76% support automation for digital bank

Account and product enquiries

Sparkle's customers reach the AI agent with a wide range of day-to-day banking questions – from account management and card requests to transaction limits, payments, savings products, and loyalty programme details. With a growing user base and query volumes rising in step, routing these interactions through human agents was becoming unsustainable.

The AI agent now handles this full range of routine enquiries 24/7 across Sparkle's website, mobile app, and WhatsApp. Customers can check account details and transaction history, get step-by-step guidance on managing their debit and virtual cards, navigate savings and group banking features, and understand their loyalty programme status – all without waiting for a human agent.

The deployment achieved 76% support automation overall – meaning the large majority of customer interactions are resolved end to end without any agent involvement.

Complaint handling and ticketing

Before Proto, every customer complaint raised across Sparkle's digital channels had to be manually logged by an agent in the ticketing system. With around 500 tickets being created each month, this represented a significant and largely avoidable overhead.

Proto integrated the AI agent directly with Sparkle's ticketing system, establishing a structured and automated complaint intake flow. When a customer raises an issue, the agent guides them through a step-by-step intake process, captures the details of the complaint, and creates a ticket automatically – without any agent involvement. When a case requires human handling, the agent escalates it directly, passing the full conversation context so the agent can pick up without asking the customer to repeat themselves.

The result: 90% fewer tickets created manually by live agents – equivalent to approximately 60 hours of agent time saved every month, based on average handling time per ticket.

Authenticated in-app support

The majority of Sparkle's users interact with the AI agent through the mobile app, where they are already logged in and verified. Asking an authenticated customer to re-identify themselves mid-session creates unnecessary friction – the kind that erodes trust in a digital bank built on seamless experience.

Proto addressed this by embedding the AI agent directly into the app via SDK integration. Because the agent operates within the authenticated app environment, it can recognise that the user's identity is already verified and skip redundant identity steps in its workflows. Customers reach the information or resolution they need faster, without interruption.

Multilingual support across five languages

Nigeria's linguistic diversity presents a real barrier for digital financial services. A support experience limited to English excludes a large portion of customers who communicate primarily in Yoruba, Igbo, Hausa, or Pidgin.

Proto built localised variants of the AI agent's workflows for each of Sparkle's five supported languages – English, Pidgin, Yoruba, Igbo, and Hausa. Each variant has its own set of workflow triggers, trained on phrases that reflect how speakers of that language naturally express banking queries. A language identification layer detects the customer's language from their first message and routes the conversation to the appropriate workflow automatically, with no manual selection required.

The result is a support experience that responds natively in the customer's language, across the website, mobile app, and WhatsApp simultaneously.

Security and data handling

As a Central Bank of Nigeria (CBN)-licensed financial institution, Sparkle operates under strict regulatory requirements for how customer data is handled across digital touchpoints.

Proto's deployment was structured accordingly. Customer interactions are handled through defined, auditable workflow logic – every response the AI agent delivers is pre-built and controlled, with no ambiguity in what is said or how data is processed. Personal details captured during complaint intake are routed directly into Sparkle's own ticketing system via a configurable webhook integration, keeping sensitive information within Sparkle's infrastructure rather than held on a third-party platform.