The Rise of AI Agents - Part 1 : Everyone talks about it, but are we all talking about the same thing?

Preliminary note: if you prefer to consume this article in a podcast format, I hired two synthetic co-hosts to discuss it :

?? Spotify ?: https://tinyurl.com/y57w7yv9

?? Apple ???: https://tinyurl.com/2xtt9zk3

The Agentic AI Revolution

In an era where technology evolves at breakneck speed, a new force is poised to redefine enterprise operations: Agentic AI. This revolutionary approach to artificial intelligence isn't just capturing the attention of tech enthusiasts; it's becoming a central focus for major industry players who recognize its potential to transform business operations fundamentally.

The emergence of Agentic AI is no accident. It's the culmination of decades of AI research, exponential growth in computational power, and the maturation of machine learning algorithms. As we stand on the brink of this new era, companies like Salesforce, SAP, ServiceNow, Accenture, and Cognizant are at the forefront, embracing AI agents to revolutionize their products and services.

Consider these groundbreaking developments:

• Salesforce, in collaboration with NVIDIA, has introduced Agentforce, a platform designed to leverage AI agents for enhancing productivity across various business functions.
• SAP introduced collaborative AI agents through its AI copilot, Joule, which includes multiple autonomous AI agents, each an expert in a particular function, collaborating to execute complex workflows..
• ServiceNow is pushing the boundaries of 24/7 collaboration with AI agents across multiple use cases, from IT to human resources.
• Accenture announced its own AI Refinery, an agentic AI system available on all public and private cloud platforms that can act on user intent, create new workflows, and take appropriate actions based on the environment.
• Cognizant has enhanced its Neuro AI platform with multi-agent AI orchestration features, aiming to boost AI-driven productivity and business growth by leveraging synthetic or anonymized data and industry-specific configurations.
• Microsoft introduced AI agents in its Microsoft 365 Copilot product, allowing users to create AI assistants that can carry out tasks across Microsoft and third-party software. To make it even easier to build Copilot agents, they announced agent builder, a new, simplified experience powered by Copilot Studio.
• Workday unveiled its next-generation AI platform, Illuminate, along with four new AI agents: Recruiter, Expenses, Succession, and Optimize. These agents are designed to automate tasks such as candidate sourcing, expense reporting, succession planning, and process optimization
• Oracle announced a new group of AI agents aimed at extending generative AI capabilities into more fully realized and automated processes. These agents are part of the next generation of AI technology in Oracle's application suite
• E42.ai reported that its multifunctional AI workers have automated over 200 processes, saved 200,000 man-hours, and resolved 3 million queries. These AI agents are used in various areas such as accounts payable and IT operations management.

And I could go on and on.

No wonder since the projected growth rates for agentic AI indicate a significant and rapide expansion:

According to Capgemini, 10% of organizations already use AI agents, with 82% planning integration within the next three years.

According to Acute Market Reports, the agentic AI market is expected to grow at a Compound Annual Growth Rate (CAGR) of 40.2% from 2024 to 2032.

Emergen Research forecasts a CAGR of 31.68% for the agentic AI market with the market size expected to reach USD 367.68 billion by 2033.

Amazon’s CEO Andy Jassy recently mentioned that their AI coding assistant, Amazon Q, has saved the company $260 million and 4,500 developer years. Meta CEO Mark Zuckerberg believes there could eventually be more AI agents than people in the world.

These advancements signal more than just incremental progress; they represent a fundamental shift in how enterprise software operates. Unlike traditional AI systems that respond to specific queries or perform predefined tasks, agentic AI systems can autonomously plan, reason, and take actions to achieve complex goals. This leap forward is poised to redefine efficiency, innovation, and competitive advantage in the business world.

Should You Be Sceptical?

While the potential of AI agents is undeniably exciting, a degree of scepticism is natural. The enterprise software landscape has seen its fair share of revolutionary promises that failed to fully materialize. Seasoned IT professionals like me will recall similar enthusiasm surrounding other technological paradigms:

• Service-Oriented Architecture (SOA) was once hailed as the ultimate solution for system integration and flexibility. Yet, many organizations still grapple with monolithic systems, finding the promise of seamless interoperability elusive.
• The initial fervor around Machine Learning (ML) led many to believe it would swiftly solve complex business problems, but widespread, transformative implementations remain a challenge for many companies.
• Even Big Data, despite its significant impact, hasn't delivered on all its early promises, with many organizations struggling to extract meaningful value from their data lakes.

However, AI agents may represent a significant leap forward, addressing many of the limitations that hampered previous technologies:

1. Advanced Natural Language Processing (NLP): Unlike earlier systems, AI agents can understand and generate human-like text, enabling more intuitive interactions and complex task execution.
2. Contextual Understanding: AI agents can grasp nuanced context, allowing them to adapt to different situations more effectively than rule-based systems.
3. Continuous Learning: These agents improve over time through experience, overcoming the static nature of many previous technologies.
4. Seamless Integration: Advanced APIs and microservices architectures make it easier for AI agents to integrate with existing systems, addressing a key challenge of past innovations.

While scepticism is healthy, the tangible successes and technological maturity of AI agents warrant cautious optimism. As Dr. Andrew Ng, co-founder of Google Brain and former chief scientist at Baidu, notes, "AI is the new electricity. Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don't think AI will transform in the next several years."

