AI Call Summarisation Assistance tool at Tier 1 Global Bank

Across the financial services industry, firms collectively spend billions annually on call center operations, with a significant portion of these costs stemming from inefficiencies in handling customer interactions. Studies show that call centers spend up to 30% of their time on administrative tasks, costing firms millions in lost productivity and suboptimal customer engagement.

THE CHALLENGE

A Tier 1 global bank faced significant challenges in its fraud call center operations. Agents were spending excessive time manually note-taking during and after calls, diverting their attention from resolving actual customer issues. This inefficiency negatively impacted customer experience and the agent’s ability to take swift actions to deliver better customer outcomes.

The bank sought a partner to identify ways to enhance the efficiency and effectiveness of call center operations. Capco, in collaboration with the bank’s platform leads and business stakeholders ran a structured four-stage process over 12 weeks, which identified six high-impact opportunities for action, projected to deliver £22 million in savings across risk and compliance operations.

One of the initiatives was the implementation of a Call Summarisation Assistance tool, designed to transform how agents handle fraud-related calls. This innovative solution leverages advanced transcription and summarisation technology to transcribe live conversations and extract critical information, including fraud details, customer concerns, and agreed actions.

The tool automatically compiles this data into structured, easy-to-read summaries, integrates directly into workflows, enabling agents to focus on resolving customer issues rather than administrative tasks.

OUR APPROACH

To address this challenge, we brought together a multidisciplinary team of data scientists, business case specialists and experts in Fraud case handling. The team combined extensive industry experience with technical expertise in data wrangling, feature engineering, process mapping, and machine learning. Leveraging tools such as Python, alongside deep regulatory knowledge in Fraud, the team was uniquely positioned to deliver impactful and innovative solutions.
The team delivered the Call Summarisation Assistance solution through the following steps:

• Process analysis: Examined call center workflows to identify inefficiencies, focusing on how manual note-taking impacts agent productivity and how auto-call summary functionality can address these challenges.
• Impact measurement: Quantification of the impact of inefficiencies in fraud call handling on operational costs and compliance risks and projecting the cost savings from proposed solutions.
• Data consolidation and preparation: Defined technical and data requirements, ensuring compatibility with internal systems. Analysed the process for identifying customer accounts at the call outset to enable seamless integration with transcription and summarisation tools.
• Data enrichment: Aggregated and enriched call data with critical variables such as call duration, operator details, timestamps, and follow-up actions to enable accurate transcription and summarization.
• Automation analysis: Identified automation opportunities, including auto-transcription and summarisation, to streamline workflows and reduce manual workloads.
• Live transcription and processing: Developed and integrated transcription capabilities, ensuring voice files extracted from internal system are processed accurately and efficiently.
• Call summarization: Designed and implemented automated summarisation models to produce concise notes in a standardised format, making case reviews easier and more consistent for call handlers.
• Workflow integration: Enabled seamless integration with existing systems to ensure summaries and original transcriptions are attached automatically to customer files or case management tools. Ensured operators could validate and update summaries post-call to improve AI model accuracy.
• Impact monitoring and feedback loop: Established metrics to track efficiency gains, such as time savings and improved call-handling capacity. Built a feedback loop for operators to validate AI-generated summaries, continuously training the model for higher accuracy and effectiveness.

VALUE DELIVERED

The solution proved to be transformative, as highlighted by a senior stakeholder at the client, who described it as ‘a significant breakthrough’ in addressing operational inefficiencies and improving fraud call handling. The measurable business benefits included:

• Improved efficiency: Time spent on post-call administrative tasks was reduced by 50%, enabling agents to handle more calls, prioritize customer interactions, and respond swiftly to urgent issues.
• Enhanced quality: Automation ensured that call summaries adhered to a consistent format, eliminating discrepancies and minimizing human error, improving operational follow-ups.
• Cost savings: Delivered significant operational savings of approximately £8m annually and free-up agent capacity to handle more customer calls.

The solution’s broader applicability extends beyond fraud call handling. It can be deployed across operational areas such as collections, claims handling and customer support, where it can accurately capture payment plans, customer commitments, and critical details, ensuring smoother follow-ups and better recovery outcomes.

Today, improved efficiency, consistent documentation, and streamlined processes have empowered the bank to enhance compliance with regulatory requirements and strengthen its ability to manage risks and deliver better customer outcomes.

The client praised the solution as ‘invaluable’ and commended Capco for its ‘brilliant execution.’