At its annual customer confab Zoomtopia, Zoom made a boatload of announcements. Seriously. Across its broad product suite, platform, and its partnerships, I counted 62 distinct announcements and significant enhancements. So, it has taken a little while for me to digest all that news and figure out the critical bits for an Opus Research audience. Sorry to the avid fans of the collaboration, conferencing, and events products—I’m going to focus on the company’s booming CX business here.

To me, the biggest CX story coming out of Zoomtopia wasn’t a new virtual agent, another AI-powered feature, or even a contact center enhancement (although, again, plenty of those to go around). The thing that really stuck with me is something called the Unified Context Layer. Maybe not the sexiest product name, but potentially the most consequential announcement Zoom made.

Consider how much customer information is locked away in conversations. A customer calls about a delivery problem, discusses a workaround with an agent, and agrees to a follow-up. The CRM system might capture a case number, a disposition, and perhaps some notes. But the circumstances behind the problem, the customer’s frustration, and the promises made during the conversation often never find their way into a system that can put them to work.

Zoom is angling to change that. Its Unified Context Layer combines three sources of information: static enterprise data from systems like CRM and knowledge bases; general knowledge available to AI models; and dynamic context drawn from conversations happening across the organization. That last piece is particularly interesting.

Imagine a customer calls an insurance company on Monday to discuss a delayed claim. During the conversation, the agent promises an update by Thursday. On Wednesday, a claims manager discusses the case in an internal videoconference and explains that the delay is caused by a problem affecting hundreds of customers. As a result, the company decides to waive certain documentation requirements to speed things up. But when the customer calls again on Thursday, the next agent knows nothing about that decision and starts asking for the very documents the company has agreed are no longer necessary.

That’s the sort of disconnect Zoom’s Unified Context Layer is designed to address. In Zoom’s vision of the world, the decision made during Wednesday’s meeting would become part of the context available to the AI agent handling Thursday’s call. The customer wouldn’t have to explain everything again, and the agent wouldn’t be working from Wednesday’s outdated information.

At the heart of Zoom’s Unified Context Layer is a dynamic knowledge graph that is designed to track people, topics, decisions, commitments, and the relationships among them in near real time. Rather than continually feeding transcripts into an LLM and hoping it finds the relevant bits, the platform builds a contextual picture that can be updated as new conversations occur. An AI agent can potentially understand what happened this morning, what was agreed last week, and how those decisions affect the customer (virtually) standing in front of it now.

From understanding to doing

Of course, having better context is only half the story. Zoom’s broader ambition, summarized well by its “conversation to completion” positioning (a theme we heard repeatedly), is to use that intelligence to get work done.

Several Zoomtopia announcements support this approach. Dynamic Customer Profiles pull together CRM records, interaction histories, and conversational intelligence. CX Insights identifies emerging customer problems across interactions. Customer Workflow Orchestrator connects those insights to actions across systems and departments. And Zoom is extending its virtual agents into physical locations through kiosks, bringing in-person customer interactions into that same contextual picture.

Individually, these are likely useful capabilities. Together, however, they point toward something more interesting: a CX operating model where information gathered in one interaction can help resolve problems elsewhere, perhaps before another customer even calls.

Zoom demonstrated this with a fictional hardware company experiencing a product defect. Customer conversations exposed the problem, analytics helped identify its cause, and specialized AI agents coordinated updates across product management, support, sales, and marketing. It was a tidy demo, naturally. Real enterprises rarely cooperate quite so neatly.

The context challenge

There are still some substantial questions to answer. How much of an enterprise’s conversational context can Zoom capture when interactions happen outside its environment? How does it reconcile conflicting information or outdated decisions? Who determines which agents can access particular conversations, and how is that access governed across applications?

Zoom’s emphasis on MCP, external AI assistants, and an open architecture is encouraging. But openness creates its own requirements for identity, permissions, policy enforcement, and oversight.

As I argued in an Opus Research blog, enterprises will need an AI agent control plane to manage shared customer context, orchestration, and governance across an increasingly fragmented collection of agents and systems. Zoom’s Unified Context Layer could become an important component of that architecture.

What really seems intriguing about all of this is that the conversations enterprises have spent decades recording and analyzing could finally become an active ingredient in how customer service gets delivered. Those conversations go from compliance necessities and coaching fodder to the fuel for actually getting things done. Zoom has laid out an ambitious plan to make that happen; now it must prove it can make all those pieces work together in the real world.

Categories: Conversational Intelligence, Articles

Tags: AI agent control plane, AI agents, Context, conversation intelligence, Zoom