Skip to content

Case studies

The End of Routing Chaos at Equifax

How Equifax replaced a legacy IVR with a voice agent that identifies the reason for the call before it routes it anywhere.


Voicing.ai Team5 min read

Contents
A telephone switchboard beside a tangle of unplugged cords, with one straight sage cord plugged in, drawn in fine ink linesFinance
The document01 / 02

The Identity Crisis in Financial Contact Centers

Calling a major credit bureau is rarely a casual activity. When a consumer picks up the phone to contact a financial data institution they are usually in the middle of a highly stressful life event. They might be trying to secure a mortgage approval that is suddenly stalled. They might have just received an alert about potential identity theft. They might be trying to place a hard freeze on their credit profile after losing their wallet. The emotional stakes are incredibly high and the tolerance for bad customer service is absolutely zero.

Despite the urgency of these situations the historical standard for routing these calls has been the traditional Interactive Voice Response system. We have all experienced this exact nightmare. You call a toll free number and are immediately greeted by a robotic voice reading a long nested menu tree. You are told to press one for credit reports, press two for disputes, press three for identity theft, and press four to repeat the options.

For Equifax this legacy phone menu was creating a critical operational bottleneck. The contact centre carried inefficiencies that degraded the customer experience from the first second of the call. The legacy system struggled with accurate intent detection and routinely sent calls to the wrong department. Customers navigated a maze of nested options only to be transferred several times before reaching an agent who could actually help them.

Beyond the routing chaos there was a massive authentication bottleneck. Financial institutions must adhere to strict security protocols. In the old system human agents were forced to spend the first several minutes of every single call performing manual verification processes. This wasted incredibly expensive tier one agent time on routine identity checks. Furthermore the legacy system was essentially a black box. Management had absolutely zero visibility into routing failures, intent patterns, or opportunities to contain queries through automation. Equifax needed a radical intervention to modernize their front door.

The Implementation Journey of the Precision Welcome Bot

Equifax partnered with Voicing AI to tear down the legacy nested menu tree and replace it with an intent driven voice automation platform. The goal was to build a Precision Welcome Bot that could understand natural human language, authenticate the caller securely, and execute dynamic routing based on the exact nature of the problem.

The implementation was broken into a short, structured timeline. Phase one covered the first two weeks and focused on discovery and architecture. The Voicing AI team read the existing call flows and built the intent taxonomy from them, then mapped the authentication requirement for every intent type. The conversational flow they designed has three steps: identify the intent, authenticate, then resolve or route.

Phase two spanned weeks three to five and tackled the hardest part of any enterprise deployment, which is secure integration. The voice platform was integrated into the Equifax enterprise stack: an API orchestration layer for real-time access to customer accounts, credit reports and dispute status, and a direct connection into the system of record for customer context, interaction history and case management. This phase also implemented compliant authentication workflows, using voice verification and knowledge-based questions, to hold the deployment inside its regulatory obligations.

A menu asks the caller to guess which department owns their problem. Most callers guess wrong, and the transfer is the cost.

Phase three took place over weeks six and seven and focused entirely on Natural Language Understanding training and routing logic. The team fine tuned the intent classification models specifically on Equifax terminology. For example the AI had to perfectly understand the nuanced difference between a customer asking for a credit freeze versus a fraud alert versus a dispute initiation. They built a dynamic routing engine equipped with a large intent taxonomy utilizing skill based agent matching and queue optimization algorithms.

Phase four was the final week of testing and optimisation. The system ran in shadow mode against live production traffic, answering silently while the human floor stayed in charge. The team validated authentication accuracy against human baselines and tested routing strategies against each other. Observability dashboards for intent confidence, authentication success and containment were in place before anyone flipped the switch to live production.

Architecting for the Speed a Conversation Needs

Voice AI Pipeline Speed

The voice pipeline operates at low latency. As a customer speaks, the audio flows through speech recognition, into language understanding, and into the intent classifier fast enough that the reply lands inside the window a caller reads as a conversation. That is what keeps the exchange feeling natural, with no awkward robotic pause.

Multi Modal Authentication Engine

The Authentication Engine operates on a multi modal verification framework. It cross references voice biometrics with knowledge based authentication and account data, running silent fraud detection flags in the background. If a caller sounds suspicious or fails a background check the system dynamically alters the routing path to a specialized fraud team.

Routing Intelligence and Observatory Platform

The Routing Intelligence module is equally sophisticated. It executes skill based matching utilizing the complexity of the intent, the specific value of the customer, the required agent expertise, and the current queue depth across the contact center network.

All of this data feeds directly into the Observatory Platform. For the first time ever Equifax management gained access to real time dashboards displaying intent classification confidence scores per interaction. They could track authentication success and failure rates separated by specific verification methods. They gained granular tracking into routing accuracy showing the exact percentage of calls reaching the correct agent on the very first try. The black box IVR was finally dead.

The Operational Impact and the End of the Transfer Loop

The before and after comparison of the customer journey is profound. Before Voicing AI a customer had to navigate a rigid nested menu tree with no natural language understanding. They were forced to wait until after they were finally routed to authenticate themselves, wasting valuable agent time.

After Voicing AI the experience is one turn. The agent opens with a plain question: what brings you to Equifax today? It identifies the intent in natural language, then says it needs to verify identity and asks for the details that establish it. It resolves the query itself when the query is a balance check or a score enquiry. When a human is needed it routes the call to a specialised agent with the full context already loaded into the system of record. The customer never repeats themselves.

The operational picture changed in four places at once. Intent accuracy rose far enough that the agent establishes the reason for the call in the opening exchange rather than through a menu. Secure verification succeeds often enough that routine enquiries complete without any human involvement. First-contact routing accuracy rose against its baseline, which collapsed the transfer loop and cut the average number of transfers on a call to a fraction of what it was. The path from greeting to resolution or routing shortened, and a material share of tier one enquiries now complete autonomously. Voicing AI will publish the measured values for this deployment when Equifax has signed them.

The Future Roadmap of Proactive Finance

Equifax has a foundation to build on and a roadmap that is deliberately ambitious. Next is predictive routing: the history of a caller sends a repeat call about an open dispute straight to the specialists who have been handling it.

After that the work turns from reactive to proactive. The agent calls out about suspicious activity and credit monitoring alerts before the customer has noticed a problem. Further out, the platform takes on native Spanish for the Latin American customer base, and self healing workflows, where the system finds its own recurring misclassifications and deploys the corrected routing logic without a human in the loop.

Equifax has proven that security and exceptional customer experience do not have to be mutually exclusive.

Read nextCase study

How Telmex scaled outbound collections on voice

4 min readRead the case study

Closing02 / 02

Bring one call type. Leave with an architecture.

An airport service desk at sunrise: a traveller with a suitcase at the counter, and an agent in a headset answering behind it.
Voicing

Voice infrastructure on the contact centre floor

A working session with an engineer who has deployed inside a bank’s perimeter. We map your telephony, data boundary and handoff rules, and tell you what we would not automate.