Poulomi Chatterjee: Now that AI agents have become the next buzzword, we see instances where companies say they’ve built an AI agent when it’s just Retrieval Augmented Generation (RAG). What is the difference between the two?

Ankur Kothari : As the landscape of automation and artificial intelligence evolves, “AI Agents” and “RAG” (Retrieval-Augmented Generation) are integral to enhancing business processes. They serve distinct purposes and offer different capabilities, such as: 

 Functionality

AI agents are designed for autonomous task execution and user interaction, while RAG focuses on enhancing response accuracy through information retrieval. 

Learning and Adaptation

AI agents can learn and adapt over time, improving their performance based on user interactions. RAG systems, while they can generate responses, do not inherently learn from interactions but rather rely on the quality and relevance of the retrieved data.  Use Cases

AI agents are ideal for scenarios requiring decision-making and complex interactions, such as customer service or process automation. RAG is better suited for tasks that involve information retrieval, such as answering specific queries or generating content based on factual data. 

At Automation Anywhere, our AI agents leverage a native RAG service to ground models with enterprise data, seamlessly integrating with platforms like AWS Bedrock, Azure AI, and Google Vertex. This helps businesses connect the LLM models to the data, to ensure they have the required context before arriving to a conclusion. This mitigates hallucinations, enriches context, and ensures accurate, data-driven insights. By combining AI agents with RAG, we empower businesses to thrive in an increasingly automated world. 

PC: AI agents are also poised to become a hotly contested area. What gives you the advantage over others? 

Ankur:  Most companies are building AI or AI agents that improve their products, helping individual users increase efficiency by 10-15%. We believe our competitive advantage in the realm of AI agents stems from several key factors that set us apart from others in the industry: 

AI + Automation Enterprise System

Our intelligent automation solutions streamline workflows, enhance decision-making, and drive growth, empowering organisations to unlock their full potential in an increasingly competitive landscape.  

Scalability and Flexibility

Automation Anywhere’s AI agents are scalable, adaptable, and versatile, automating tasks from routine to complex, tailored to meet the unique needs of any industry or organization. 

User-Centric Design

We prioritise user experience in our AI agent development. Our agents are built to interact naturally with users, providing intuitive and engaging experiences. This focus on user-centric design ensures that our AI agents can effectively understand and respond to user needs, leading to higher satisfaction and adoption rates.

Continuous Learning and Improvement

Our AI agents are equipped with machine learning capabilities that enable them to learn from interactions and improve over time. This ensures that our agents become more effective and accurate in their responses, adapting to changing business environments and user expectations.

Comprehensive Support and Resources

We offer extensive resources, training, and support to help organisations successfully implement and manage AI agents. Our commitment to empowering our customers ensures they can fully leverage the potential of AI agents, driving greater business outcomes.

Focus on Security and Compliance

Our AI agents are developed with built-in security features and compliance measures. This focus on safeguarding sensitive information provides organisations with peace of mind as they adopt AI solutions. 

By combining these strengths, Automation Anywhere is well-positioned to lead the charge in the competitive landscape of AI agents.

PC: AI agents have revolutionised Automation Anywhere’s core from Robotic Process Automation (RPA) as is expected to happen with automation overall. What does this change entail?

Ankur: The evolution from Robotic Process Automation (RPA) to Agentic Process Automation (APA) signifies a major advancement in how organisations automate their processes.

RPA has traditionally focused on automating repetitive, rule-based tasks, allowing organisations to increase efficiency and reduce human error. However, as business environments become more complex, the limitations of RPA have become apparent. It primarily relies on predefined scripts and rules, which can hinder adaptability and responsiveness to change.

In contrast, APA introduces a more sophisticated approach by incorporating elements of AI/ML. This allows for the automation of not only straightforward tasks but also complex, mission-critical processes that require cognitive decision-making. APA systems can learn from data, adapt to new situations, and even generate automations autonomously, thus providing a significant enhancement in operational efficiency.

With APA, organisations can expect to see efficiency gains of 40-80%, compared to the 20-30% typically associated with RPA. With a human-in-the-loop approach, APA ensures automation remains autonomous, yet guided by human oversight.

PC: The age-old question around what will the impact of automation be on jobs, still remains.

Ankur: The integration of AI with automation is poised to transform industries and redefine job roles, shifting the focus from repetitive tasks to more meaningful and creative work. AI and automation are expected to handle mundane aspects of jobs, enabling workers to channel their efforts towards innovation and complex problem-solving. It simultaneously prepares the workforce for more advanced roles, fostering long-term growth and adaptability. 

It is integral to realise the importance of skill development while keeping a positive mindset that sees AI as an augmentation tool rather than a threat. The transition is more about embracing a mental shift than acquiring new skills. With responsible deployment and a commitment to upskilling, organisations can unlock AI’s full potential, driving faster, smarter, and more efficient operations while fostering innovation and growth in the workplace. 

PC: Which sectors will be disrupted most by AI agents?  

Ankur: The sectors that will likely experience the most significant disruption due to AI agents include healthcare, financial services, retail, manufacturing, logistics and supply chain, and customer service.

In healthcare, they can streamline administrative tasks and assist with diagnostics. The financial sector will see automation in fraud detection and personalised customer service. Retail will benefit from optimised inventory management and personalised shopping experiences. Manufacturing can leverage AI for predictive maintenance and quality control. Logistics will improve efficiency with route optimisation, and customer service will see enhanced interactions through automation.

Each sector stands to gain from the operational efficiencies and cost reductions brought by AI agents. 

PC: When they were initially introduced, AI agents were flawed and expensive. How have they improved over time? 

Adi Kuruganti: AI agents have undergone significant improvements since their inception, evolving from flawed and compute-heavy systems to more sophisticated, efficient, and effective solutions. While Large Language Models (LLMs) historically required significant compute power, making them costly to create and prone to hallucinations—particularly in enterprise scenarios—AI Agents offer a more practical, action-oriented solution. AI Agents combine AI Skills with Actions, allowing customers to select from a variety of LLMs, including foundational, custom, and pre-configured models from Automation Anywhere (AAI), tailored for specific use cases.

AI Skills are further tested using real automation data and governed by a robust AI Governance layer that ensures real-time auditability, traceability, and PII (Personal Identifiable Information) masking. These high-performing AI skills seamlessly connect to automation actions like RPA, API, or document processing, delivering exceptional enterprise value. Importantly, AI Agents are not limited to AI/ML or deep learning, which pertain primarily to model architecture. Advances in smaller models (e.g., LLaMA) and techniques like Compound AI Systems have dramatically reduced compute costs—by up to 10,000%—further enhancing accessibility and efficacy, solidifying AI Agents as a transformative solution for enterprise automation. 

Published - December 27, 2024 04:43 pm IST