Every day, hundreds of millions of phone calls are made, customer support chats resolved, video conferences concluded, and emails threaded to their end. Each one is a conversation—a real-time exchange of intent, information, and meaning. And for the most part, when it ends, the data it produces scatters into proprietary silos and incompatible formats, where it is poorly searchable, difficult to audit, and hard to feed into AI systems that could learn from it.
That is where vCon—short for Virtualized Conversation—comes in. At its technical core, a vCon is a formatted digital container designed to hold everything pertaining to a single human conversation. That includes call detail records and metadata, the identities of participants, the actual content of the exchange in whatever form it took (audio, video, or text), real-time or post-conversation analysis such as transcriptions and sentiment scores, and any files or attachments exchanged along the way.
Think of it as “a PDF for conversations”—a portable, self-contained record that works across systems and decades. Just as the PDF made a document portable and self-contained, vCon aspires to do the same for a phone call, a web chat, a video conference, an SMS thread, or an email exchange—wrapping all of it into a single, tamper-resistant, optionally encrypted object that can be stored, transmitted, analyzed, and audited on its own terms.
Jeff Pulver, an Internet communications pioneer, sees sweeping implications. “vCon’s ability to connect messaging realms is a game changer for companies,” he says. “It will give management an unparalleled overview of dark operational data and AI agents the ability to view the conversations that took place and the actions recommended.”
Pulver has formed the vCon Foundation, a global nonprofit that brings together hardware and software service providers as well as pioneers in the voice-over-IP sector to help advance the technology. And a working group at the Internet Engineering Task Force (IETF) is developing a global standard for vCon, potentially paving the way for universal industry adoption.
The IETF’s track record is why that matters. For more than four decades, this volunteer body of engineers has quietly written the rulebook for the internet. Email moves between providers because of an IETF standard. The web addresses you type, the padlock that signals a secure connection, the way your phone finds a website by name, the protocols behind Zoom and FaceTime calls, and the modern anti-spam checks that keep your inbox usable all began as IETF documents. When the IETF settles on a standard, it tends to disappear into the background as invisible infrastructure that nobody markets but everything depends on. A vCon standard from the same body carries the same promise of becoming the default plumbing for conversational data, owned by no single vendor and usable by all of them.
While the standard is still in progress, implementations are already emerging. vCon-based infrastructure has been deployed in the automotive sales sector, using the containers to capture and analyze dealership conversations for training and performance insights.
Vconic, a vCon pioneer and one of the first companies to join the Foundation, has already begun signing up corporate customers. “We have a large financial institution that has just gone live, producing literally millions of vCons per day,” says Vconic CEO Perry Evans. “We expect them to increase that to a million an hour.”
The technology is attracting interest from contact center operators, telecommunications providers, healthcare technology companies, and compliance professionals, as well as other sectors where conversations are both operationally critical and legally sensitive.
The Need For A Standard Way Of Handling Conversations
Conversational data is among the richest and most strategically valuable information a business generates—and among the most poorly managed.
Customer support calls reveal product failures and training gaps. Sales conversations contain competitive intelligence and signals about what messaging works. Clinical consultations are medical records in audio form. Yet most organizations treat conversations as ephemeral events: recorded, perhaps, but siloed in whatever platform captured them, inaccessible to other systems without custom integrations, and nearly impossible to manage coherently at scale.
While most systems offer some way to store conversational data, a lack of standards or interoperability in storage and transmission mechanisms leads to fragmentation. For example, a healthcare provider’s call center, its electronic health records system, and its AI-powered quality assurance tool may each hold a piece of a conversation, none in a format the others can natively read.
The situation has become more urgent as AI has entered the picture. Large language models can summarize calls, flag compliance issues, detect sentiment, and surface insights at scale—but only if they have structured, reliable data to work with. Feeding raw audio files or proprietary transcripts into AI pipelines is messy and inconsistent. A standardized container changes the equation.
AI Alignment By Design
Pulver argues that the timing of vCon’s emergence as an IETF standard is not incidental to AI—it is central to its value.
Because large language models were trained heavily on IETF documentation and open standards, they already understand the structure and semantics of specifications like vCon. That means AI systems processing vCon-formatted data are less likely to misinterpret it, hallucinate about its structure, or require custom fine-tuning for every new format they encounter. A standardized conversational container, in this view, doesn’t just organize data—it makes that data more reliably legible to the AI systems that will act on it.
That matters because the alternative is an AI landscape in which every organization’s conversational data remains locked in proprietary formats, each requiring its own integration work, each introducing its own failure modes. vCon proponents believe the standard offers an opportunity analogous to what the Session Initiation Protocol (SIP) did for voice-over-IP: a common language that lets an ecosystem of tools, services, and platforms interoperate freely.
The Driving Force Behind vCon
Thomas McCarthy-Howe, Vconic’s CTO and the creator of the standard now being developed in the IETF vCon Working Group, is the driving force behind the technology.
For more than 30 years, McCarthy-Howe has worked at the intersection of real-time communications, distributed systems, and applied AI—from early SIP and Digital Subscriber Line (DSL) deployments through large-scale contact-center platforms to today’s privacy-preserving conversational AI.
