Swisscom saved time and budget: how AI agents optimized customer support and sales

Swisscom, one of Switzerland's largest telecommunications providers, faced the challenge of scaling AI solutions in a constantly evolving ecosystem and under stringent regulatory requirements. The implementation of AI agents for customer support and sales not only allowed the company to optimize operational processes but also significantly reduced the time for developing new AI initiatives, achieving initial stakeholder demos in just 3-4 weeks.

In a large telecom company with billions in revenue and thousands of customers, seemingly minor issues constantly arise: router resets, tariff changes, billing questions. Each such task involves manual labor from an employee who must navigate dozens of systems, gather information, and provide it to the customer. Multiply this by millions of inquiries, and you get colossal time and money expenditures that could otherwise be directed towards development. Today, there's a solution that automates these processes while maintaining security and compliance with all standards.

The reality of the problem: the "automation ceiling"

Swisscom, as a leader in innovation and sustainable development, constantly strives for excellence. However, like many large enterprises, the company encountered what it termed the "automation ceiling." Traditional automation approaches could no longer meet growing business demands, especially in sensitive areas like customer support and sales. The core problem was that a customer inquiry, for instance, about a non-functioning internet connection, could stem from three potential causes: billing issues, a network outage, or an incorrect router configuration. Each of these causes typically falls under different departments, necessitating cross-departmental coordination and access to disparate systems. This consumed significant employee time and slowed down issue resolution.

The path to AI agents: from isolated solutions to a unified architecture

Previously, Swisscom had already utilized various AI solutions, including conversational AI based on Rasa and fine-tuned large language models. However, these solutions were fragmented and did not scale well to address complex tasks requiring interaction across multiple systems and departments. The company realized that to overcome the "automation ceiling," a fundamentally new approach was needed – a unified architecture for creating and managing AI agents capable of seamless communication among themselves.

This led to the idea of developing AI agents that could not just perform individual tasks, but also coordinate their actions, exchange information, and work as a cohesive team to resolve complex customer inquiries from start to finish, all while adhering to Switzerland's strict data protection standards.

How the multi-agent architecture was designed

Swisscom developed a multi-agent architecture focused on addressing the key challenges of scaling AI solutions. The core principles included:

• Seamless agent communication. To achieve this, Model Context Protocol (MCP) and Agent2Agent (A2A) protocols were developed, enabling agents to exchange information and coordinate actions across different domains and departments.
• Security and authentication. Under strict data protection laws, a robust authentication and authorization system was required to ensure least-privilege access. Every request necessitated the use of temporary access tokens, validated by both the agent and the customer context.
• Integration and compatibility. Agents needed to integrate easily with existing Swisscom systems, such as the central API system SAIL (Service and Interface Library) and the corporate network via AWS Direct Connect.
• Customer data collection and utilization. For long-term improvement of the customer experience, it was essential to effectively collect and analyze information from all agent interactions.

Within this architecture, a customer-facing AI agent, deployed in a containerized environment, could interact with other agents and MCP servers, which also ran as containerized applications within Swisscom's shared Virtual Private Cloud (VPC). This ensured flexibility, scalability, and security.

Implementation and overcoming challenges

To realize its ambitious multi-agent architecture, Swisscom faced a number of critical challenges, particularly concerning security, integration, and ensuring common standards. The company decided to use a specialized platform for AI agents, which allowed them to:

• Ensure secure and transitive authentication. Built-in integration mechanisms with Swisscom's existing identity provider allowed for both inbound and outbound authentication management, eliminating the need for custom token exchange servers.
• Manage agent memory. The system allowed for storing both session-based and long-term memory, which is critical for understanding customer context across multiple interactions. This also ensured user data segregation, supporting security and compliance.
• Optimize deployment and scaling. The agent runtime environment enabled developers to focus on building agents, while the system automatically handled secure, cost-efficient hosting and scaling through Docker container deployment.

Implementation began with two B2C scenarios: generating personalized sales pitches and providing automated customer support for technical issues (e.g., router troubleshooting). These agents were integrated into Swisscom's existing SAM chatbot system, which required high-performance agent-to-agent communication protocols due to the high volume of customers and strict latency requirements. Development teams achieved initial demonstrations for stakeholders in just 3-4 weeks, even without prior experience with the chosen framework.

Results and benefits

The use of AI agents brought significant benefits to Swisscom:

• Accelerated development. Development teams were able to achieve initial demonstrations for business stakeholders in just 3-4 weeks, significantly reducing time-to-market for new solutions.
• Reduced complexity. Some project teams migrated from other implementations to the chosen framework, citing reduced complexity and faster development cycles.
• Improved security and compliance. Built-in security and identity management mechanisms ensured strict adherence to Swiss data protection laws, which is critical for the financial sector.
• Scalability and performance. The agent runtime environment allowed for efficient handling of thousands of requests per month with low latency while optimizing costs.
• Long-term customer insights. The ability to track and analyze customer interactions across various channels allows for continuous improvement of the customer experience.

How to implement this in your company

Swisscom's experience demonstrates that successful AI agent implementation requires not only innovative technology but also a comprehensive approach to infrastructure and operations. If your company faces similar challenges, here's where to start:

• Build a robust architectural foundation. Address the fundamental challenges of secure cross-organizational authentication, standardized agent orchestration, and comprehensive observability. This will accelerate deployment rather than constrain it.
• Choose the right framework. Development tools that simplify agent creation and integrate with existing infrastructure can significantly reduce time-to-value.
• Turn regulatory compliance into an advantage. Develop solutions that not only adhere to regulations but also leverage them as a basis for innovation, ensuring data sovereignty and user privacy.

If this case sounds like what's happening in your company, our manager can help: he'll analyze your business and niche for free and point out where an AI agent would bring a real result in your case. Message the manager