Building an Agent in CCM (Part 1) — The Case for Open Formats

2025 was supposed to be the year of the agent. In some ways it is, and in others not quite yet.
Before discussing what it takes to build an agent within Customer Communication Management (CCM), it is worth clarifying what an agent actually is and what makes one valuable in the first place.

AI agents combine large language models for reasoning and decision-making with tools that allow real-world interaction. They are designed to complete complex tasks with limited human involvement. A simple and already common example would be ChatGPT using web search to answer a user query. In that case, the user activates a tool and the agent then decides how to use it to improve its response. The purpose of such tools is to extend the agent’s context with additional information that enables it to generate a more accurate or complete answer.

Once a combination of information and query leads to consistent results, the next challenge becomes scalability: how to apply this successful pattern to similar cases, or how to enable the agent to adapt dynamically when inputs, data or expectations change. That is essentially what agents are meant to achieve — turning dynamic, variable systems into reliable workflows that make users more productive in their specific task domain.

Within CCM software, this concept is particularly relevant. Enterprises often dedicate entire teams to maintaining templates, designing reusable building blocks and ensuring compliance. Thousands of person-days each year are spent on repetitive work such as specifying, creating and testing communication templates. Agents have the potential to assist across that entire chain — from design to validation and maintenance — by automating or accelerating parts of these processes. Yet before this potential can be realized, one key prerequisite must be met: the availability of open and standard formats.

The Case for Open Formats

The software-as-a-service evolution of the past decade has shifted many IT-heavy tasks toward business users. Tools such as Canva have done for design what CCM platforms are now doing for enterprise communication. But while those tools simplify the interface, AI does not need a friendly user interface to function. What it needs is structured, machine-readable input — source files, configuration files or templates that clearly describe the desired output. The crucial question then becomes whether those files exist in open, widely used formats such as XML, JSON or JavaScript, or whether they are locked inside proprietary or binary ones.

Commonly available language models such as GPT, Claude, Gemini or Mistral are not trained on a vendor’s proprietary resource format or internal scripting language. While it is possible to fine-tune a model for that, the additional cost and complexity quickly outweigh the benefits, especially for enterprise customers using CCM tools. If, on the other hand, a CCM platform already provides access to standard, open formats, organizations can begin to deploy AI today to improve their processes and to build agents that create or modify resources without the need for costly customization.

The advantages of open formats extend far beyond template creation. CCM platforms are highly integrated within the enterprise landscape, constantly exchanging data with systems such as CRM, ERP, marketing automation or document management. In the near future, agents that monitor Kubernetes clusters, perform health checks on databases or trigger build processes in Git will likely become common. In all these cases, both the human operator and the agent benefit from an environment built on transparent, standardized components. It improves interoperability, reduces security risks and simplifies automation.

Conclusion

In this first part of Building an Agent in CCM we have defined what an agent is, looked at simple examples of how it works and discussed where such systems could add value within CCM. Most importantly, we established that open formats and the consistent use of standards are the foundation for building effective agents. They increase efficiency, reduce complexity and make both human and machine collaboration more productive.