How Swisscom is Revolutionizing Customer Support with AI

They are not just focused on business goals—they have also committed to achieving net-zero greenhouse gas emissions by 2035.

The telecom giant has hit what industry experts call the "automation ceiling." Traditional automation tools can only take customer service so far. You need something more intelligent, more adaptive. That's where Amazon Bedrock AgentCore comes in.

Swisscom is transforming customer support and sales with enterprise agentic AI solutions powered by Amazon Bedrock AgentCore. These solutions go beyond basic chatbots and offer advanced AI agents that can collaborate across different departments, uphold security measures, and provide tailored customer experiences on a large scale.

In this article, you'll learn how Swisscom uses Amazon Bedrock AgentCore to create enterprise agentic AI for customer support and sales. We'll explore the architectural choices that enable AI-driven sales operations while ensuring compliance with Swiss regulations. We'll also discuss the difficulties of implementing multi-agent systems in large organizations and the strategies Swisscom has developed to overcome these challenges.

The insights shared here offer a guide for any company seeking to use AI agents in their customer-facing operations.

Swisscom's AI Strategy and Commitment to Sustainability

Swisscom is a prime example of how enterprise technology and environmental responsibility can go hand in hand. For three years in a row, World Finance magazine has named them the Most Sustainable Company in Telecom. This Swiss telecommunications giant has set an ambitious goal: achieving net-zero greenhouse gas emissions by 2035. This commitment goes beyond what the Paris Climate Agreement requires, making Swisscom a leader in sustainable business practices.

The company's sustainability achievements aren't just empty promises—they are the backbone of their digital transformation strategy. With $19 billion in revenue and a $37 billion market cap as of 2025, Swisscom proves that being profitable and sustainable can go together.

Bridging Sustainability and AI Innovation

Swisscom's AI strategy directly addresses what they call the "automation ceiling"—the point where traditional automation approaches reach their limits. By using Amazon Bedrock AgentCore, they're pushing past these constraints while still keeping their sustainability commitments. The integration of AI technologies within their operations serves two purposes: reducing operational inefficiencies that contribute to carbon emissions while also improving service delivery.

The company's approach to responsible AI technologies reflects their broader sustainability framework. You can see this in their decision to deploy solutions in the AWS Europe Region (Zurich), ensuring data sovereignty while minimizing the environmental impact of data transfer. Their containerized deployment strategies optimize resource utilization, reducing computational waste and energy consumption.

Key sustainability-aligned AI initiatives include:

• Energy-efficient containerized runtimes that maximize resource utilization
• Localized data processing reducing cross-border data transfer emissions
• Automated systems that minimize human travel for routine maintenance tasks
• Intelligent routing systems that optimize network efficiency

This strategic alignment between sustainability goals and AI implementation creates a sustainable telecom company model that other enterprises can study and adapt for their own digital transformation journeys.

Architectural Foundations of Swisscom's Agentic AI Framework

Building enterprise-scale AI solutions requires addressing fundamental architectural challenges that can make or break deployment success. Swisscom's approach centers on a multi-agent architecture designed specifically to handle the complexity of coordinating specialized agents across customer support and sales operations.

Containerized Runtimes within a Shared VPC Environment

The foundation relies on containerized runtimes within a Shared VPC environment. This setup enables secure hosting of generic customer agents alongside internal services while maintaining strict isolation between different operational units. Each agent runs in its own Docker container, providing the scalability needed to handle thousands of customer interactions monthly without compromising performance or security.

Data Sovereignty with AWS Europe Region (Zurich)

Data sovereignty stands as a non-negotiable requirement for Swiss enterprises. Swisscom's deployment in the AWS Europe Region (Zurich) ensures all customer data remains within Swiss borders, meeting stringent regulatory compliance standards. This regional deployment strategy addresses both legal requirements and customer expectations around data protection.

Integration with Existing Corporate Infrastructure

The architecture connects seamlessly with Swisscom's existing corporate infrastructure through two critical components:

• AWS Direct Connect: Establishes dedicated network connections between AWS and Swisscom's corporate network
• VPC Transit Gateway: Enables secure communication pathways across different departments and organizational units

This integration approach solves a common enterprise challenge—how to deploy modern AI solutions without creating isolated systems that can't access existing corporate resources. Your agents need access to internal applications, customer databases, and departmental systems to deliver real value.

