ABN AMRO
A better customer experience with Voice AI
Summary
A state-of-the-art voice AI system that handles millions of customer service calls annually, with natural conversation, autonomous actions, and responsible AI guardrails.
DEUS partnered with ABN AMRO to create a natural-sounding, AI-powered customer service experience. The voice assistant handles customer identification, intent recognition, and autonomous actions, from card blocking to limit adjustments, while ensuring responsible AI performance through real-time guardrails.
Industry
Services
The Challenge
Millions of routine calls consuming human capacity, with no scalable path forward under the existing model.
ABN AMRO's contact center handled enormous call volumes, the vast majority of which were structured, repeatable queries that required no human judgment. Balance checks, recent transactions, card blocking, limit adjustments. Each one answered by a human agent, but multiplied by millions of calls, the cost and capacity burden was immense. The bank needed a system that could enable better customer experiences for all credit and debit holders while ensuring responsible performance with guardrails in real time.
Problem 01
Unsustainable call volumes
- ▪Millions of inbound calls per year
- ▪Majority are structured, tier-1 queries
- ▪Headcount growth not viable at this volume
- ▪Increasing cost-to-serve year on year
Problem 02
Bank-grade accuracy and responsibility
- ▪Financial data requires zero-error tolerance
- ▪Authentication and identity verification in-call
- ▪Regulatory compliance across all interactions
- ▪Responsible AI guardrails needed in real time
The Solution
We built the AI-based customer service automation on three pillars: a modular, model-agnostic backend; a natural conversation layer with sub-second latency; and a three-level guardrail system.
01
Customer identification and secure verification
The voice assistant identifies the customer through natural conversation, verifying their identity securely before proceeding, removing friction while maintaining bank-grade security.
02
Intent recognition and autonomous actions
Using natural language understanding, the system determines what the customer needs, retrieves live account data, and autonomously carries out actions, from providing balance information to blocking cards.
03
Intelligent human handoff
When complexity exceeds the AI's scope, the system seamlessly routes to a human agent, providing them with full context so the customer never has to repeat themselves.
Client
quickly and easily solve your problem
Voice Assistant
Human Agent
filter out repetitive tasks and help you focus on helping the client
Impact
Millions of calls handled autonomously, freeing human agents to focus on the interactions where human empathy and judgment genuinely matter.
Autonomous Resolution
The vast majority of incoming calls are routed through the AI, handling customer identification, intent recognition, and actions autonomously at full production scale.
Natural Conversation
Significant reduction in average handling time per call, with sub-second latency creating a natural conversational experience that feels effortless rather than robotic.
Responsible AI
Three-layer guardrail system ensures responsible AI performance without impacting the customer experience: system-level, programmatic rules, and utterance-level filtering.
Methods
A modular voice AI engineering approach built for natural conversation, autonomous actions, and responsible performance.
Voice AI
Audio LLM
Natural language understanding
Intent recognition
Customer identification
Secure verification
Autonomous actions
Human handoff
System guardrails
Programmatic rules
Utterance filtering
Observability
Modular backend
API integration
Solution architecture
Engineering