Companies plan to use AI agents – scaling takes centre stage

Reply [EXM, STAR: REY] has published the study “Agentic AI: From Copilot to Autopilot” conducted by market research firm Lünendonk & Hossenfelder among CIOs, IT leaders, and department heads across German-speaking countries. According to the findings, 43 per cent of respondents are already planning to deploy AI agents. However, only about one in four AI projects has successfully transitioned from proof of concept to production. For enterprise leaders, this highlights a critical imperative: establishing how to reliably integrate agent-based applications into daily operations and expand them across additional business processes.

With agents, artificial intelligence becomes actionable for complex, multi-step workflows. They can autonomously break down objectives into sequential steps, synthesize data across disparate sources, and interact directly with enterprise applications, automating end-to-end processes that previously demanded extensive manual intervention.

To prepare for production-grade deployment, companies are establishing the necessary technological and organisational foundations: 98 per cent plan to invest in data quality and system integration, while 94 percent are targeting investments in platforms and tools for AI development and orchestration. Centralised governance is also taking priority, with 62 per cent planning to implement a central registry of pre-approved AI agents to monitor and control internal use.

Driving these investments is a clear financial mandate: 96 per cent of respondents expect AI agents to deliver higher efficiency and lower operational costs. Furthermore, 89 per cent see significant potential in modernising legacy environments. AI agents can systematically analyse and document legacy applications, unlocking the institutional knowledge required for long-term maintenance and modern development.

Two customer projects from Reply illustrate how agentic AI is being deployed and scaled across real-world business environments. At Audi, a multi-agent system developed together with Comsysto Reply supports software development for the central B2B used-car platform, spanning requirements engineering, code generation, and automated quality assurance. Siemens Healthineers, together with Ki Reply and Data Reply, developed the “Cerebra” Agent Factory, which standardises and accelerates the end-to-end development of specialised AI agents.

Filippo Rizzante, CTO of Reply, commented: “The most critical phase begins after the pilot project. In order to be scaled successfully, AI agents must understand the specific business context of an organisation and be securely connected to its data, legacy systems and governance. This is where the “Reply Model Factory” comes in: It translates institutional knowledge into custom AI models and agents that can be seamlessly integrated into existing processes. In doing so, we lay the groundwork for turning standalone AI applications into scalable, enterprise-wide process automation.”

The full study, “Agentic AI: From Copilot to Autopilot,” is available for download here.