Guest Post by Melissa Incera, S&P Global Market Intelligence
This is the final blog in the three-part series on emerging generative AI use cases (read the first and second blog here), a companion to the 451 Research report, “Generative AI: Balancing Innovation and Privacy.”
In the first two installments of our series, we explored how enterprises have quickly moved to harness the capabilities of generative models. A new wave of AI development is opening the door to more complex, high-value use cases as techniques for improving model output quality, and capabilities continue to evolve. In this blog, we will examine two more sophisticated applications of GenAI: advanced information retrieval and agentic AI.
Advanced information retrieval
Advanced information retrieval is one such use case that has emerged from the enterprise need to streamline access to critical data. As application sprawl complicates employees’ ability to efficiently access information, organizations are leveraging LLMs to centralize discovery, enhance productivity and reduce the time spent searching for documents, policies and procedures.
This use case relies heavily on a mechanism called retrieval-augmented generation (RAG), a natural language processing technique that merges retrieval and generation models, allowing an LLM to draw on an organization’s information sources to improve the quality and relevance of model outputs. As a result, enterprises are dramatically increasing the volume and diversity of data — e.g., customer, operational, employee and market intelligence data — that they leverage in their machine learning workflows.
Advanced information retrieval helps users access information more quickly and in more useful formats. It is especially useful in applications where factual accuracy and up-to-date information are critical, such as question-answering systems, chatbots and content generation tasks. Through its use, businesses can improve the speed and accuracy with which employees can locate necessary information, enabling them to focus on higher-value tasks and decision-making processes.
Agentic AI
While advanced information retrieval has a wide variety of applications, agentic AI takes this a step further by allowing systems to not only find and summarize relevant data but also to interpret it, make recommendations, and take actions based on user needs and insights.
An AI agent is a decision-making engine powered by a large language model (LLM). Its key features include the ability to gather information, reason, plan, act, and learn from its behavior. Typically, agents use an LLM to orchestrate interactions across various AI components or services. Task- or domain-specific specialization is achieved by fine-tuning the LLM with data and providing access to specific tools (such as calculators and search engines) and resources (such as databases and instructions). Here, RAG also plays a crucial role as a foundational mechanism for retrieving information.
Today, agents operate at constrained levels of automation, but as they advance, they will increasingly function like digital knowledge workers, autonomously executing workflows, completing tasks, and managing other agents or processes. According to a recent 451 Research AI & Machine Learning survey, 58% of respondents are seeking opportunities to implement AI assistants and agents in their organizations, while another 40% are open to exploring them.
Looking ahead
Both of these use cases demonstrate the expanding potential of generative AI as deployments become more sophisticated. However, these advancements also increase the complexity of implementation. As we move from stand-alone LLMs to integrated AI systems that can act and respond to user inputs in real-time, addressing security implications and establishing secure frameworks within a trusted environment becomes even more essential. The future holds promising opportunities, but success will require careful execution and continuous adaptation to the evolving AI landscape.
About the Author
Melissa Incera is a research analyst on the 451 Research Data, AI & Analytics team within S&P Global Market Intelligence. Her focus lies in generative AI, its emergent use cases in enterprise applications and the foundation models that underpin it. Specifically, she tracks the text-generation landscape and its offshoots in enterprise search, AI assistants, marketing and writing tools. Prior to joining the AI team, Melissa spent a few years covering technology M&A and financial markets looking at high-level trends, noteworthy deals and investments, and forecasting activity across technology subsectors. She also did some work with the Workforce Productivity & Collaboration team, where she researched the future of work and emerging applications to enhance the employee experience. Melissa holds a bachelor’s degree in English Literature from Boston College and an MBA from ESADE, Ramon Llull University in Barcelona.
