Risk & Compliance

Headcount Avoidance

The pipeline required integrating multiple AI models (OpenAI, Gemini, Claude) on AWS infrastructure with custom orchestration, evaluation infrastructure for accuracy benchmarking, confidence scoring, and a feedback loop mechanism for continuous improvement. Ingesting diverse unstructured input formats — images, PDFs, spreadsheets — and mapping them reliably to a proprietary taxonomy at scale is a complex engineering challenge.

Best Fit For

PE-owned companies with a significant BPO or offshore cost center performing repetitive, high-volume document processing or data mapping work — particularly operators who want to reduce a large recurring cost line and bring critical IP back in-house within a single fiscal year.

Chris Taylor

CEO & Co-Founder

Fractional AI

CEO & Co-Founder of Fractional AI, helping PE firms and portfolio companies implement AI workflow automations, product features, and diligence at scale.

Connect with the Expert

Industry:

Retail & E-Commerce

Business Function:

Operations

Supply Chain & Procurement

Company Size:

251-1,000

Project Cost:

Cost not disclosed

Ownership:

Private Equity-Backed

Organization Type:

Private Company

AI Pattern:

Document Processing & Extraction

Process Automation (RPA + AI)

Value Type:

Cost Reduction

Time Savings

Risk & Compliance

Headcount Avoidance

AI Model:

Frequently Asked Questions

How did a mid-sized e-commerce company cut processing cost 84% with document-processing AI?

The experts stood up evaluation infrastructure first, so they could benchmark accuracy against the existing offshore baseline before go-live. A pipeline on AWS ingests shopping lists arriving as images, PDFs, and spreadsheets, with different models handling extraction, taxonomy classification, and intermediate steps, plus a confidence-scoring layer that flags low-certainty outputs for review. The AI replaced the BPO as the primary engine and cut processing cost 84% in year one.

What AI tools and models were used for the processing pipeline?

The pipeline was built on AWS and used Google Gemini for text extraction, OpenAI models for core taxonomy classification, and Claude for select intermediate steps, with each model chosen for the subtask it handles most reliably. The approach combined document processing and extraction with process automation.

What results did the e-commerce company achieve?

Processing cost fell 84% in year one, per-list turnaround dropped from 24-plus hours to about 30 seconds, and accuracy benchmarking showed the AI exceeded the human baseline, revealing the BPO had been less accurate than assumed.

How long did the engagement take?

The engagement ran about 2–4 months.

Who is this document-processing AI approach best for?

PE-owned companies with a significant BPO or offshore cost center doing repetitive, high-volume document processing or data mapping, particularly operators who want to cut a large recurring cost line and bring critical IP back in-house within a single fiscal year.