Unlocking the Agentic AI Advantage

Leo

Welcome to our podcast, where we explore the cutting edge of AI and technology. I'm your host, Leo, and today we're joined by the incredibly insightful Bella. Today, we're diving into the world of agentic AI and how it can revolutionize business processes. Bella, what are you most excited about in this discussion?

Bella

Hi, Leo! I'm really excited to learn more about the agentic AI revolution. I’ve heard a lot about the gen AI paradox, where there's widespread adoption but minimal impact. Can you explain what that means and how agentic AI can help?

Leo

Absolutely, Bella. The gen AI paradox refers to the imbalance between horizontal and vertical use cases. Horizontal use cases, like enterprise-wide copilots and chatbots, have scaled quickly but deliver diffuse, hard-to-measure gains. On the other hand, vertical use cases, which are function-specific and have the potential for significant impact, are often stuck in pilot mode. This imbalance is a major challenge, but agentic AI can help bridge that gap by integrating deeply into core business processes.

Bella

That makes sense. Can you give us some examples of horizontal and vertical use cases to better understand the difference?

Leo

Sure thing! Horizontal use cases include tools like Microsoft 365 Copilot, which can help employees with tasks like writing emails, summarizing documents, and generating meeting minutes. These tools are widely accessible and easy to implement. Vertical use cases, however, are more specific, such as a supply risk assessor that generates risk profiles for suppliers to boost sourcing resilience, or a sales assistant that prioritizes and qualifies accounts. These use cases have the potential to significantly impact revenue and operational efficiency but often face technical and organizational barriers.

Bella

I see. So, what is the role of agentic AI in transforming these vertical use cases and making them more impactful?

Leo

Agentic AI is the key to unlocking the full potential of vertical use cases. Unlike traditional AI tools, which are reactive and isolated, AI agents are proactive and goal-driven. They can understand complex goals, break them into subtasks, interact with both humans and systems, and execute actions with minimal human intervention. This means they can transform entire business processes, not just optimize isolated tasks. For example, an AI agent in a supply chain can monitor disruptions, dynamically replan transport and inventory flows, and negotiate directly with external systems, significantly improving service levels and reducing costs.

Bella

That sounds incredibly powerful! But what about the challenges? You mentioned a new architectural paradigm called the agentic AI mesh. Can you explain what that is and why it’s necessary?

Leo

Absolutely. The agentic AI mesh is a new architectural paradigm that addresses the challenges of deploying AI agents at scale. It’s a composable, distributed, and vendor-agnostic environment that enables multiple agents to reason, collaborate, and act autonomously across various systems and tools. This mesh is essential because it helps manage the new class of risks that come with agentic AI, such as uncontrolled autonomy, fragmented system access, and lack of observability. It also allows organizations to blend custom-built and off-the-shelf agents while staying agile and avoiding vendor lock-in.

Bella

Wow, that’s quite comprehensive. But I imagine the bigger challenge isn’t just technical, right? It’s also about the human and organizational aspects. How do companies ensure trust and drive adoption of these agents?

Leo

You’re absolutely right, Bella. The biggest challenge in the agentic era is human and organizational. Earning trust, driving adoption, and establishing the right governance are crucial. This involves building transparency into how agents operate, ensuring they behave predictably, and integrating them intuitively into daily workflows. It also means creating a ‘human + agent’ mindset through cultural change, targeted training, and supporting early adopters as internal champions. New roles, such as prompt engineers and agent orchestrators, will also be essential to manage agent workflows and handle exceptions.

Bella

That’s a lot to consider. How do companies start reimagining their business processes around AI agents? It seems like a significant shift from the current approach.

Leo

Indeed, it is a significant shift. In the early days of gen AI, most vertical initiatives focused on plugging a solution into a specific step of an existing process, which delivered narrow gains without changing the overall structure of how work is done. With AI agents, the paradigm shifts entirely. The opportunity now lies in transforming entire business processes by embedding agents throughout the value chain. This means moving from the question, ‘Where can I use AI in this function?’ to ‘What would this function look like if agents ran 60 percent of it?’ It involves rethinking workflows, decision logic, human–system interactions, and performance metrics across the board.

Bella

That’s a fascinating approach. But what about the practical steps? How do companies equip their workforce and introduce new roles to support this transformation?

Leo

Equipping the workforce is crucial. Companies need to foster a ‘human + agent’ mindset through cultural change and targeted training. They should introduce new roles such as prompt engineers, who refine interactions, agent orchestrators, who manage agent workflows, and human-in-the-loop designers, who handle exceptions and build trust. This involves not just technical training but also a shift in how people think about their roles and the role of AI in their daily work.

Bella

That sounds like a holistic approach. What about governance and ensuring autonomy control? How do companies manage the risks associated with AI agents acting independently?

Leo

Governance is indeed a critical aspect. Companies need to define clear frameworks that establish agent autonomy levels, decision boundaries, behavior monitoring, and audit mechanisms. Policies for development, deployment, and usage must be formalized, and classification systems should group agents by function, each with an appropriate oversight model. This ensures that agents operate safely and transparently, building trust and preventing uncontrolled sprawl.

Bella

Building a foundation for interoperability and scale sounds like another big challenge. How do companies approach this?

Leo

To build a foundation for interoperability and scale, companies need to evolve their AI architecture from LLM-centric setups to an agentic AI mesh. This involves creating a modular, distributed, and vendor-agnostic environment that allows agents to operate across a fragmented ecosystem of systems, data, and workflows. In the long term, organizations should also start preparing for a next-generation architecture where all enterprise systems are reshuffled around agents in terms of user interface, business logic, and day-to-day operations. This ensures that agents can seamlessly integrate and scale across the enterprise.

Bella

And what about data? How do companies ensure they have the quality and accessibility needed for agentic AI to be effective?

Leo

Data is the lifeblood of AI agents. Companies need to accelerate data productization and address quality gaps, especially in unstructured data. This involves transitioning from use-case-specific data pipelines to reusable data products and extending data governance to ensure that data is accessible, accurate, and up-to-date. By doing so, companies can provide the high-quality data that AI agents need to operate effectively and deliver value.

Bella

Those are