Scalable Coordination for Multi-Agent Interactions

The scalable AI agent architecture handles the coordination complexity inherent in multi-agent systems. When a customer interaction requires input from multiple specialized agents—perhaps one for technical support, another for billing, and a third for product recommendations—the architecture orchestrates these interactions efficiently while maintaining security boundaries between different organizational domains.

Session-Level Isolation for Data Privacy

Session-level isolation within the shared infrastructure ensures that customer data from one interaction never leaks into another, even when multiple agents process requests simultaneously on shared resources.

Salesforce CRM and AI : Boost Sales, Service & Marketing

Salesforce is a CRM tool that has transformed how businesses manage customer relationships across the globe. As the leading platform in customer relationship management, it’s become the backbone of operations for companies ranging from small startups to Fortune 500 enterprises.

Customer ServiceEmily Carter

Key Components Enabling Swisscom's Agentic AI Solutions

Swisscom's agentic AI framework relies on several interconnected components that work together to enable seamless agent operations across the enterprise.

Model Context Protocol (MCP)

At the heart of this system lies the Model Context Protocol (MCP), which serves as the communication backbone for agent interactions. MCP servers provide a standardized interface that allows agents to access external tools and data sources consistently, regardless of their origin or location within the infrastructure.

Agent2Agent protocol (A2A)

The Agent2Agent protocol (A2A) complements MCP by enabling direct communication between specialized agents operating across different organizational domains. This protocol allows agents to collaborate on complex tasks that require expertise from multiple areas—such as when a customer support agent needs to consult with a billing agent or a technical troubleshooting agent. The A2A protocol ensures these cross-domain interactions happen efficiently while maintaining security boundaries.

SAIL (Service and Interface Library)

SAIL (Service and Interface Library) acts as the gateway between AI agents and Swisscom's internal applications. Rather than requiring each agent to implement custom integrations with dozens of backend systems, SAIL provides a unified access layer. This standardization dramatically reduces development time and ensures consistent behavior across all agents accessing corporate resources.

Docker container deployment

The technical implementation leverages Docker container deployment strategies that provide both scalability and isolation. Each agent runs within its own containerized environment, allowing Swisscom to:

• Deploy multiple agent instances dynamically based on demand
• Isolate agent workloads for security and resource management
• Update individual agents without affecting the broader system
• Maintain consistent runtime environments across development and production

Strands Agents Framework

Swisscom accelerated their development timeline by adopting the Strands Agents Framework, which simplifies agent construction through pre-built components and patterns. The framework includes built-in support for OpenTelemetry integration, enabling Swisscom to export performance traces directly to their existing observability infrastructure. This capability proved invaluable during development and continues to support ongoing optimization efforts as Swisscom builds enterprise agentic AI for customer support and sales using Amazon Bedrock AgentCore.

Security and Compliance in Enterprise AI Deployment at Swisscom

Managing multiple AI agents across various departments presents significant security challenges. Each agent requires access to sensitive customer data, internal systems, and corporate resources while ensuring strict separation between organizational units. Granting broad permissions is not an option due to Switzerland's strict data protection laws.

Solving the Authentication Challenge with AgentCore Identity

AgentCore Identity addresses this authentication challenge by directly integrating with Swisscom's current identity provider. The system implements strong authentication and authorization processes that operate smoothly throughout the entire multi-agent network. Whenever an agent needs to interact with an MCP server or use an internal tool, AgentCore Identity verifies its credentials against Swisscom's centralized identity management system.

Ensuring Secure Communication with Temporary Access Tokens

The authentication procedure depends on temporary access tokens that allow two-way verification. These tokens have two main functions:

• Agent-to-server validation: When an agent seeks access to an MCP server or tool, it presents a temporary token that the server verifies before granting access
• Server-to-agent validation: MCP servers can confirm they're communicating with legitimate agents authorized to access specific resources

This token-based method removes the necessity for long-lasting credentials that could pose security threats if compromised. The tokens have a set expiration time, requiring regular re-authentication and minimizing potential attack vectors.

Meeting Data Protection Compliance Requirements through Session-Level Isolation

Session-level isolation fulfills Switzerland's stringent data protection compliance demands. Each customer interaction operates in its own separate environment within the containerized runtime. Customer data from one session never mixes with another, upholding the privacy boundaries required by Swiss law. This isolation applies to memory, processing, and storage, creating secure compartments for every interaction.

