AI and ML
From assistants to agents
In part two of our Agentic AI article series, we’ll explore some of the ways that change will manifest and help you reimagine how humans and machines can collaborate across a huge range of business contexts.
Of all the possible use cases for agentic AI, some with the highest potential to deliver significant value are in customer experience (CX). In a field where automated experiences have seen mixed success, agentic AI could close the gap between automated and human-assisted interactions, bringing qualities like hyper-personalization to automated elements of the customer experience.
But agentic CX only works when organizations keep customer data, systems and processes accessible, connected, structured and governed. Without that foundation, agents cannot act reliably across a journey. They can only respond to whatever is visible at a single touchpoint.
Dynamic personalization is just the beginning
The most obvious use case for agentic AI in CX is dynamic personalization. When someone interacts with an AI agent, they can receive personalized responses, experiences and recommendations.
But agentic AI can do much more than just interact with customers; its true potential in CX lies in orchestrating and accelerating key processes in customer journeys. It’s a shift from an AI world focused on automating processes to one where the focus is on orchestrating outcomes.
For example, we worked with a leading insurance firm to create an autonomous AI agent that can file a dispute for wrongly denied claims on a customer’s behalf. This accelerates and simplifies a stressful process for customers, while saving valuable time for the insurer’s employees.
The rising quality of AI-driven buying journeys is driving the biggest shift in the CX landscape. We’ve all experimented with chatbots and automated retail, for example, but often those experiences don’t deliver everything we need. Through AI agents, we can move from receiving faster, better responses to experiencing better decisions across the full arc of a journey: agents that coordinate across systems, connect context from previous interactions and act on what matters to the customer, not just what is visible in a single session. For many customers, that will make agentic buying their primary channel of choice. And that’s where the real change begins.
It’s all about choice, and that choice is still human-driven
As AI agents become a shopping and service channel in their own right, some customers will embrace them with open arms. Others will still opt to visit brick-and-mortar stores or connect with a human representative. Sometimes, people may want to browse and make the decision themselves, and sometimes they’ll want an AI agent to do it for them. It’s all about choice.
In that sense, the core tenets of CX aren’t going to change. Brands still need to facilitate great experiences through numerous channels and touchpoints, so customers have the freedom to engage with them however they please. It’s an innately human-driven process, which means the way people work and how teams organize themselves will have to change, too.
Why the composable enterprise is the right foundation for agentic CX
For agentic ecosystems to function properly and drive meaningful transformation, they can’t be kept in silos. When customer data lives in systems that cannot share context, agents cannot act coherently across a journey. Agentic technology will only fulfill its potential in a ‘composable enterprise’ that provides an adaptable platform to evolve capabilities as technology advances and customer expectations change.
In a composable enterprise, organizations break the business down into independent capabilities, data products, AI models and agents that can be reused and assembled to serve new use cases or create new capabilities. This approach involves using agents to automate and orchestrate processes, with humans in the loop to provide strategic direction and handle ambiguous exceptions. Humans also define the guardrails within which agents operate, set the intent those agents pursue and own the accountability for the outcomes they produce. That design work is not a residual function. It is what makes agentic CX trustworthy.
In a CX context, these agents plan and reason using domain-specific models trained on customer, sales and service data, leveraging APIs to execute actions in the real world.
Importantly, this approach also requires strong governance, with policies and guardrails to ensure agentic capabilities behave within specified boundaries. This is especially critical in CX, where the reputational, legal and financial impact of straying beyond these boundaries could be catastrophic.
With these foundations in place, organizations can dynamically redraw the boundaries as models and agents become smarter and more capable, as organizations refine workflows, update rules and deliberately expand or adjust agent boundaries based on what they learn. The models do not drive this evolution. Organizations drive it by intentionally redesigning how agents operate within the composable architecture.
The composable enterprise is not just the starting condition for agentic CX. It is the ongoing structure that makes continuous improvement possible: modular enough to change a workflow without rebuilding a platform, governed enough to tighten a guardrail without disrupting a journey.
