This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our cookie policy>

  • From China to the World: Speakly AI Explores the Path to Agentic Banking with Huawei

    From China to the World: Speakly AI Explores the Path to Agentic Banking with Huawei

    Huawei RONGHAI Financial Partner's Story

Backed by the RONGHAI Program, Huawei and Speakly AI have partnered to design globally replicable AI capabilities for real-world scenarios. This partnership aims to scale up banking AI from isolated applications to synergizing across multiple scenarios.

Banking AI is beginning to move beyond simply answering questions to understanding customers. It is actively following up on business, handling collections, and scaling across regions, showing that a global shift toward agentic banking is underway.

Global Banking Enters the Agentic Stage

As Large Language Models (LLMs) integrate into core banking systems, the role of AI in banks is evolving.

Early AI handled basic queries, Q&A, and analytical support. Then, AI copilots entered frontline operations like customer service and sales, where they helped staff process information, provide recommendations, and boost efficiency. Today, AI agents are stepping directly into business workflows—understanding goals, invoking system capabilities, and executing tasks within predefined rules and permissions.

Banking AI is advancing from AI assistants and copilots to AI agents. It is moving from providing information or recommendations to participating in business collaboration. It now takes on standardized, high-frequency business workflows. Agentic Banking is fast becoming a key direction for intelligent banking. This trend creates new opportunities for Chinese fintech companies as they go global.

Years of digital transformation have given Chinese banks a large number of complex, high-frequency, and large-scale business scenarios, allowing local fintechs to accumulate product and implementation expertise. However, going global requires navigating local regulatory requirements, data environments, business workflows, and customer habits, alongside local customer engagement, project delivery, and long-term operations.

Huawei RONGHAI Digital Finance Global Partner Program serves as a vital bridge connecting these core capabilities. Powered by global ICT/AI infrastructure, global sales and service networks, and local ecosystem resources, Huawei connects technology companies to international markets with essential infrastructure, market, and ecosystem support.

Speakly AI is a key practitioner in this collaboration model. Starting with a contact center project for a top private retail bank in Indonesia, Huawei and Speakly AI expanded their partnership to physical branch services, customer engagement, and intelligent collections management, later bringing these capabilities to the Middle East. Together, they have proven how banking AI can grow from single-scenario applications to multi-scenario use cases across regions.

Starting with a Contact Center Upgrade

As an early mover in LLM development and adoption, Speakly AI has partnered with over 50 top financial institutions in China. Its deep experience in core banking scenarios such as sales, customer service, and collections management lays a strong foundation for delivering value to global markets.

The first stop for Speakly AI and Huawei's cross-border partnership was a leading private retail bank in Indonesia.

Managing tens of millions of accounts and annual interactions, the bank relied on thousands of customer service representatives across phone, web, app, email, and social media.

As service volumes grew, the bank was juggling siloed data across channels, limited coverage of traditional sampling quality inspection, and increasing challenges in service standardization and risk management. For the bank, the goal had changed from "answering faster" to understanding every interaction, resolving issues quicker, and giving managers real-time visibility into service quality, risks, and sales leads.

Combining their strengths, Huawei and Speakly AI launched an Intelligent Contact Center Solution built on Huawei Atlas AI and AICC platforms. Huawei provided the compute infrastructure and contact center platform, while Speakly AI delivered core capabilities like AI copilot and intelligent quality inspection—embedding real-time assistance, customer insight, risk management, and operational analysis into daily workflows.

The AI copilot assists frontline staff by understanding customer intents in real time and offering on-screen knowledge, response recommendations, and compliance reminders. Once the service ends, it automatically logs the interaction, creates customer profiles, and generates follow-up tasks. For managers, the LLM processes total communication data to summarize customer feedback and needs, evaluate the service workflow and team performance, and cross-reference business outcomes to pinpoint key conversion drivers.

Focused on service efficiency, issue resolution, quality management, and risk control, the project improved Average Handling Time (AHT) across voice and digital channels, enhanced First Contact Resolution (FCR), and enabled 100% quality inspection coverage. Meanwhile, AI expanded into customer engagement by understanding customer needs, interests, and purchase intents during routine interactions to explore potential business opportunities.