This architecture is especially beneficial when dealing with sensitive telecommunications data. The combination of AgentCore Identity management, temporary access tokens, and session-level isolation establishes multiple layers of security that meet both technical needs and regulatory responsibilities.

Operational Excellence Through Amazon Bedrock AgentCore Features

Amazon Bedrock AgentCore Runtime transforms how Swisscom manages its AI infrastructure at scale. The platform delivers secure, cost-efficient hosting through Docker container deployment, allowing you to run multiple agents within a shared infrastructure while maintaining strict isolation between sessions. This containerized approach means Swisscom can deploy agents rapidly without provisioning dedicated resources for each use case.

Efficient Resource Management with Automatic Load Balancing

The automatic load balancing capabilities respond dynamically to demand patterns across Swisscom's customer base. When customer inquiries spike during peak hours, AgentCore Runtime scales resources seamlessly, ensuring consistent performance without manual intervention. You get the benefit of handling thousands of requests per month while maintaining the low latency requirements critical for real-time customer interactions. The system optimizes costs by scaling down during quieter periods, eliminating the expense of maintaining idle infrastructure.

Enhanced Memory Management for Personalized Experiences

AgentCore Memory provides Swisscom with sophisticated memory management capabilities that separate into two distinct layers:

• Session-based memory: Maintains context throughout individual customer conversations, allowing agents to reference earlier parts of the dialogue without requiring customers to repeat information
• Long-term memory: Stores insights across multiple interactions, building a comprehensive understanding of customer preferences, previous issues, and service history

This dual-memory architecture enables Swisscom's agents to deliver truly personalized experiences. When you contact support, the agent already knows your service configuration, past technical issues, and preferred communication style. The memory system tracks customer interactions across all touchpoints, creating a unified view that improves service quality with each engagement.

Data Control and Compliance with Custom Storage Strategies

The memory capabilities integrate with custom storage strategies, giving Swisscom control over data retention policies and ensuring compliance with Swiss data protection requirements. You can define precisely how long different types of customer information persist, balancing personalization benefits against privacy obligations.

Building Strong Customer Relationships for Business Growth

Your relationship with customer success determines whether your business thrives or merely survives. In today’s customer-focused market, building strong connections with the people who buy from you isn’t just a luxury—it’s the key to long-lasting business success.

Customer ServiceEmily Carter

Real-world Applications: Enhancing Customer Support and Sales at Swisscom

Swisscom's use of Amazon Bedrock AgentCore focuses on two important business areas that directly affect customer satisfaction and revenue generation. The SAM chatbot system is the main way these AI-powered interactions happen, handling thousands of customer requests every month while making sure response times are quick for real-time support.

Improving Technical Support with AI

One example of how this implementation works in practice is through technical support. When customers have problems connecting to their internet router, the AI agent uses conversational AI powered by Rasa and fine-tuned LLMs deployed on Amazon SageMaker to walk them through troubleshooting steps. What's impressive about this solution is that it's not just another basic scripted chatbot. Instead, the agent understands the context of each conversation, adjusts its responses based on the customer's level of technical knowledge, and retrieves live network diagnostics using secure MCP server connections.

This system has proven capable of managing complex troubleshooting situations involving multiple steps—something that previously required human intervention. As a result, resolution times have decreased significantly and support staff can now focus on more intricate issues.

Personalizing Sales Pitches with Customer Insights

The second major application where Amazon Bedrock AgentCore Memory truly stands out is in delivering personalized sales pitches. This system has the ability to analyze a customer's interaction history across various touchpoints, such as previous inquiries or expressed preferences.

When a customer reaches out to Swisscom, the sales agent can access this comprehensive profile in order to provide recommendations tailored specifically for them. For example:

• A customer who frequently travels internationally might receive targeted offers for roaming packages.
• A family with high data consumption gets recommendations for upgraded internet plans.

By leveraging these insights from past interactions, Swisscom is able to offer highly relevant suggestions that resonate with individual customers.

Enabling Smooth Communication Between Agents

Another key aspect of both applications is the use of high-performance agent-to-agent communication protocols. These protocols allow different agents within Swisscom's organization to share information seamlessly across departmental boundaries.

For instance:

1. If a support agent identifies an opportunity for upselling during a conversation with a customer, they can quickly pass that information along to a sales agent.
2. In cases where service status needs verification, support agents can coordinate directly with network monitoring agents.

