At any moment, the bar for what counts as good can be reset. Not by your competitors, but by the best experience any customer has had, anywhere. A seamless retail checkout reshapes what people expect from their bank. A one-tap rideshare reshapes what they expect from their insurer. And when an experience falls short of that bar, loyalty doesn’t erode slowly. It disappears.
That pressure is a big part of why so many organizations are racing to add AI agents into their customer experience (CX). But something bigger than another wave of automation is underway. Every major category of enterprise software – CRM, service management, HR, finance, marketing, cloud platforms – is becoming agentic. This isn’t a future state. It’s happening now. Gartner® predicts that “by 2028, 80% of organizations will see AI agents consume the majority of their APIs rather than human developers, up from less than 20% in 2026”¹.
This moves the industry beyond a shift in who’s calling an API. It’s a shift in what enterprise systems are built to serve. Agents are becoming a new class of consumer for the tools, data and workflows that used to be reserved for people to use.
The question that matters
Faced with this seismic shift, many organizations ask a version of the same question: whose agentic AI should we use?
It’s understandable. But it doesn’t go far enough.
In practice, you won’t use one agent. You’ll use many – built by different vendors, deployed by different teams, running on different data and governed by different rules.
A typical brand supports a handful to hundreds of distinct customer journeys: order status, billing disputes, returns, renewals, technical support and more. Specialized agents may handle some or all those journeys well. But no single agent or provider can handle every step of every journey, and increasingly, multiple agents and platforms need to work together to reach an outcome. Build those agents on different technology, with no central orchestrator tying them together, and fragmentation becomes its own problem — both across journeys and within them. This can result in inconsistent handoffs, duplicated context and no shared view of the customer.
A customer’s issue doesn’t respect those boundaries either. They don’t care whether their problem touches your CRM, your billing system, your knowledge base or a human expert. They just want it solved — completely and fast.
That gap, between how agents are being deployed and how customer journeys unfold, is where most agentic AI investments quietly underdeliver. And the pressure to close it is intensifying: A Gartner survey finds ninety-one percent of service and support leaders surveyed reported pressure from executive leadership to implement AI – marking a sharp increase in urgency for AI‑enabled transformation². At the same time, IBM reports that 77% of organizations say AI adoption is already outpacing their governance capabilities, and only 11% feel fully ready for the scale of agent deployment expected in the next year³.
Put simply: AI agents will continue to show up across the business. The question that matters: what will make them work together?
The answer to this question is less about choosing the perfect agent and more about building and orchestrating the right ecosystem around a new operating model designed to run it.
An agentic ecosystem is the connected network of AI agents, tools, resources and data that work together to turn customer intent into an outcome. Left alone, that network is chaotic: agents that can’t find each other, tools no agent is allowed to use, context and memory that don’t carry across interactions, and no consistent way for an agent to know what the customer has already said or another agent has done. And customers feel it.
The organizations getting this right are building around a new kind of platform – one designed to become the gravitational center of the CX ecosystem. By staying close to where intent is expressed and maintaining the context and memory to understand and carry it forward, the platform can move the enterprise quickly from intent to governed outcome. That matters even more once multiple agents are involved. Something has to attest that the intended outcome was achieved — not just that each agent along the way reported success.
Three qualities tend to separate a platform that can do this from one that can’t:
- It works in real time. CX happens in the moment. A platform must capture signals and context as they happen, not minutes or hours later.
- It’s open by design. No single agent or vendor can cover every use case, so the platform must connect broadly to the systems an organization already runs today and the specialized agents and providers it adds tomorrow.
- It’s built for trust. As agents gain autonomy, governance becomes a competitive advantage that enables scale, rather than a compliance checkbox.
It’s a platform that connects, orchestrates and governs AI agents, tools, resources and data to deliver memorable customer experiences – the kind that raise the bar and become everyone’s new normal.
