Single copilots are becoming less prevalent, as competitive differentiation now focuses on deploying networks of specialized agents that can collaborate and self-evaluate. The recent AI Impact Series by VentureBeat, held in San Francisco and sponsored by SAP, examined the deployment and governance of multi-agent AI systems.
Yaad Oren, managing director of SAP Labs U.S. and global head of research and innovation at SAP, alongside Raj Jampa, SVP and CIO at Agilent (a provider of analytical and clinical laboratory technologies), discussed the implementation of these systems in practical settings while adhering to cost, latency, and compliance constraints. Oren emphasized the importance of ensuring customers can safely scale their AI agents, highlighting the necessity of continuous monitoring and checkpoints for improvement.
Agilent is currently in the process of integrating AI across their operations, with promising results but ongoing challenges related to vulnerability and scaling. Jampa noted that they are moving beyond initial exploration and are now addressing issues related to monitoring and cost optimization.
AI deployment at Agilent is structured around three strategic pillars: enhancing product innovation by integrating AI into instruments, identifying valuable AI capabilities for customers, and improving internal operations through solutions like self-healing networks. Jampa underscored the significance of governance frameworks that balance compliance and security with operational efficiency. An example provided was an agent that caused issues during a configuration update due to a lack of monitoring checkpoints, although the system quickly detected the problems thanks to robust auditing practices.
Addressing integration challenges between AI agents and existing enterprise solutions is another focus area, with Oren stating that leveraging cloud-based systems simplifies these connections. SAP’s Business Data Cloud serves as a unified data platform, allowing for effective indexing and integration of business data.
Key elements for successful agent deployments include a unified data layer, effective orchestration of connections, and strong privacy and security measures. Oren pointed out that as agents become more integrated within enterprise frameworks, the management of their access and identity becomes increasingly critical.
Source: https://venturebeat.com/ai/vb-ai-impact-series-can-you-really-govern-multi-agent-ai/