Context
A major international bank was developing a new corporate-banking platform with conversational AI intended as a core capability. The work reframed the proposition from a conventional support chatbot into an assistant capable of combining contextual guidance, banking information and customer actions.
The project defined a focused first experience around payment tracking, a reusable system for AI responses and interface components, and a longer-term vision for proactive orchestration.
The challenge
The organisation was rebuilding its corporate-banking experience to simplify legacy journeys and introduce new capabilities faster. Conversational AI was part of the platform vision, but the initial ask was close to a traditional support chat. The opportunity was to integrate assistance into customers’ work so it could understand context, retrieve relevant banking information and enable action. Any solution also had to meet the expectations of a high-trust, permission-sensitive environment by making behaviour clear, protecting customer control and providing safe recovery or human support when it could not proceed.
Approach
The work began by clarifying the initial brief, technology context and constraints with product and technology stakeholders. A three-stage capability model described how the experience could progress from answering queries, to offering contextual guidance and actions, and eventually to proactive orchestration. This gave the programme a shared day-one scope and longer-term direction.
Research into perceptions of AI in financial services was combined with a review of conversational products across banking and adjacent sectors. The findings were translated into requirements spanning interaction, human handover, task enablement, trust and compliance, and error recovery. The first use case, tracking a payment, was then mapped across customer intent, information gathering, data retrieval, permissions, response and next actions. Entry-point options were evaluated before a persistent assistant was selected for the initial proposition. Design principles, a structured response architecture and a reusable conversational component system turned the strategy into an actionable experience framework.
Solution
The resulting concept was a persistent AI assistant embedded within the corporate-banking interface. It combined natural-language conversation with structured UI, enabling customers to ask about a payment, see its status and supporting facts, understand what had happened and take an appropriate next step without leaving the journey.
A reusable library gave the assistant consistent patterns for progress, balances, transactions, tables, selections, forms, documents, cards and error recovery. Response logic accounted for successful, partial, ambiguous, restricted and failed scenarios. A separate future-state prototype showed how the same foundation could evolve to identify a need, assemble a plan and invite the customer to review, edit and approve it before execution.
