Who Owns Your Company's Conversations?
We spent fifteen years untangling CRM lock-in. The same mistake is being made right now, one layer up — in voice.
Photo Credit: Craig Pattenaude on Unsplash
David Haber’s a16z piece, Everything is Recorded Now, is the kind of essay that lands differently depending on where you sit. If you run internal operations at a fast-moving startup, it reads as permission — an intellectual mandate to stop debating recording and start building institutional memory. If you’re in enterprise architecture, compliance, or customer-facing operations, it reads as a warning. A category is being defined, and the governance decisions being made in the next 18 months will either replicate the lock-in patterns of the last two decades or break them for good.
Haber is right about the irreversibility. The shift from “opt-in to record” to “assume you’re being recorded” is already underway, driven by the same compounding logic that made email searchable and Slack discoverable. The model that has ingested two years of company conversation is a better assistant than the one that only read the documentation. That argument is settled.
The question Haber leaves open — who owns the substrate that makes this useful — is where the real strategic decision gets made.
The Intelligence Economy Has a Memory Problem
I have argued before that the intelligence economy is being built on systems that forget. We’re pouring billions into models that can reason and generate, but the conversational context those models need to be genuinely useful — the nuance in a customer call, the real argument in a product review, the offhand comment in a leadership meeting that quietly changes the roadmap — evaporates the moment the call ends, or gets locked inside a tool that doesn’t talk to anything else.
Haber frames this beautifully: the highest-value context doesn’t live in structured data. It lives in conversation. And LLMs are uniquely equipped to take that unstructured voice data and make it structured, searchable, queryable. The software category being built around this insight is real, and it’s large. The voice AI market reached $12.5 billion in 2026 and is projected to hit $47.5 billion by 2034, with enterprise voice — contact centers, virtual assistants, customer-facing operations — accounting for $7.8 billion of that total. Gartner projects that conversational AI will reduce contact center labor costs by $80 billion globally this year alone. The category isn’t speculative. It’s already in production.
What’s not resolved is the architectural question underneath all of it.
Granola Is a Proof Point, Not the Blueprint
Haber uses Granola as his clearest example: it has better context on a16z’s culture, investments, and operating logic than almost any other tool they use, because it’s been in every meeting. That’s a compelling proof point, and the market validated it quickly — Granola raised $125 million in March 2026 at a $1.5 billion valuation, a sixfold jump from its previous round less than a year earlier. Revenue grew 250% in the quarter before the raise.
But Granola is also, at this stage, a silo — and Haber acknowledges it. It has built enterprise APIs, SSO, SCIM, and team-wide context management. It integrates with tools like Claude and Figma. It is doing exactly what a well-funded, well-designed product should do. And none of that changes the structural reality: the context it builds lives inside Granola, accessible to other systems only on Granola’s terms.
This is not a criticism of Granola’s engineering. It’s a pattern. Salesforce did the same thing, and we spent fifteen years complaining about it. The lock-in wasn’t malicious — it was the natural consequence of running a business on top of a system of record for long enough that the data accumulated, the workflows ossified, and switching became prohibitively expensive. Vendor lock-in erodes organizational agility and compromises long-term value. The conversational layer is about to do the same thing, at a faster pace, with stakes that are higher — because the data being captured includes regulated customer interactions, legally discoverable communications, and the institutional knowledge of the company itself.
The Harder Question Is Not Internal Meetings
Here is where Haber’s framing is worth sharpening. His examples skew toward internal context: a16z’s culture, OpenAI’s agent-in-meeting model, Bridgewater’s institutional recording policy. These are real and important. But the volume, regulatory weight, and interoperability challenge are orders of magnitude larger in customer-facing conversations.
Consider the actual scope: 81% of businesses have already implemented AI in contact centers. By year-end 2026, 80% of customer service organizations will use generative AI in some form. The conversational AI market in intelligent contact centers alone is growing at 18.66% CAGR through 2030. Customer conversations happen across contact centers, messaging platforms, voice channels, email threads, and video calls — all of it siloed, all of it subject to sector-specific regulation (HIPAA in healthcare, PCI-DSS in financial services, GDPR across borders), and none of it currently connected to a unified governance layer.
The CRM lock-in problem we spent fifteen years on must not get repeated one layer up, in voice.
When enterprises ask “who owns the conversational data from our contact center?”, the answer today is: whoever sold them the CCaaS platform. That’s not a governance strategy. It’s an accident waiting to become a compliance liability.