In an interview with The Innovator McCarthy-Howe says he saw a pressing need to build a standard file format for conversations.“Everything we’ve digitized in computing eventually had a file format,” he says. “We digitized pictures, spreadsheets, documents, contact cards, and appointments. But at some point in the standards work, we stopped defining them. And I found that because I couldn’t define a conversation, I couldn’t build tools to work on it. And because it wasn’t a file, it couldn’t travel. That made data silos.”
Determined to fix the problem, McCarthy-Howe took a circuitous route to market. He convinced Vinnie Micciche, a friend and owner of Strolid—a New Hampshire company that specializes in helping auto dealers sell more cars—to fund the research and development. Over three years of R&D investment, vCons were used by Strolid to listen to customers at scale, processing millions of real conversations to deliver better customer experiences. The effort generated millions of automotive appointments and proved the platform’s reliability and scalability in a demanding, high-volume production environment.
McCarthy-Howe argues that CRMs flatten everything into text notes and represent “random acts of integration that don’t capture a true picture of what’s going on between a company and its customers. But “if every conversation is deeply mined for the intelligence inside it, you start to build things like better intelligence for sales forecasting,” he says. “You get the content, the tone, and the context of the conversation—so if you’re a CFO, you will have a much deeper understanding of what’s really going on in the pipeline.”
vCon also converts recordings into training opportunities. “Strolid uses vCon to study the performance of the best sales reps: What’s their call pattern? Do they mention price? Do they mention the competition? How do they lead in? How can the company use that to continually coach and improve the performance of their outbound sales team?”, he notes.
The technology proved so successful that Strolid spun it out as a separate company—Vconic. The commercial validation, combined with McCarthy-Howe’s vision for a privacy-first vCon standard, proved compelling enough to attract Perry Evans, co-founder of MapQuest, as CEO, and Jeremie Miller, creator of Jabber and XMPP, as strategic advisor. The company’s “Conserver” platform—essentially a pipeline for processing and routing vCon objects—is being positioned as open infrastructure for building conversational AI applications.
In the age of agentic AI, Vconic presents itself as an important alternative to proprietary solutions. Evans points to Salesforce’s Agentforce—autonomous AI agents designed to handle tasks across service, sales, marketing, and commerce—as an example of the complexity enterprises face.
“If you’re an enterprise CIO, that sounds like a really high-risk, complex, big enterprise project,” says Evans. “We can work with the data that’s already in the call center and bring it into the enterprise model in a much more rapid, lower-risk, and higher-impact way by structuring the data in a comprehensive model.”
Vconic is building enterprise applications on top of the open-source Conserver platform. “Our philosophy is to build proof-of-value applications that show enterprises how they can use vCon data in customer support, marketing, product management, sales operations, financial operations, and compliance,” says Evans. “If we can hand over the prompts, the skills, and the learning we’ve built—and create a library of ways they can take advantage of this new model—then the enterprise gets a lot more value creation and acceleration.”
The Privacy Tension At The Heart Of vCon
As compelling as the idea of total recall is for corporations, vCon challenges companies to balance two opposing forces: the privacy of personal data and the rewards of analyzing conversations at scale.
Conversations are, by their nature, intimate. They contain names, medical information, financial details, emotional states, and confidences. Any standard designed to make conversational data more portable and analyzable must grapple seriously with what that portability means for the people whose voices are in the recordings.
The IETF standards group has sought to build privacy mechanisms directly into the standard rather than treat them as an afterthought. vCons support data minimization, allowing different portions of a container to be disclosed to different parties while others remain protected.
A proposed consent structure—developed by McCarthy-Howe, privacy specialist Diana James and long time Microsoft veteran Steve Lasker—embeds structured consent metadata directly into the container, recording who consented to what, under what time limits, and with cryptographic proof that the consent record has not been altered.
This consent framework is designed to work with regulations such as GDPR and CCPA, where consent can be revoked and organizations are obligated to coordinate deletion across all systems holding the relevant data. By making consent a first-class object within the container itself, vCon makes it possible, at least in principle, to automate compliance in ways that are currently manual, error-prone, and often performed after the fact.
A parallel effort explores using the IETF’s Supply Chain Integrity, Transparency, and Trust (SCITT) protocol to create auditable, tamper-evident logs of a vCon’s lifecycle, from creation through distribution to deletion. The goal is to give regulators—and the individuals whose conversations are captured—a verifiable record of how their data was used. At the same time, it records the provenance of the conversations, enabling detection of deep fakes.
There are, however, hard questions that no technical standard can fully resolve. Even the most sophisticated consent framework cannot force organizations to honor it. Encrypted containers can be decrypted by those with the keys. And the same infrastructure that makes it easier to protect conversational data also makes it easier to aggregate it at unprecedented scale.
The Conversation About Conversations Is Just Beginning
The vCon standard sits at the confluence of several powerful and converging forces—the explosion of AI, growing regulatory scrutiny of data practices, and enterprises’ long-frustrated desire to use the conversational intelligence they generate. What is clear is that the conversation about conversations is just beginning. The trillions of exchanges that flow through the world’s enterprises each year have long been treated as exhaust—a byproduct of doing business, quickly vented and forgotten. vCon is a bet that they are, in fact, the signal. And that capturing them faithfully, managing them responsibly, and analyzing them intelligently could reshape how organizations understand their customers, train their people, and govern themselves. The infrastructure for total recall is being built. The harder question—who controls the memory, and to what ends—will define what comes next.
This article is content that would normally only be available to subscribers. Become a subscriber to see what you have been missing