**London**: Former Oracle executive Bhat elaborates on the distinctions between generative AI and agentic AI, highlighting the demand for intelligent applications and the challenges enterprises face in integrating autonomous agents into their systems amid evolving AI technologies.
At a recent presentation, Bhat, a former Oracle executive and now technology leader in the agentic AI sector, shared insights into the rapidly evolving landscape of artificial intelligence, particularly focusing on the distinction between generative AI (GenAI) and agentic AI. The discussion, reported by InfoQ, traced Bhat’s journey through the AI field, highlighting the ongoing demand from enterprises for solutions that leverage their existing data for more intelligent applications.
Bhat opened the presentation by recalling his early experiences at Oracle, where he was tasked with heading AI strategy. He noted that discussions around artificial intelligence back then often stalled when IT leaders expressed their frustration, saying, “the data goes to die” in traditional data lakes and demanding actionable insights from their collected data. This led him to explore potential acquisitions and join the startup Rockset, where he helped develop a database that integrated search and AI functionalities, ultimately acquired by OpenAI.
The conversation proceeded to the adoption rates of generative AI, which Bhat asserted were almost double those of personal computers and the internet two years post-commercialization. He emphasised that interest in generative AI remains strong, but the pressing question for enterprises is how to transition from merely creating applications to developing “intelligent agents.”
Bhat introduced the concept of agentic AI, which he described as a sophisticated evolution beyond GenAI. While GenAI often operates on individual requests with limited context (zero or one-shot prompting), agentic AI is characterised by multi-step operations. Citing examples from software development, he articulated the human-like processes that AI must adopt to be effective, which includes gathering information, iterating through problems, and improving based on past experiences.
A notable aspect of agentic AI is its capacity for autonomy and collaboration. Bhat explained that these agents could engage with one another to accomplish tasks, such as rebooking flights by leveraging various educational agents. This collaborative aspect is crucial as additional intelligence can improve overall performance, he claimed.
He presented data citing a comparative analysis of AI models, revealing that in certain workflows, earlier versions of GPT could outperform later ones when employed in agentic frameworks. This finding highlighted the potential for micro-agent collaboration, similar to microservices in software architecture, where specific agents are designed for narrowly defined tasks and work together.
Addressing the future direction of agentic AI, Bhat underscored the challenges of integrating such technology into existing enterprise systems. Early attempts to create agents that could handle full roles were largely unsuccessful, demonstrating the necessity to refine agent roles down to discrete tasks. This will result in an ecosystem of specialized agents, each optimised for specific functions within the organisation’s data landscape.
The complexity of building and managing these agents was another key consideration, with discussions around orchestration and governance cited as ongoing challenges. Companies are proactive in contemplating how they can create agentic workflows, drawing upon their existing data resources and customising solutions for their specific needs. Notably, enterprises like Salesforce and OpenAI are venturing into the domain of custom enterprise agents because of the varied and unique demands of their clients.
Bhat painted a picture of a future where agents can operate independently while still adhering to corporate governance structures and security protocols. Concerns surrounding privacy, data access, and cybersecurity were stressed, as the introduction of autonomous agents could also pose new risks if not properly managed.
In a broader context, the conversation pivoted to the human element of adopting agentic AI. Reflecting on industry trends, Bhat suggested that successful early adopters invest significantly in adapting their processes to integrate AI technologies effectively. He noted, “Most of the leaders who are embracing this change are spending 70% of their time and energy and investment into people and processes, and only 30% on the rest.” This consideration underlines the potential disruption in workplace dynamics as AI technologies reshape traditional task execution.
Overall, the discourse around agentic AI presents an array of technological advancements and infrastructural challenges, with the promise of significant productivity gains tempered by the need for careful integration into existing systems and processes. As companies continue to navigate this evolving landscape, understanding the distinctions between generative and agentic AI will be vital for optimising performance and realising the full potential of AI-driven operations.
Source: Noah Wire Services
Subscribe to Industry Updates
