AI Voice Is Ready. What We Learned About Fixing the Data Layer

This week’s The Inner Circle session was one of those discussions where you feel the industry quietly shifting.

A small group of CPaaSAA members—operators, vendors, analysts—came together to go deeper on a topic that’s still early, but increasingly unavoidable.

The session was led by Amandeep Khanuja (QKS Group), building on their ongoing research into conversational data and enterprise AI.

The central question was simple:

If AI Voice is accelerating… why is it still so hard to scale?

The answer, once again, comes back to something much less exciting than AI itself:

the data layer.

Conversations Are No Longer Just Interactions

One of the strongest themes—grounded in QKS research—was how conversations have fundamentally changed role.

They’re no longer just moments between a business and a customer. Today, they carry intent, context, operational signals, and even decision history. They don’t just reflect what happened—they shape what happens next.

In that sense, conversations are becoming the nervous system of the enterprise.

They influence analytics, compliance, automation, and increasingly, the performance of AI itself. And once you start looking at them that way, the importance of how they are structured—and preserved—becomes obvious.

The Real Problem Isn’t AI. It’s Fragmentation.

Across industries, the pattern is remarkably consistent.

Voice sits in one system. Chat in another. CRM and contact center platforms each maintain their own version of the truth. AI models are then expected to make sense of all this—often with incomplete or inconsistent data.

The result is not a lack of data, but a lack of usable data.

Enterprises are generating enormous volumes of conversation data every day. But they struggle to move it, connect it, and reuse it across systems. Migration becomes painful. Governance becomes unclear. And AI initiatives fail to deliver the expected return.

Not because the ambition is wrong—but because the foundation is weak.

Why vCons Are Getting Attention

This is where the discussion moved into vCons (virtualized conversations).

The simplest way to understand them is this: a vCon is like a “PDF for conversations.”

Not a new platform. Not another tool. But a way to package a conversation—across voice, chat, or video—together with its metadata, analysis, and business context, into a single structured object.

That might sound like a small shift. It isn’t.

Because it changes how conversations can be stored, moved, and used across the enterprise.

From Fragmented Data to a Usable Asset

Once conversations are treated as structured data objects, a number of things start to fall into place.

AI becomes more reliable, simply because it is working with more complete and consistent input. Migration becomes less painful, because data can move with integrity across platforms. Compliance becomes easier to manage, as consent and lineage travel with the conversation itself. And integration becomes more scalable, reducing the need for constant custom mapping between systems.

This is where the discussion became very practical.

Because this isn’t theoretical. It’s already happening.

What Happens When You Actually Use This

One example shared during the session stood out.

When organizations begin to analyze conversations in a structured way, entirely new insights emerge. Churn signals become visible in real time. Follow-ups that were promised but never executed suddenly surface. Compliance gaps appear much earlier in the process.

And perhaps the most striking observation was this:

Once you see the data at that level, you can’t unsee it.

That’s the moment where this stops being interesting—and starts becoming essential.

The Mistake the Industry Keeps Making

There was also a very honest reflection during the discussion.

Too many players are trying to position vCons as a technology to be sold.

But that misses the point.

No one buys a format. They buy outcomes.

vCons are not the product. They are the enabler. The real value lies in what becomes possible once the data layer is fixed.

Why This Matters Now

Timing matters.

AI adoption is accelerating quickly, while regulation is becoming stricter. Enterprises are rethinking data ownership and becoming more aware of the risks of platform lock-in.

At the same time, a large percentage of AI initiatives still struggle to meet expectations. Not because the models are insufficient, but because the data feeding them is incomplete, fragmented, or poorly structured.

That gap is becoming harder to ignore.

The Bigger Picture

What became clear in this Inner Circle session is that this is not just about voice, or contact centers, or even AI alone.

It’s about how enterprises treat one of their most valuable assets: conversations.

Whether they remain fragmented by-products of systems, or become a durable, portable data layer that can support the next generation of applications.

That choice will define a lot of what comes next.

What Comes Next

vCons are still early. Awareness is limited, adoption is just beginning, and standards are still evolving.

But the direction—highlighted both in this session and in ongoing QKS research—is becoming clearer.

The next phase of AI in communications will not be won by better models alone, or by adding more features.

It will be won by those who fix the foundation.

Inside CPaaSAA

These are exactly the kinds of discussions we run in our Inner Circle—small, focused, and grounded in real-world experience.

Members can access the full session recording and continue the conversation on our community platform.