Swisscom unlocked thousands of hours: how AI agents transformed telecom sales and support

Swisscom, a leading Swiss telecom provider with a turnover of approximately $19 billion, faced the challenge of scaling AI solutions: traditional automation approaches could no longer meet growing business demands. The implementation of AI agents allowed them not only to reduce the development time for new solutions to 3-4 weeks but also to ensure unprecedented scalability for processing thousands of requests per month in sales and customer support departments, while maintaining the strictest data privacy standards.

In a large telecom company, routine tasks consume thousands of hours that employees spend searching for information, drafting standard proposals, and answering common questions. This is not just wasted time; it's lost opportunities to improve service quality and business development. Every hour spent on routine tasks represents direct costs that can and should be reduced. AI agents are not just tools for automation; they are a way to give businesses back time and focus on strategic objectives.

Swisscom’s automation ceiling and strict regulations

Swisscom, like many large companies, had long and successfully automated many processes. However, with the increasing volume of data and the growing complexity of customer inquiries, traditional automation methods reached their "ceiling." The company faced critical challenges that required a new approach:

• Scalability. How to coordinate the work of dozens and hundreds of AI agents across various departments while maintaining a high level of security and performance?
• Security and Authentication. There was a need for secure, transitive access with minimal privileges, considering overlapping permissions for customers, agents, and departments. Data privacy in Switzerland is not just a requirement; it is the foundation of trust.
• Integration. How to centralize and ensure the compatibility of new AI solutions with the existing complex infrastructure and numerous internal systems?
• Analytics and Cross-Departmental Collaboration. Efficient collection and utilization of customer data from multiple interactions were extremely difficult. For example, if a customer's router failed, the problem could be related to billing, network, or configuration, each belonging to different departments. This required seamless cross-departmental coordination, which in manual mode consumed a lot of time.

All these factors led to employees spending too much time on routine operations, even with a high level of automation, instead of focusing on strategic tasks and improving customer experience.

The path to AI agents: why traditional solutions were insufficient

Swisscom already used various scripts and automated systems, but they were effective only for strictly formalized and predictable processes. As soon as a request went beyond the template, for example, a non-standard document, a complex customer question, or the need to retrieve information from three different systems simultaneously, the process slowed down and reverted to manual mode.

The company realized that it needed not just a tool to automate individual tasks but an intelligent system capable of understanding context, making autonomous decisions within defined rules, and coordinating actions between various data sources and services. This led to the need for AI agents – systems that can operate autonomously, interacting with other systems and people, while being integrated into the existing IT infrastructure.

How the AI agent was designed: architecture and functionality

Swisscom chose a flexible and scalable platform to build its AI agents, which allowed developers to focus directly on the agents' logic rather than on infrastructure tasks. Key architectural components included:

• Agent Runtime Environment. A system that provides secure hosting, automatic scaling, and session isolation for each agent. This allowed developers to quickly deploy new agents without needing deep DevOps expertise.
• Identity Management. Integration with Swisscom's existing authentication system ensured secure and seamless interaction between agents, tools, and data sources, adhering to the principle of least privilege.
• Memory Mechanism. Reliable storage for both short-term (session) and long-term agent memory. This is particularly crucial for B2C operations, where understanding customer context across interactions is vital for personalizing services.
• Agent Development Framework. A framework was chosen that simplified agent creation, accelerated development cycles, and had built-in capabilities for tracing, evaluation, and logging, which significantly streamlined debugging and optimization.

The agents were required to perform several key functions: generate personalized commercial offers, handle technical customer inquiries, and gather information from various systems to provide comprehensive answers, all while maintaining conversation context and interaction history.

Implementation: from pilot to widespread adoption

The implementation of AI agents began with two key B2C scenarios: generating personalized commercial offers and automated customer support for technical issues. Both agents were integrated into Swisscom's existing SAM chatbot, which demanded high performance and low latency.

The implementation process was iterative and included the following stages:

• Pilot Projects. Development of the first agents for specific tasks. Development teams, even those without prior experience with the chosen framework, were able to present the first working prototypes to the business within 3-4 weeks.
• Training and Adaptation. Customer support and sales staff were trained to work with the new agents. It was important to show them that agents are assistants who free them from routine tasks, not replacements.
• Scaling. Gradual expansion of agent functionality and their application in other departments. Thanks to the chosen architecture, the system could efficiently handle thousands of requests per month for each agent, maintaining low latency and optimizing infrastructure costs.
• Feedback and Improvement. Continuous collection of user feedback and analysis of agent performance allowed for continuous improvement of their functionality and accuracy.

One team, which previously used a more complex framework for graph computing, migrated to the new platform and noted a significant reduction in development complexity and acceleration of release cycles for new features. This confirmed the correctness of the architectural choice.

Results: quantitative and qualitative changes

Thanks to the implementation of AI agents, Swisscom achieved significant improvements:

• Accelerated Development. The time required to create and deploy new AI solutions was reduced from several months to 3-4 weeks. This allows the company to respond faster to market changes and launch new services.
• Sales and Support Efficiency. Agents took over routine tasks, allowing employees to focus on more complex cases and personalized customer interactions. This led to improved service quality and increased customer satisfaction.
• Scalability. The system easily handles thousands of requests per month, ensuring low latency and optimizing infrastructure costs.
• Regulatory Compliance. Built-in mechanisms for data sovereignty and user privacy ensure compliance with strict Swiss regulatory requirements.

Overall, Swisscom was able to significantly accelerate AI adoption, improve operational efficiency, and strengthen its market position, while maintaining high security and customer trust.

How to implement this in your company: steps to AI transformation

The Swisscom case demonstrates that AI agents are not just a trend but a powerful tool for solving real business problems, especially under strict security and scalability requirements. If your company faces similar challenges, here's where you can start:

• Identify bottlenecks. Find processes where employees spend a lot of time on routine tasks, information retrieval, or repetitive requests. These are ideal candidates for automation with AI agents.
• Start with a pilot. Choose one or two non-critical but high-volume tasks where the impact of implementation will be immediately noticeable. This will allow you to gain experience, gather feedback, and demonstrate the value of AI agents to management and the team.
• Focus on security and integration. Choose a platform that provides robust authentication, access management, and seamless integration with your existing IT infrastructure. This is critical for scaling and regulatory compliance.
• Train your teams. Ensure that employees understand how to work with AI agents and see them as assistants, not threats. Team involvement is key to successful implementation.
• Utilize frameworks. Using ready-made frameworks for agent development will significantly accelerate the process and reduce complexity, allowing you to focus on business logic.

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