Agents shouldn’t be limited to one application, workflow or system. Emerging standards make that possible. Agent-to-agent (A2A) protocols let an agent delegate work to another trusted agent instead of trying to do everything itself. Model Context Protocol (MCP) gives agents a standard way to access approved tools and resources, instead of requiring a custom integration for every new use case.
Underneath both, agents need access to data and services spread across the rest of the enterprise — systems that hold and manage information, including:
- CRM, ERP, HR, ITSM and data platforms
- Tools and partners that help drive outcomes, including marketing automation, ticketing, and knowledge and content management
- The AI layer itself, including specialized agents, voice and speech services, and the foundation models underneath them
Connection alone doesn’t create value. Orchestration is what turns access to agents, tools, resources and data into a specific customer or employee outcome.
That’s what transforms individual services into a unified system. Agents can work across workflows, reuse tools and data that already exist, and build on capabilities other teams have already deployed, rather than reinventing them. Done well, this creates a reusable foundation – one where the next use case is faster to build than the last. No more starting over.
None of this scales without governance, which can’t be an afterthought.
As agents gain more autonomy, organizations need control over what can be discovered, accessed or invoked, and by whom. They need auditability and observability that spans every system a workflow touches, so activity can be monitored, policy enforced and accountability maintained even when the workflow crosses vendors.
This takes the shape of a control layer: a place where organizations define goals, policies and guardrails, and design how AI is allowed to behave. It connects models to enterprise systems. And it monitors performance, ensures observability and explainability, and meets compliance needs across the entire ecosystem, not just one corner of it.
Connect, orchestrate and govern aren’t isolated actions. They’re interdependent capabilities of an agentic orchestration platform, and their value comes from working in unison. A common, governed access layer lets enterprises move agentic CX workflows toward meaningful outcomes with far more consistency and scale than a patchwork of point solutions ever could.
This is also what separates organizations further along the maturity curve from those still experimenting. Leading organizations are the ones whose agents, tools and workflows are orchestrated around outcomes – with the governance needed to do it safely and to prove it.
Picture a customer asking an agentic virtual agent to expedite a delayed order:
- The agentic virtual agent authenticates the customer, confirms the intent and hands the task to a trusted order-management agent elsewhere in the ecosystem using an A2A connection instead of a custom build.
- The order-management agent authenticates the agentic virtual agent and then proceeds to check carrier and inventory data, determine the earliest available delivery date and return a structured result.
- The agentic virtual agent presents the option to the customer, confirms they’d like to proceed, and the order-management agent creates the case and updates the order status.
- Finally, the agentic virtual agent confirms the outcome with the customer and provides a case number for tracking.
From the customer’s side, it’s one seamless conversation. Behind the scenes, it’s an ecosystem of agents orchestrating around a single customer’s intent, governed the entire way through.
The shift to an agentic ecosystem is already underway inside every major enterprise software category, whether or not organizations deliberately engage with it.
The starting point isn’t picking the “best” agent. It’s mapping the systems and data your agents will need to execute real workflows, and choosing a platform built to connect, orchestrate and govern services within your CX ecosystem on behalf of customers and employees. Organizations that get this right will have an agentic ecosystem that works together to create and keep loyal customers, and the foundation to scale safely with confidence.
If you’re attending Xperience this year, look for the breakout sessions and keynote moments on agentic ecosystems. And consider scheduling an Agentic eXperience Lab workshop to start building your first governed agentic workflow.
¹Gartner, “CX Product Leaders Must Adopt A2A to Support Composable Multiagent Experiences”, Manoj Bhatia, 10 August 2026
²Gartner press release, Gartner Survey Finds 91% of Customer Service Leaders Under Pressure to Implement AI in 2026, February 18, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-02-18-gartner-survey-finds-ninety-one-percent-of-customer-service-leaders-under-pressure-to-implement-ai-in-2026
GARTNER is a trademark of Gartner, Inc. and/or its affiliates
³IBM press release, New IBM Study Finds CIOs and CTOs Face Growing AI Control Gap as Enterprise Deployment Scales, June 8, 2026