Such coordination happens almost instantaneously—within milliseconds—to ensure that customers experience uninterrupted conversations similar to those found in modern digital platforms.

Overcoming Multi-Agent System Challenges in Enterprise Settings

Orchestrating AI agents across departments/systems presents unique complexities that Swisscom had to address head-on. When you're managing multiple specialized agents operating across various organizational units, coordination becomes exponentially more difficult. Each department has its own workflows, data structures, and security requirements. Swisscom's multi-agent architecture needed to ensure these agents could communicate seamlessly while respecting organizational boundaries.

The challenge intensifies when agents need to access shared resources. You can't simply give every agent unrestricted access to all systems—that creates security vulnerabilities and performance bottlenecks. Swisscom implemented mechanisms for controlled resource sharing that balance operational efficiency with security constraints. Through their use of temporary access tokens and bidirectional validation between agents and MCP servers, they created a system where agents can access only the resources they need for specific tasks. This approach maintains strict isolation while enabling the collaboration necessary for complex customer interactions.

Cross-domain integration requires robust governance frameworks. Swisscom recognized early that without proper oversight, their multi-agent ecosystem could quickly become unmanageable. They're focusing their future roadmap on implementing governance with centralized agent registry systems that provide:

• Version control for tracking agent iterations and ensuring consistency across deployments
• Standardized documentation that makes it easier for teams to understand and utilize existing agents
• Usage monitoring to identify performance patterns and optimization opportunities
• Security audits that verify compliance with Swiss data protection regulations

This governance strategy transforms what could be a chaotic collection of independent agents into a coordinated system. You gain visibility into which agents are being used, how they're performing, and where potential conflicts or redundancies exist. The centralized registry becomes your single source of truth for agent management, enabling teams to discover and reuse existing capabilities rather than building duplicate solutions.

Agentic AI in Customer Service : Transforming Support

Agentic AI represents a significant shift in how businesses approach customer service. Unlike traditional AI systems that follow predetermined scripts, agentic AI operates with autonomy—making independent decisions, learning from interactions, and adapting its behavior.

Customer ServiceEmily Carter

How Developer Experience & Framework Adoption Impact Velocity

The Strands Agents framework has fundamentally transformed how Swisscom's development teams approach agentic AI implementation. It provides simplified APIs that hide the complexity of managing multiple agents, allowing developers to focus on business logic instead of infrastructure issues. The framework also includes tools for tracking and evaluating agent performance, which makes it easier to understand how agents are behaving and reduces the time spent on debugging.

Benefits of Strands Adoption

One of the biggest advantages Swisscom gained from using Strands is the ability to deliver value faster. With this framework, teams can quickly create and test new agent features without having to completely rewrite existing code. This was especially helpful when Swisscom moved their existing conversational AI components to a new architecture powered by AgentCore. Instead of taking months to migrate with custom solutions, they were able to complete the process in weeks while keeping everything running smoothly.

Here are some specific ways Strands has helped Swisscom:

• Faster Prototyping: Teams can quickly build and test new agent capabilities.
• Smooth Transition: Moving away from old frameworks is easier without major rewrites.
• Operational Continuity: The migration process doesn't disrupt ongoing operations.

Ongoing Productivity Gains

The benefits of Strands go beyond just the initial development phase. Here are some ways it has improved productivity for Swisscom's development teams:

• Standardized Patterns: Using consistent methods for building agents makes it easier for new team members to get up to speed.
• Performance Tracing: Integrating OpenTelemetry allows for seamless export of performance data into existing monitoring systems.
• Quality Assurance: Pre-built evaluation frameworks speed up the quality assurance process.
• Comprehensive Documentation: Documentation that follows enterprise development standards ensures everyone is on the same page.

Increased Development Velocity

After adopting the Strands framework, Swisscom saw a noticeable increase in their development speed. Teams can now make multiple iterations on agent capabilities each week instead of just once a month. This allows them to respond more quickly to customer feedback and business needs, something that was previously difficult due to longer development cycles.

Allocating Resources for Innovation

With the reduced complexity in developing agents using Strands, Swisscom can now allocate more engineering resources towards innovative projects rather than maintaining existing ones. This shift in focus is crucial for continuously improving customer-facing AI systems and staying ahead in a competitive market.

