How TrueLayer improved interaction and problem resolution with Ema

TrueLayer is a fintech pioneer in the UK and EU transforming financial services with open banking APIs. It empowers developers to securely access and leverage financial data, driving innovation in payments, lending, and personal finance. TrueLayer prioritizes transparency, security, and compliance, enabling businesses to create groundbreaking financial solutions while safeguarding customer data.

As a fast-growing FinTech company responsible for managing critical payments for their customers, timely and accurate customer support is paramount to TrueLayer. TrueLayer needed a solution where they could free up their customer support team to focus on the most complex support cases while software automates the rest.

The technical bar was high:

• Deep technical domain: TrueLayer provides clients with a set of financial APIs to connect to payment rails in each country and region. A big challenge was understanding the code and APIs TrueLayer provides — and figuring out which APIs apply to which countries, regions, and accounts.
• Knowing when to hand over: It was critical to figure out when escalation to human agents was needed. For sensitive issues related to fraud, security, and compliance policies, TrueLayer wanted a seamless handover back to their human agents.

Ema's Customer Support Assistant AI Employee seamlessly integrated with TrueLayer's Zendesk system in minutes, using a pre-built integration, then learned TrueLayer's support processes by ingesting their Notion knowledge base.

• Self-bridging technical knowledge: Thanks to her domain understanding features, Ema systematically bridged her technical knowledge gap on her own once provided with resources and feedback — now operating with the same vocabulary and tone a seasoned technical support agent would.
• Trained to abstain and escalate: Ema correctly detects which tickets to hand over to human agents — and proactively assists those agents in resolving escalated tickets faster.
• Rapid feedback loop: Conversational feedback improved accuracy rapidly, with Ema resolving complex queries it hadn't seen before.

Within just a few weeks of training, over 82% of cases could be satisfactorily resolved by Ema. Every week, Ema's Customer Support Assistant AI Employee kept improving its accuracy and efficiency, instilling a level of trust that is undeniable.

• 82%+ of cases resolved by Ema within weeks of deployment
• Human agents refocused on the limited set of high-value, complex cases, including sensitive fraud, security, and compliance issues
• Faster escalations: tickets that do reach human agents arrive with Ema's proactive assistance already attached
• Compounding accuracy: week-over-week improvement in accuracy and efficiency

I've been truly impressed by the evolution of our partnership with Ema. Every week, Ema's Customer Support Assistant AI Employee kept improving its accuracy and efficiency. The rapid progress is a testament to Ema's unique approach to building Gen AI applications, and I am genuinely excited to see how our collaboration will continue to elevate our customer experience.

Elizaveta Shevchenko

Technical Support Lead, TrueLayer