MIC Analyzes AI Agents Moving into Enterprise Core via On-Premises Architecture and Open Models

The Market Intelligence & Consulting Institute (MIC) reported that AI is transitioning toward real-world task execution and autonomous value generation, fostering an agent economy characterized by autonomous collaboration and transactions. Synthesized international research indicates that 53% of enterprises anticipate generative AI becoming widespread within three years, with 39% already utilizing AI Agents and 23% expanding agent applications internally. MIC Deputy Director Chung-Chieh Yang stated that as open-source capabilities near proprietary models, enterprise evaluations increasingly prioritize inference cost, deployment autonomy, data governance, and application integration. Enterprises are combining open-source models, proprietary databases, and Retrieval-Augmented Generation (RAG) to develop on-premises solutions, such as petrochemical industrial safety prediction models and sensitive workflows in finance and healthcare. Regarding computing architecture, evolving AI workloads are driving tiered computing structures across the cloud and edge. AI PCs are shifting from isolated features toward contextual understanding and cross-application collaboration for immediate endpoint tasks. MIC anticipates the next phase will feature governed, cross-organizational AI Agents interacting across digital value networks.

Key takeaways

• MIC reports AI Agents are driving an agent economy with capabilities spanning demand discovery, bargaining, and contract delivery.
• Survey findings reveal 53% of enterprises expect generative AI adoption within three years, while 39% have begun using AI Agents.
• Enterprises are adopting on-premises architectures combining open models and RAG for domain tasks like industrial safety while managing costs and governance.
• Computing architectures are restructuring into tiers across cloud, desktop, and AI PCs based on computational needs and privacy requirements.