A fast-growing digital-first bank was facing an increasingly familiar problem: success was starting to create operational drag.
As its customer base expanded, so did the pressure on its contact centre, with routine queries steadily eating into agent time. The risk was obvious. Without intervention, servicing would scale linearly with growth, requiring headcount to rise in step with customers. That was neither sustainable nor aligned to the bank’s operating model.
At the same time, the organisation had a reputation to protect and extend. Innovation wasn’t a side quest, it was part of the brand. The bank wanted a conversational channel that could do more than respond with scripted answers. It needed to feel genuinely intelligent, reduce friction for customers, and support the business as it scaled.
Objectives
- Scale service without scaling headcount by offloading routine interactions to automation and reserving human agents for complex, high-value work
- Create a first-to-market experience that felt meaningfully different, not another “me too” chatbot
- Later, objectives expanded to include hyper-personalisation to further drive customer engagement and offer extended client servicing
Benefits
- Service scalability without linear headcount growth
- Hyper-personalised customer interactions at scale
- A governed, trusted conversational service channel
- Clear visibility into call drivers and service friction
- Reduced operational and compliance risk
- A future-ready foundation for intelligent automation
The opportunity
The rise of large language models accelerated internal expectations. Customers were increasingly comfortable with conversational interfaces, and the bar for “intelligent support” shifted almost overnight. The bank’s leadership wanted to move quickly, but without compromising security, governance, or customer trust.
Approach
BBD’s involvement began earlier than most delivery engagements: as an advisory partner.
The bank believed it needed to partner with a globally recognised AI platform provider to bring its vision to life. As a trusted advisory partner, BBD supported the selection process by helping evaluating vendors against the bank’s needs, ecosystem complexity, and the realities of enterprise delivery.
Once a provider was chosen, BBD shifted into implementation mode where we integrated the vendor’s platform into the bank’s environment and connected it to key systems and channels. It was at this stage however where the challenges in partnering with the global AI platform provider surfaced.
The proof of concept worked, but the overall programme struggled to deliver at the pace, quality and collaboration model the bank needed. Progress slowed. Frustration rose alongside inflexibility. And the experience, while functional, wasn’t living up to the promise that had driven the investment.
Rather than abandon the vision, the bank made a strategic pivot to step away from a vendor-led model and build a governed platform that could evolve on its own terms.
BBD proposed a pragmatic next step: a rapid set of proof points to validate what was possible, using a small team to prototype key capabilities, establish the foundations for LLM readiness, and demonstrate a path forward that balanced innovation with control.
In parallel, the bank’s internal machine learning team entered the picture. They had strong data and ML infrastructure capabilities, and a clear engagement mandate. Together, we aligned on a model that played to each team’s strengths:
- The bank’s internal ML team owned AI infrastructure, data pipelines and model iteration
- BBD built the platform around it: conversational orchestration, deterministic journeys, systems integration, and the live-agent experience
It was the right split for a high-stakes, non-deterministic channel as the AI couldn’t exist in a vacuum. It needed guardrails, workflow, integration, auditability, and a controlled route to human support.
Overview of the solution
Over approximately a year, the joint team built a scalable conversational platform designed to support both deterministic and AI-driven interactions through a single customer experience.
At a high level, the platform and surrounding ecosystem BBD developed and integrated included:
- A secure custom message ingestion and routing layer, enabling conversational interactions via an external channel while maintaining governance and control
- A custom conversation orchestration layer responsible for session management, context handling, and message interpretation
- A decision engine to determine whether an interaction should follow a known, deterministic journey or be routed through AI-driven interpretation
- A task and journey engine to design and run structured workflows for common service needs (for example, steps that require specific validations and controlled API calls)
- Live-agent escalation, enabling customers to transition seamlessly to a human agent when needed without breaking the conversational flow
- End-to-end auditing and tracking, ensuring interactions were logged and traceable, supporting risk, compliance, and continuous improvement
The result was not “a chatbot”. It was a managed platform capable of answering, routing, automating, escalating, and learning.
The bank launched the solution in phases: a closed pilot followed by a public release. Early uptake was deliberately measured. Introducing an AI-enabled conversational channel into a regulated environment carries real risk, and trust is earned in increments, not demanded on day one.
The benefits have been meaningful across several dimensions:
- Improved customer engagement through hyper-personalised support
Customers began asking questions specific to their own accounts and context, unlocking more relevant, higher-value interactions. As data quality and model performance improved over time, the experience became increasingly consistent and useful - A first-to-market innovation milestone
The launch reinforced the bank’s position as a digital innovator, delivering a conversational service channel that moved beyond scripted automation - Deflection insights that drive continuous optimisation
The channel became a listening post. By analysing what customers ask, where they struggle, and why they escalate to agents, the bank could prioritise the deterministic journeys that would have the biggest impact on call drivers. Over time, this creates a flywheel: identify friction, build a journey, reduce calls, refine the experience
Contact centre reduction wasn’t immediate, and that was expected. Early in the rollout, customers often escalated to agents because trust was still forming, while agents had to support both voice and chat. But with ongoing training, iterative improvements, and a growing set of deterministic journeys, the bank is steadily increasing containment and reducing unnecessary transfers.
In short: the platform is doing what it was designed to do, not by chasing overnight transformation, but by building confidence and capability over time.
Why it matters
Technology highlights
Designed for scalability, resilience and control, the platform combines modern cloud-native architecture with enterprise-grade integration patterns.
- Cloud-native runtime in Microsoft Azure
The conversational services run in Azure, designed to scale up or down based on demand - Containerised services and orchestration
Core components are deployed as containerised services, supporting independent scaling and clear separation of responsibilities across the platform - Asynchronous, message-based architecture
Interactions are processed through an event-driven, asynchronous model, reducing coupling between components and improving resilience under load - Conversation orchestration framework
A dedicated orchestration layer manages sessions, context, routing and response handling, enabling consistent customer experiences across both deterministic and AI-driven interactions - Deterministic workflow and task engine
Common banking journeys are implemented as controlled workflows, making it possible to deliver predictable outcomes where governance is essential, while still presenting everything through one conversational interface - Enterprise integration across service and support systems
The platform integrates into existing operational systems, including contact centre tooling and customer support workflows, enabling seamless escalation to live agents and full visibility of conversation history - Auditability and risk controls built in
Logging, tracking and verification steps support safe operation, enabling the bank to iterate responsibly while maintaining trust in a regulated context
Impact of BBD’s partnership
BBD helped the bank move from a stalled, vendor-led initiative to a platform it could scale and evolve with confidence, combining enterprise-grade integration and orchestration with the bank’s own AI and data capability.
By building the “everything around the AI” layer, BBD enabled a conversational experience that is auditable, extensible, and designed for real-world banking journeys, not just demos.
