AI Companies Acquire Internal Data from Messaging and Email Systems

AI development companies are increasingly acquiring internal business data such as corporate messaging, emails, and video conference recordings. This trend is driven by the recognition that publicly available internet data has limitations when it comes to training AI agents to handle real work processes.

Deinformation reported that, based on interviews with over 20 founders, data companies, and financial advisors, the demand for corporate data, including internal messaging conversations, emails, video conference recordings, and code change histories, has surged in recent months. Companies like OpenAI and Anthropic are seeking actual work interaction records beyond publicly available web data.

Bobby Samuels, CEO of Protege, stated, "AI labs have already scraped the entire internet. Now what they want is real interactions between people." The total transaction volume handled by Protege has increased from $30 million (approximately 42.5 billion KRW) last year to at least $100 million (approximately 141.8 billion KRW) this year.

The transactions involve emails that capture how developers solve problems, conversations where CFOs discuss performance during conference calls, and videos demonstrating software usage. The buyers remove personally identifiable information before selling the data to AI model training companies. However, most of these transactions are not disclosed publicly.

The demand for corporate data is intertwined with the competition among AI agents. AI agents focus on performing actual work procedures, such as customer service, clinical trial management, and invoice verification. We previously reported that the competition among AI agents is shifting from model performance to corporate systems.

Corporate work involves procedures that intertwine multiple programs, such as document searches, customer relationship management inputs, internal messaging notifications, and report writing. Due to the interplay of access rights, exception handling, audit trails, and human approval processes, automating the entire workflow is challenging with just the ability to manipulate screens.

The data that AI companies seek is aligned with this point. Rather than simple documents, the sequence of human judgment, the messages used for collaboration, and how exceptions are handled are evaluated as learning materials. While public web documents are closer to outcomes, internal conversations and work records are seen as closer to the process.

A real case has emerged. The AI agent startup Warmly agreed to be acquired by customer management and marketing software company HubSpot on June 30. Warmly subsequently received several offers to acquire its codebase and communication records among employees for up to $300,000 (approximately 425.4 million KRW).

Maximus Greenwald, CEO of Warmly, revealed that he received such offers four times but rejected them all. Buyers showed interest in routine work communication data such as internal meeting transcripts, emails, and messaging conversations that were not included in the acquisition price.

HubSpot stated regarding the acquisition of Warmly, "The purpose was talent acquisition, and no data was included in the acquisition target." Warmly also explained in an official announcement that existing customer contracts, pricing, account managers, product experiences, and integration features would remain unchanged.

The issue of data transactions revolves around privacy. Internal messaging, emails, and meeting transcripts may contain sensitive information about employees, customers, and business partners. Even if the data is anonymized, compliance and rights infringement controversies can arise depending on what information is removed and what context remains.

Shub Sinha, CEO of Integral, stated, "The demand for corporate data is exploding for the development of industry-specific customized apps in sectors such as healthcare, finance, and energy," adding that anonymization processes are conducted to maintain data value while adhering to privacy regulations. Industry-specific work data is considered more directly relevant for training specific work applications than general web documents.

There are opposing views as well. Jason Pressman, managing director of Shasta Ventures, criticized that in acquisitions where companies may be valued at $100 million to $300 million (approximately 141.8 billion to 425.4 billion KRW), offers for licensing worth several million to $10 million (approximately 14.2 billion KRW) are often made, stating, "They are trying to take only the most valuable assets of the company and leave just the shell behind."

This trend indicates that the AI training data market is expanding from public web data to internal corporate records. There has been no confirmation of content directly linked to specific token prices or virtual asset demand. However, as AI agents spread into the realm of corporate work automation, discussions surrounding data rights, security, and privacy protection are bound to grow.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.