What is for sure is that at this stage there may be more questions than definite answers on the topic.

Are all these players talking about the same thing?

Understanding AI Agents

To appreciate the true potential of AI agents, we must understand what sets them apart from previous technologies. At their core, AI agents are autonomous software entities that combine the power of large language models (LLMs) with the ability to interact with their environment, make decisions, and take actions.

Think of AI agents as highly skilled digital assistants. Unlike traditional software that follows predetermined rules, these agents can understand complex requests, adapt to new situations, and even anticipate needs before they’re explicitly stated.

Key Capabilities of AI Agents:

1. Perception and Interpretation: AI agents can "read" and understand complex environments, not just in terms of data analysis but also grasping context and nuance.
2. Abstract Reasoning: They can think about concepts and solve multi-step problems, mirroring human cognitive processes.
3. Planning and Execution: AI agents devise strategies and carry out sequences of actions to achieve goals.
4. Adaptive Learning: Through experience and feedback, these agents continuously improve their performance.
5. Collaboration: AI agents excel in working alongside humans and other AI systems, unlocking new possibilities for problem-solving and process automation.

The advent of Generative AI (GenAI) and Agentic AI represents a quantum leap in these capabilities. GenAI creates new content, while Agentic AI adds autonomous decision-making and action-taking to the mix. This combination is pushing the boundaries of enterprise software and business processes.

Dr. Fei-Fei Li, Professor of Computer Science at Stanford University, emphasizes: “AI agents are powerful tools that can automate complex tasks that were once thought to require human-level cognition, but they are only as good as the data they are trained on and the systems they are integrated with.” This underscores the potential and limitations of AI agents, as their effectiveness relies on the quality of their training data and integration into business systems.

What is Common Across AI Agent Announcements?

The recent wave of announcements from companies like Salesforce, SAP, Microsoft, and others highlights how AI agents are transforming the business landscape. While these announcements vary by industry and use case, they share several common elements:

1. Autonomy: All AI agents operate independently, requiring minimal human intervention. They perform tasks like customer service automation, data analysis, and decision-making across various business functions
2. Task Execution: AI agents are deployed to handle a range of tasks, from repetitive administrative work to more complex problem-solving, enhancing productivity and efficiency
3. Contextual Understanding: By leveraging natural language processing (NLP) and machine learning, these agents interpret and act based on contextual data, enabling more dynamic and human-like interactions
4. Integration with Data Systems: The effectiveness of AI agents depends heavily on their ability to access and process vast amounts of enterprise data, both structured and unstructured, to provide real-time insights and solutions

Variations in Approaches:

• Domain-Specific vs. General-Purpose: Some platforms, like Workday’s Illuminate, focus on specific business areas such as HR or finance, whereas On the other hand, Salesforce Agentforce and ServiceNow AI agents offer broader applications across customer service, sales, and IT support, but may not delve as deeply into niche tasks as domain-specific agents. Another company, Fractal and their Genesis platform offers both general-purpose and specialized agents for broader adaptability: their agents are designed to augment human decision-making in areas like market research, finance, and supply chain management. Fractal's agents are more specialized and built for scalability and flexibility. Their general-purpose agents handle broader tasks, while task-specific agents are designed for detailed, niche business functions. The platform emphasizes the integration of GenAI for more sophisticated, tailored solution
• Complexity of Tasks: Accenture’s AIRefinery and Oracle’s AI agents focus on complex process automation across industries, such as supply chain optimization, by integrating with large-scale enterprise systems. These agents are designed to handle sophisticated workflows and decision-making; In contrast, Microsoft’s Copilot is centered on automating routine tasks, such as drafting emails or preparing presentations, which are typically less complex than those addressed by enterprise-level solutions like Accenture's
• Collaboration vs. Task Specialization: SAP’s Joule excels in collaborative agents working across multiple business functions, while companies like E42.ai and Supervity.ai build agents that can automate specialized processes like IT management or finance
• Human-Like Interaction: Companies like Salesforce, in collaboration with NVIDIA, focus on creating AI avatars that provide human-like experiences, enhancing the interactive nature of these agents; other platforms, like SAP’s Joule, are more focused on functional collaboration between agents, rather than making them feel "human"

Where we need to be clear and careful, is that AI Agent does not equal Gen AI: the LLM part usually only represents about 20% of what makes a real AI Agent. Some companies tend to talk about Agents as soon as they have adopted a GenAI powered Chatbot as the customer interface to their application. This is only the visible part but far from being the most important.

Conclusion

The agentic AI revolution is only starting and some of the Big Tech claims are certainly ovestated as marketing goes faster than developments. "Fake it until you can actually make it", as we say. But the direction is clear and it represents more than a technological advancement—it's a fundamental shift in how enterprises will operate, think, and innovate. While questions remain about implementation and integration, the convergence of major players and rapid technological progress suggests we're not just witnessing another tech trend, but the dawn of a new era in enterprise computing. Those who embrace this transformation thoughtfully and strategically will likely find themselves at the forefront of the next great leap in business evolution.

In part 2 we will go deeper into the very nature of this paradigm shift in Enterprise Systems.