Building on initial successes, conversational intelligence spread into more banking scenarios.

Expanding into More Customer Touchpoints

Physical Branches: Integrating In-Person Interactions into Business Data Systems

Branches remain vital for customer relationships, but in-person chats historically lacked data tracking, leaving the service process of relationship managers (RMs), customer needs, and sales leads buried in unrecorded conversations.

At a policy bank in Thailand, Huawei and Speakly AI introduced conversational intelligence to physical branches, creating an intelligent branch solution that includes smart badges, LLM conversation insights, and an AI sales assistant called SalesMate.

Smart badges capture in-branch customer interactions. SalesMate helps RMs manage customer information and analyze conversations, connecting with the bank's app and Customer Relationship Management (CRM) system to generate personalized follow-up strategies. Meanwhile, LLM conversation insights extract customer needs, profiles, and sales execution status from high-volume customer communications.

By converging data from online contact centers, physical branches, and other touchpoints, the bank gained a unified view of customer interactions to drive engagement, sales follow-ups, and service improvements.

Customer Engagement: Feeding Omnichannel Data into Operations

Speakly AI unifies voice and text data from branches, contact centers, WhatsApp, and other channels into a single conversational intelligence system. By linking interaction history with customer behavior, it empowers banks with the full context needed to spot changing needs, re-engage dormant customers, and recommend tailored financial services, products, and solutions.

To scale this, Speakly AI introduced a dual-agent model: the RM Agent and the Service Agent.

The RM Agent addresses the challenge of scale—where a typical RM oversees 800 to 3,000 customers but manually reaches just 10% to 15%. By scanning interaction histories and behavioral data, it identifies dormant customers, initiates personalized outreach via WhatsApp, recommends marketing campaigns, and handles lead nurturing and appointment scheduling. When a customer shows strong interest, the RM Agent transfers the customer's information, interaction history, and current needs to the RM for deeper engagement.

The Service Agent primarily handles routine inquiries and standardized services. When encountering high-intent customers, complex issues, or requests beyond its authorization, it hands off the conversation, along with its complete history, to customer service representatives to ensure uninterrupted support.

Conversational intelligence keeps learning from customers, AI agents manage high-volume outreach and standardized services, and human staff focus on high-value, complex needs.

Intelligent Collections Management: AI Agents in High-Frequency Workflows

Collections management is a high-frequency, highly regulated scenario in banking, and requires both strict rule compliance and personalized customer communication.

Speakly AI applies AI agents across the entire 0-to-90-day overdue lifecycle, to tailor collections strategies aligned with distinct customer profiles and business objectives.

At a bank in Southeast Asia with over 2,000 branches, more than 5,000 RMs were responsible for customer engagement, product sales, and collections follow-ups. Facing a high volume of overdue accounts and rising communication demands, the bank needed a better balance between outreach efficiency, follow-up strategies, and compliance management.

The Collection Agent takes on different tasks across the recovery lifecycle: delivering quick responses and gentle reminders in the initial 0–15 days, targeted interventions and strategic follow-ups from 15 to 30 days, and controlled escalations with end-to-end traceability for late-stage accounts between 30 and 90 days.

During actual interactions, the AI agent checks customer profiles, credit ratings, loan types, and debt status within authorized limits to offer tailored repayment plans. It negotiates terms, guides repayments, and manages follow-ups over multi-turn chats.

If a case exceeds authorized limits, hits an anomaly, or dissatisfaction is expressed, the agent summarizes the customer profiles and context and hands over to a human staff.

In practice, for 0–15 day accounts, the first-response time dropped by about 70%, digital outreach grew 2–3x, and promise-to-pay (PTP) rates rose by 15%. For 15–30 day accounts, case preparation time fell by 60%, and average case handling capacity per collections specialist rose by about 15%. For 30–90 day accounts, 100% of interactions became traceable, reducing compliance risks by about 40%. Driven by intelligent multi-turn conversations, personalized negotiation, and guided repayments, the PTP conversion rate jumped from about 3% under conventional bots to 28%.

From quick response and accurate follow-up to compliant escalation, AI agents are integrating into high-frequency banking workflows, handling standardized tasks at scale within clear business rules and authorization boundaries, while transferring complex cases to human staff.

