Case Study Case Study
Perspectives

Case Studies

Explore expert perspectives on industry transformation, emerging technologies, and business innovation. Stay ahead with insights from WNS thought leaders.

Customer Stories

Testimonials

ALM Media, LLC

Josh Gazes, Senior Vice President – Operations

British Gas

Jess Johnson, Head of Operational Excellence

Mosaic Insurance

Mitch Blaser, Co-CEO

Mosaic Insurance

Krishnan Ethirajan, COO

Oxford Nanopore Technologies

Jason Hendrey, Senior Director, Global Customer Services

WS Audiology (WSA)

Christof Steube, Director of Finance Excellence

Kiwi.com

Leonard McCullie, Director, Vendor Management

Kiwi.com

Petra Reiter, Vice President, Customer Services

Flight Centre

Aaron Fadelli, Business Leader

Healius Pathology

Alex Cook, Head of Finance Operations

Varo Bank

Breanna Rivers, Partner Performance Manager

Yorkshire Building Society Group (YBS)

Jessica Lockwood, Process Automation Manager

WS Audiology (WSA)

Sharang Patil, Director of Group Finance Excellence

Priya Madan Mohan, VP for Group Accounting & Controlling

United Airlines

Chris Kenny, VP and Controller

GFG Alliance

Phillip Irish, General Manager, Shared Services Delivery, Quality & Governance

Energy Australia

Steve Corden, Outsource Operations Leader

Delaware North

Christopher Lozipone, Senior Vice President and Global Business Services Head

Moneycorp

Nick Haslehurst, Chief Financial & Operating Officer

Prodigy Finance

Nico Barnard, Head of Operations

M&T Bank

Chris Tolomeo, Senior VP & Head of Banking Services

Minerals Technologies Inc. (MTI)

Khem Balkaran, CIO

Church's Chicken

Louis J. Profumo, CFO & EVP

Scaling Document Intelligence with an AI-led, Human-in-the-Loop Model

Read | Apr 22, 2026

AUTHOR(s)

A WNS Perspective

Key Points

  • As Intelligent Document Processing becomes foundational to scaling consumer platforms, real-world data variability continues to challenge pure automation — making accuracy, trust and speed harder to sustain at volume.
  • For a leading rewards platform, WNS architected a hybrid AI and human-in-the-loop model that stabilized operations in real-time while systematically improving underlying document intelligence.
  • The outcome signals a broader shift, from reactive exception handling to self-learning, scale-ready operating models that enhance data integrity, strengthen customer trust and improve efficiency as volumes grow.

The Industry Landscape
Scaling Consumer Rewards in a Data-intensive Economy

Consumer rewards platforms operate at the intersection of Customer Experience (CX), data accuracy and monetization. As digital adoption rises, these platforms must process increasingly larger volumes of consumer-submitted receipts. Intelligent Document Processing (IDP) engines remain foundational to receipt processing, but data variability – driven by multiple formats, missing information and physical damage – creates systemic accuracy challenges that pure automation struggles to resolve.

To stay competitive, leading firms are embracing hybrid models combining AI technologies with Human-in-the-Loop (HITL) verification to deliver precise data attribution and near real-time brand insights, critical to sustaining platform trust and customer loyalty.

The Client Challenge
Sustaining Trust and Insight Quality at High Volumes

As daily document submissions scaled into the millions, the client needed to protect frictionless consumer engagement while maintaining trust across its rewards platform. Although its IDP engine performed well on clean documents, a growing share of real-world submissions were poorly illuminated, damaged, unusually formatted or missing key data fields, leading to misreads of store names, line items and transaction totals.

At scale, these exceptions exposed the limits of a purely automated approach and created the need for a more resilient operating model that could:

Why Debt Behaves Differently Infographic
← Swipe →

The Solution
A Hybrid AI Training and HITL Ecosystem

As a strategic, domain-led digital transformation partner, WNS worked with the client to implement a dual-track operating model that combined human expertise with advanced AI Training capabilities. This solution was purpose-built to deliver immediate operational reliability while strengthening automation capability over time.

Overview of the Dual-Track Operating Model

Track 1

Experience and Trust Layer

24×7 human verification of IDP exceptions
Near-real-time document correction
Quality governance and delivery control
Consumer experience protection
Track 2

Automation Improvement Layer

Structured AI Training feedback and re-training
Error pattern capture and classification
Continuous model learning
Progressive reduction in exception dependency

Key Solution Components

1. Human Oversight

To safeguard CX and service reliability, WNS established a globally distributed, always-on receipt verification operation, focused exclusively on AI-generated exceptions. It was designed to:

Target
Business Impact
Review all IDP-flagged receipts
Prevent point allocation errors
Ensure <24-hour turnaround for exceptions
Protect consumer trust and engagement
Correct eligible documents instantly
Maintain frictionless rewards experience
Achieve high accuracy
Deliver consistent quality at scale

2. AI Feedback and Continuous Training

Rather than limiting the engagement to manual corrections, WNS operationalized every exception as a learning opportunity for the AI models.

Overview of the AI Feedback Loop
Why Debt Behaves Differently Infographic
← Swipe →
Feedback Mechanisms Embedded into Operations
Feedback Element
Purpose
Error capture and categorization
Identified systemic failure modes
Field-level tagging (store name, line items, totals)
Enabled targeted model tuning
Structured data handoff to AI-training teams
Accelerated re-training cycles
Continuous monitoring of flagged rates
Measured automation improvement over time

This closed-loop system was designed to directly strengthen the client’s core automation engine while reducing future reliance on human intervention.

Across finance organizations, core close activities, including journal preparation, reconciliations, task management and reporting, are often managed using spreadsheets, e-mail workflows and fragmented data sources. While functional, these environments lack workflow governance, real-time visibility and audit-ready traceability.

For enterprises operating in real estate and joint venture structures, additional complexities, such as “at-share” calculations, portfolio-level grouping and entity-based close dependencies, further increase manual effort and control risk. As regulatory expectations tighten and transaction volumes grow, finance leaders are focusing on end-to-end financial close automation and integrated close governance platforms.

The Outcomes
A Self-improving Operating Model Built for Scale

The engagement transformed receipt verification from a reactive exception-handling function into a resilient, intelligence-led operating model. By integrating always-on human oversight with structured AI feedback loops, the solution strengthened consumer trust, protected experience quality during demand spikes and ensured service consistency as volumes scaled.

Beyond operational performance, the model materially improved data integrity across the receipt ecosystem. Cleaner, more reliable data enhanced downstream analytics for brand and retail partners and quality-led governance created a structural cost advantage that improved efficiency as automation accuracy matured.

Key Tangible Outcomes:

 

0% progressive reduction in exceptions (from 5 percent to 0.12 percent)

0% accuracy sustained consistently vs. 95 percent target

0% reduction in average handling time per receipt

0% turnaround reliability achieved consistently vs. 80 percent target

24-hour verification maintained regardless of volume spikes