Synopsis: Babak Hodjat, CTO for artificial intelligence (AI) for Cognizant, traces the evolution of AI agents from the initial launch of Apple Siri to a new era of agentic AI.

In this interview, Babak Hodjat, CTO for AI at Cognizant, explains how the journey to generative AI (GenAI) and intelligent agents began long before Siri, with early AI agents operating in limited environments like the web. These early systems evolved into today’s more powerful AI agents, thanks to GenAI’s ability to better understand and reason with natural language. Hodjat highlights the growing importance of multi-agent systems that can interact and collaborate to carry out tasks on behalf of users. This interconnected network of agents reflects an emerging digital ecosystem that mirrors the structure of modern enterprises—working much like microservices with the ability to coordinate across functions.

As organizations begin adopting these AI agents, interoperability and negotiation between agents become critical. Hodjat describes a future in which users will rely on multiple agents that communicate behind the scenes to provide recommendations or negotiate services, much like human assistants. He shares an experiment where his personal agent negotiated vacation deals across third-party providers, illustrating how autonomous agents can operate within defined parameters to support consumer decision-making. However, he notes that current AI models tend to agree too readily, and new fine-tuning strategies may be needed to simulate more realistic and productive negotiations among non-aligned agents.

Looking ahead, Hodjat emphasizes that successful agent adoption should be gradual and modular. He commends early adopters who recognize the incremental nature of “agentification” and who prioritize agent security, autonomy, and fallback mechanisms. Agents must be designed to operate responsibly, with clear lines between autonomous functions and those that require human or rule-based oversight. As AI agents take on more complex roles, particularly in sensitive or critical enterprise processes, securing their behavior through code, access control, and organizational safeguards becomes essential. Hodjat closes by stressing that despite the automation potential, human intelligence will always remain the foundation for responsible AI deployment.