Cross-Market Success: Scaling Proven Solutions to the Middle East

Through multiple projects, Speakly AI modularized its core capabilities—customer service assistance, conversational intelligence, quality management, and customer insights—making them easily configurable for different bank workflows, regulatory requirements, and data environments. Building on this, Huawei and Speakly AI deployed the solutions at a top bank in Saudi Arabia.

The partnership focused on the customer service center, covering voice and digital channels to boost service efficiency and quality, customer experience, and sales conversion. Designed to serve banking, digital payments, and insurance, AI dynamically adapts to scenario-specific product rules, service workflows, and customer needs, retrieving the relevant data, knowledge, and rules. For example, it supports account management, loans, and credit card services in retail banking, manages transaction processing and account security in digital payments, and assists with policy administration, claims, and coverage details across insurance operations.

As interaction data accumulates, AI helps identify service-to-sales opportunities based on customer needs and interests. When a customer expresses interest in a loan, credit card, insurance, or wealth product, AI detects the intent from context and passes the lead and chat history to human staff for follow-up.

As a result, the customer service center transitioned from a reactive support channel into a strategic hub connecting customer services, customer insights, and business operations.

From Southeast Asia to the Middle East, Speakly AI established a model of "modular capabilities + localized adaptation," allowing proven banking AI capabilities to scale quickly across regions in alignment with local business, regulations, and data environments.

Financial-Grade Foundation: Powering Multi-Scenario Deployments at Scale

Extensive deployments across diverse scenarios and regions have turned project wins into comprehensive, financial-grade solution capabilities.

At the solution level, it pre-integrates AI agents, conversational intelligence, process orchestration, knowledge bases, customer context, and system APIs, complete with human handoff, audit trails, and risk boundaries to meet banks' requirements for business applications, security management, and operational control.

At the infrastructure layer, Huawei provides secure and scalable enterprise-level digital infrastructure and platform capabilities, while Speakly AI focuses on banking scenarios, AI agents, and interaction intelligence. They collaborate on solution design, implementation, and continuous optimization.

Software-hardware synergy enables AI integration with existing banking systems and workflows, forming a full support framework that spans from infrastructure and AI capabilities to business applications.

For banks, this means deploying proven solutions with local tweaks—saving time, cutting integration costs, and avoiding building from scratch.

RONGHAI Program: Moving from Isolated Projects to a "4-Win" Model

As an independent software vendor (ISV) partner in the Huawei RONGHAI Program, Speakly AI leverages Huawei's global ecosystem to work alongside local system integrators (SIs) and bank customers, jointly driving the deployment of banking AI solutions across global markets.

Under this ecosystem framework, Huawei provides global sales and service networks and ICT/AI infrastructure platform capabilities. ISVs offer market-validated industry solutions and software capabilities. Local SIs, leveraging local market resources and customer relationships, are responsible for local adaptation, compliance support, and project delivery. Bank customers drive solution implementation based on actual business needs and gain corresponding business value.

For Speakly AI, the RONGHAI Program provides a clear collaboration path that connects global infrastructure, global markets, and local ecosystems. Through this framework, Speakly AI brings its banking AI capabilities, already proven in markets like Southeast Asia, into new regional markets. These capabilities are then adapted to local business processes, regulatory requirements, and data environments, enabling Speakly AI to work with Huawei, local SIs, and bank customers to deliver projects on the ground.

Through a mature global ecosystem, this model allows proven banking AI capabilities to rapidly expand into new markets with localized adaptation and delivery. It also provides a practical foundation for Speakly AI to scale its banking AI scenarios in global markets.

Advancing Global Banking AI Practices for Agentic Banking

From building capabilities in China to scaling across Southeast Asia and the Middle East, Huawei and Speakly AI continue to pioneer the global implementation of banking AI.

Looking ahead, the partnership will continue to advance in core banking scenarios under the RONGHAI Program, driving the adaptation and implementation of proven AI and agent capabilities in more markets.

Moving from isolated applications to multi-scenario synergy, from AI-assisted to agent-involved process execution, and from local wins to global deployment of proven capabilities, banking AI is entering a new stage. Huawei and Speakly AI remain committed to real-world impact as they shape the future of Agentic Banking globally.

TOP