An Open Container Already Exists
The good news is that the infrastructure problem has already been solved at the standards level. vCon — the Virtualized Conversation standard developed under the IETF — is the open container format for conversational data that the enterprise needs.
vCon is not a product; it’s a standardized JSON framework for representing any conversation in a platform-independent way, regardless of whether that conversation happened over a phone call, video conference, SMS exchange, web chat, or email thread. A vCon container can hold metadata, participant identity, the actual audio or video, transcripts, real-time analysis, and attachments — all in a single, portable, governance-ready object. The standard was co-authored by Thomas McCarthy-Howe, now of Vconic, and has been in IETF working group development for three years, with an ecosystem of 30+ participants already building around the standard.
The IETF vCon working group has also drafted a lawful basis extension — specifically addressing how to record, verify, and manage the legal grounds for processing conversational data within each container. That’s not a compliance checkbox. That’s governance baked into the data format itself, at the schema level, before any application layer touches it.
This is the substrate Haber is pointing toward when he writes that the enterprise has a window to build the right governance while there’s still a choice to make. The container exists. The question is whether enterprises will demand it, or accept proprietary silos until it’s too late to change.
The Open Standard Wins When Multiple Players Build On It
One thing worth naming clearly: vCon is not a single-vendor play. Platforms offer vCons-as-a-Service, capturing, containerizing, and connecting enterprise conversation data using the open standard without requiring changes to existing systems. Other players are building commercial and infrastructure layers on top of the same open substrate. That’s not fragmentation; it’s the ecosystem dynamics that make a standard real.
The TCP/IP analogy is overused, but it’s accurate here. The internet didn’t grow because one company owned the packet format. It grew because the packet format was open and any company could build on top of it. The enterprises that understood that early — that insisted on open infrastructure before deploying proprietary applications — were the ones with flexibility when the market shifted.
A growing ecosystem of vCon-native builders is a feature, not a bug. It means enterprises can choose the commercial layer that fits their architecture, their regulatory environment, and their vendor relationships — without losing the ability to move their conversational data. That’s the outcome that avoids the CRM mistake.
The Enterprise Window Is Real — And Narrow
Haber’s closing point deserves to be read carefully: “The question isn’t whether this happens. It’s whether you get there first, and build the right governance around it while you still have the advantage of choosing.”
That window is measurable. Every enterprise contact center that signs a multi-year CCaaS agreement without a vCon portability clause is making a data governance decision, even if it doesn’t know it. Every internal productivity tool that records meetings into a proprietary silo is creating the next migration project. The data accumulates fast, the workflows ossify faster, and the exit costs compound.
For enterprise operators: the conversation about recording is no longer the right conversation. The conversation about where that data lives, who can access it, and whether you can take it with you is the one that matters. The governance infrastructure — open container format, lawful basis extension, portable queryable memory — exists today and is production-ready.
For channel partners and technology operators: the category is being defined now. The enterprises that figure this out in the next 12 months will be ahead of the ones who revisit it in 2028, after three years of proprietary lock-in and a migration they didn’t budget for.
The substrate has to be open. That’s not an ideological position. It’s a lesson the enterprise learned the hard way once already.
If you’re an enterprise operator, architect, or channel partner thinking about conversational data governance, the vCon ecosystem is the right place to start. Explore the standard here and the open-source production platform at Conserver.io. Connect with me on LinkedIn here or subscribe to Global AI Leaders for intelligence on where enterprise AI is actually heading.
About the Author
Ken Herron is a B2B SaaS strategist working at the intersection of enterprise data infrastructure and conversational AI. With more than 30 years of experience across telecommunications, contact centers, and AI-driven operations on five continents, his focus is on a question the industry keeps deferring: who actually owns the conversational data your enterprise generates every day — and what happens when you find out the answer is someone else?
He started Global AI Leaders for operators, architects, and technology leaders navigating the moment when AI stops being a productivity tool and starts being a system of record. His work centers on open standards — particularly the IETF vCon standard — as the foundation layer that prevents the next generation of enterprise lock-in from being built on proprietary voice silos instead of proprietary CRM data.
Ken’s perspective is grounded in commercial reality: governance decisions made in the next 18 months will compound for a decade. He covers them here, before they become migration projects.
About Global AI Leaders
Global AI Leaders is a practitioner-led briefing on the systems being built around AI — not just the models themselves.
The focus is where most coverage stops: infrastructure, governance, and the decisions that compound quietly until they’re too expensive to reverse. How AI reshapes enterprise operations, redistributes data ownership, and exposes the gaps in systems that were never designed for a world where every conversation is a strategic asset.
If you’re building, deploying, or governing AI inside a real organization — especially where customer conversations, compliance, and institutional memory intersect — this is written for you.
Less hype. More signal. And a sharper argument every week.
If something in this piece maps to what you’re seeing inside your own systems, add your perspective in the comments. That’s how this conversation gets better.