Topics In Demand
Notification
New

No notification found.

Entering the Dawn of Enterprise Agentic AI
Entering the Dawn of Enterprise Agentic AI

43

0

Entering the Dawn of Enterprise Agentic AI 

Author: Ankit Goyal, Senior Vice President — EPM Practice, Polestar Solutions 

 

While 2022-2023 revolved around Generative AI and its implementation, we're now witnessing the growth of a much more evolved framework: Agentic AI for enterprise intelligence. 

In case you missed it: Google has launched Gemini 2.0— the AI model for the agentic era with improvements like multimodal reasoning, long context understanding, complex instruction planning & more. NVIDIA has launched AI Blueprints to build the agentic AI. Overall, AI agents are emerging as a transformative force in how AI operates in enterprise settings, marking a fundamental shift in the field. 

Understanding the Agentic Evolution 

Though various definitions exist, at their core, agents are autonomous systems that can operate independently by perceiving its environment and acting on it. Especially with LLMs maturing in their capabilities like planning (reasoning) and ability to understand complex inputs and breaking the task into multiple steps—the foundation for true agentic systems is fully formed. 

But before the transition to agents, understanding the need for agentic systems is vital. Currently there is confusion about workflows and automation in the name of Agentic AI. All three of them have distinct use cases. For example: 

1. Traditional Automation: Rule-based execution of predefined tasks like sending a notification when someone signs up. 

2. Workflow Orchestration: Structured paths coordinating with LLMs and/or tools – like sending different types of customer complaints along different pipelines 

3. True Agents: Autonomous systems capable of handling tasks even non-deterministic in nature with feedback loops in place; like an agent that can automatically close resolvable tickets by itself. 

The distinguishing feature of genuine agentic systems is their ability to initiate from user interaction, independently determine task parameters, and autonomously execute complex sequences while leveraging available tools - all while maintaining alignment with enterprise objectives. 

Though it is very easy to point out the flaws in the agentic models now, think about the progress made with image and video generation, for example, over the past year. We now have dogs having a podcast on mountains! So, there’s a lot of scope for experimentation to build agents which can be a part of a multi-agent systems with multiple components. 

Building Agents for Enterprise Planning 

Can you imagine doing business without Excel or even laptops? AI is going be the new omnipresent presence in the business environment. But, to enable this AI should be able to integrate many other tools to bring sufficient capabilities (Example: If the LLM is bad at math, give it access to a tool like calculator). Many model providers already support tool use with their models, a feature often called function calling. 

We’ve been experimenting with Gen AI and Agents for quite some time, and we’ve seen huge benefits in terms of accessibility for users specially for extracting & analysing information even with Image Character Recognition (ICR). We’re continuing this journey to experiment with how such agents can help with Enterprise planning.  

The key use cases we’ve been working on are to enable enterprises to plan more effectively. Some of the agents that we’re looking at currently (some in the build and some in testing are) 

  • Supply chain agents that can automatically adjust inventory levels based on real-time demand signals 

  • Workforce planning agents that can dynamically allocated workloads and resources 

  • Finance compliance agents that assess risk in real time and respond to threats and anomalies 

  • Other alternative: Finance agents that are capable of giving personalized financial advice with personalized offers 

A diagram of a process

Description automatically generated 

<Automated agent workflow> 

The ultimate vision extends beyond isolated agents to integrated systems capable of end-to-end business process optimization. Consider an agent that not only detects demand pattern variations but autonomously adjusts S&OP plans while maintaining cross-functional alignment. 

Strategic Considerations for the Future 

When considering the future of AI, there are two major challenges to be mindful of: 

  • System reliability with task complexity: As tasks grow more complex, AI systems tend to become less reliable, leading to issues like hallucinations (false or misleading outputs) and incorrect responses. 

  • Integration of the agent-sub agent framework: While giving AI agents access to more tools enhances their capabilities, it also makes the integration of various agents (and sub-agents) more complicated. 

These challenges highlight the importance of failsafe mechanisms and tight access control to ensure that systems remain effective and trustworthy. 

Despite these hurdles, we’re in the early stages of AI-powered agents, with no established frameworks for defining, developing, or evaluating them. All of this is uncharted territory, and the future transformation of enterprises from AI-capable to AI-driven will largely be shaped by agentic systems. The key to staying ahead? Being proactive, adapting, and engaging with these systems—simply watching won’t cut it. 


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


About Polestar Insights Inc. Founded in 2012 by Chetan Alsisaria (CEO & Co-Founder), Amit Alsisaria (COO & Co-Founder) and Ajay Goenka (CFO & Co-Founder), Polestar, is a leading AI & analytics solutions company that serves Fortune 1000 companies, startups and the government across various industries, including CPG & retail, manufacturing and pharmaceuticals, among others. Headquartered in Dallas, Texas, the company enables businesses across North Americas, Asia Pacific, ANZ, and the UK with analytics foundation, data science and AI initiatives, offering a comprehensive range of services to help succeed with their data. For more information visit: https://www.polestarllp.com/

© Copyright nasscom. All Rights Reserved.