Presented by Outshift by Cisco
Gartner predicts that by 2028, one-third of human interactions with generative AI will shift from user prompts to direct interfaces with autonomous, intent-driven agents. This marks a significant leap from the reactive AI assistants many users currently know.
“Agents represent the next evolutionary step in generative AI,” says Vijoy Pandey, SVP/GM of Outshift, Cisco's incubation arm. “For executives, the message is clear: prepare now. With just three years left, begin by implementing assistants and tackling manageable use cases before evolving to more critical applications.”
AI agents are akin to tireless, specialized employees tailored to specific tasks that collaboratively address business challenges. Tim Tully, partner at Menlo Ventures, highlights an ongoing trend: “We're witnessing customer success companies replace and enhance their teams with agents, boosting scalability. This is evident in marketing automation and code generation, and I expect agents to proliferate further into software engineering.”
The Big Three—Google Cloud, Microsoft’s Copilot, and AWS's Q—are actively developing generative AI agents, indicating the emergence of a transformative technology.
The Distinction Between Agents and Assistants
So, what differentiates AI agents from earlier AI-powered assistants? AI assistants react to user prompts, employing large language models (LLMs) and natural language processing (NLP) to provide answers and contextual content in a conversational interface.
In contrast, AI agents are proactive and autonomous, capable of making decisions and taking actions without human intervention. They continuously analyze domain-specific data in real-time, managing complex workflows independently while working towards specific objectives.
Unlike traditional assistants, agents produce high-quality content that can shorten review cycle times by 20% to 60%, thanks to an easily accessible audit trail of tasks and data sources. “Think of them as specialized employees that specialize in particular tasks and collaborate to tackle broader business problems,” Pandey explains. For instance, in financial services, an agent can detect and prevent fraud in real-time, while in HR, it can analyze data to identify top talent or predict turnover.
When integrated into a multi-agent framework, these systems can collaborate across various skill areas, make informed decisions, and manage complex workflows autonomously. However, a dedicated orchestration layer for agent collaboration is still in development, representing a significant opportunity for startups.
“There’s a need for a Kubernetes-like infrastructure for agent technologies—something that’s just right for running these specialized workloads,” Tully notes. The goal is to connect these thin agents, enabling seamless communication through yet-to-be-established protocols.
Transitioning from Assistants to Agents
The Cisco AI Readiness Index reveals that while 97% of organizations wish to leverage generative AI, only 14% have implemented it—highlighting a substantial gap. Common hurdles include understanding where to begin, ensuring ROI, and addressing trust, safety, and security challenges.
“There are limitations to the internal reasoning and planning required for agents to tackle ambiguous problems,” Pandey explains, pointing to the importance of clear instructions to guide agents effectively.
Organizations should start with straightforward business cases rather than ambitious projects. Empowering citizen developers—individuals within business functions who understand processes and how to enhance them—is critical, especially given the limited availability of generative AI developers.
Before embarking on their AI journey, organizations must prioritize data cleansing to ensure proper identity management and access controls. “Start with manageable business cases rather than moonshots,” Pandey advises. This approach allows organizations to navigate and refine their pipeline while educating citizen developers, ultimately laying a robust foundation for future advancements in AI.
As industries transition from assistants to agents and LLMs continue to improve, every organization stands to benefit from the transformation brought by agentic generative AI.
Watch the entire conversation with Outshift’s SVP/GM Vijoy Pandey, Tim Tully from Menlo Ventures, and VB editor-in-chief Matt Marshall [here].
Ranking