Supporting Ambitious AI Strategy

The streamlined workflows created by Strands directly align with Swisscom's ambitious AI strategy. Even as their ecosystem of agents becomes more complex, they are able to maintain their pace of innovation. This ensures that technical infrastructure never becomes a barrier to achieving business goals.

Long-Term Insights & Data Utilization from AI Interactions

Swisscom's implementation of Amazon Bedrock AgentCore Memory transforms every customer interaction into a valuable data asset. The system captures detailed conversation histories, decision patterns, and customer preferences across all touchpoints. This persistent storage enables agents to recognize returning customers and recall previous interactions, creating continuity that traditional support systems struggle to achieve.

Customer Intelligence from AI Interactions

The customer intelligence from AI interactions extends beyond simple conversation logs. AgentCore Memory stores contextual information about customer behavior patterns, product preferences, and historical issues. When a customer contacts support about internet connectivity problems, the agent accesses their complete interaction history—previous technical issues, resolution methods that worked, and even the customer's communication style preferences. This depth of context allows for truly personalized support experiences.

Sophisticated Session Analysis

Sophisticated session analysis happens at multiple levels within Swisscom's framework. The system analyzes individual sessions to identify successful resolution patterns and common pain points. You can track which agent responses lead to customer satisfaction and which trigger escalations. This granular analysis feeds directly into agent improvement cycles.

Long-Term Memory Storage Architecture

The long-term memory storage architecture supports both operational and strategic objectives. Sales teams leverage historical interaction data to identify upselling opportunities based on customer usage patterns and expressed needs. When an agent detects a customer frequently experiencing bandwidth limitations, it can proactively suggest upgraded service plans with relevant context from past conversations.

Analytical Infrastructure for Actionable Insights

Swisscom's analytical infrastructure processes this stored data to generate actionable insights. The team identifies trending issues before they escalate into widespread problems. Service improvements emerge from pattern recognition across thousands of interactions. Product development teams receive direct customer feedback synthesized from agent conversations, creating a feedback loop that drives continuous innovation.

This data-driven approach to How Swisscom builds enterprise agentic AI for customer support and sales using Amazon Bedrock AgentCore demonstrates the strategic value of treating AI interactions as intelligence assets rather than ephemeral exchanges.

Future Roadmap & Best Practices for Enterprise Agentic AI at Swisscom

Swisscom's journey with Amazon Bedrock AgentCore continues to evolve as the company refines its approach to enterprise agentic AI platforms. The future roadmap centers on expanding agent sharing capabilities across business units, creating a more interconnected ecosystem where specialized agents can serve multiple departments without compromising security or performance.

Best Practices for Cross-Domain Integration

Cross-domain integration best practices form the backbone of Swisscom's expansion strategy. The company is implementing a centralized agent registry that catalogs all available agents, their capabilities, and access requirements. This registry serves as a single source of truth for teams looking to leverage existing AI solutions rather than building from scratch.

The governance framework receives continuous attention through:

• Standardized documentation practices ensuring every agent includes clear usage guidelines, API specifications, and integration requirements
• Version control mechanisms tracking agent evolution and maintaining backward compatibility
• Usage monitoring dashboards providing visibility into agent performance across departments
• Regular security assessments validating compliance with Swiss data protection regulations

Prioritizing Reusability in Enterprise Agentic AI

You'll notice how Swisscom builds enterprise agentic AI for customer support and sales using Amazon Bedrock AgentCore by prioritizing reusability. Agents designed for specific customer service scenarios can be adapted for sales applications, reducing development time and maintaining consistency across customer touchpoints.

Supporting Infrastructure Improvements

The technical infrastructure supports this vision through enhanced MCP server capabilities and refined A2A protocols. These improvements enable faster agent discovery and more efficient resource allocation across the shared VPC environment.

Ensuring Security Compliance

Security audits occur quarterly, examining authentication flows, access token management, and data sovereignty compliance. This rigorous approach ensures that as Swisscom scales its agentic AI deployment, regulatory requirements remain satisfied while operational efficiency improves.

Lessons Learned for Building Enterprise-Scale AI Solutions

The blueprint Swisscom has established demonstrates that enterprise-scale AI solutions require careful planning around governance, security, and cross-functional collaboration. Your organization can apply these principles when building similar systems, adapting them to your specific regulatory environment and operational needs.

FAQs (Frequently Asked Questions)