With AI in the spotlight, tech giants and startups are exploring the potential of AI agents. Yet, some companies are confusing AI assistants and chatbots with fully developed, autonomous agents. Though there is a need for clarity on this, AI agents are being adopted across industries.
Reflecting on this, Dominic Pereira, vice president of product management at software firm Automation Anywhere, cited the example of a commercial bank. In this case, a person would review a 50-page document for a loan application and analyse the terms to make a financial decision. According to him, the process can be simplified with AI agents.
“With an AI agent trained on the same 50-page document, it can access the criteria and suggest optimal loan terms for a customer, grounded in real, company-specific data,” Pereira told AIM.
To bring in more clarity, here’s a list of the AI agents that were introduced this year.
Oracle’s Miracle Agent
In September, Oracle announced its plans to introduce over 50 AI agents with its Fusion Cloud suite in sectors including finance HR, supply chain, quality control, scales, and customer service. One interesting example is Oracle’s document agent, which could streamline tasks for a sales executive by snapping a photo of an overseas vendor’s product quote, collecting relevant data, translating it, and automatically generating a purchase request. Further, it can even assist with processing vendor invoices for final payment approval.
Microsoft’s Work Recall with Copilot Vision
In October, Microsoft introduced new capabilities with its Copilot to transform business processes through autonomous agents. With Copilot Studio, users can create and manage these agents. And now, with ten new autonomous options available in Dynamics 365, it can support sales, service finance, and supply chains.
According to Microsoft, these agents are like the new “apps” for an AI-driven world, autonomously executing tasks and managing business functions on behalf of individuals, teams, and departments.
Anthropic’s Human-Like Experience
Anthropic’s latest model, Claude 3.5, includes features like ‘computer use’ that enable its AI to perform human-like tasks on a computer screen. It operates like navigating websites by entering text, selecting buttons, and browsing in real time. This additional capability enables the AI to interact with software just like a human user would.
This new update was released in October in public beta for developers. The team hopes to expand access to consumers and enterprise clients over the next few months or early 2025.
Google’s ‘Iron Man’ Inspiration
Google also plans to develop an AI agent named Project Jarvis, a clear nod to Iron Man’s famous assistant. This AI is designed to navigate a user’s web browser autonomously and carry out tasks like researching, purchasing products, or booking flights.
Autonomous Agents in Real-World Organisations
Not only tech giants but also major organisations are developing custom agents to address specific challenges. For example, UK pet retailer Pets at Home built an AI agent to aid its profit protection team, which could yield seven-figure annual savings.
McKinsey & Company has developed an agent to speed up client onboarding, showing a potential 90% reduction in lead time and a 30% cut in administration work. Meanwhile, Thomson Reuters has launched an agent to streamline the legal due diligence process, cutting task time by half and helping drive new business opportunities.
AI Agent Framework
Many companies are working to advance the AI field by offering open-source frameworks. OpenAI’s experimental framework, Swarm, allows developers to build networks of cooperative AI agents that can tackle complex issues with minimal human intervention.
Similarly, Google Cloud’s Vertex AI Agent Builder consolidates tools and resources for developers aiming to create custom AI agents. At the same time, IBM’s watsonx.ai platform provides enterprise-grade tools to support specialised, scalable agent development.
Another interesting framework is CrewAI, which aims to orchestrate role-playing AI agents. It allows developers to create a “crew” of AI agents, each assigned with a specific task, to work together on complex activities. This framework is particularly useful for building collaborative AI systems that can address multifaceted problems requiring diverse expertise and coordinated support.
Microsoft’s Magentic-One is another such framework that aims to assist developers and researchers in creating agentic applications that manage complex and multi-step tasks autonomously.
The post Top AI Agents of 2024 appeared first on Analytics India Magazine.
