Written by Julien Ricciarelli-Bonnal
26 October 2026
The Essentials
American AI data company Micro1 has announced plans to spend $1 billion over twelve months acquiring and licensing operational data from businesses. According to a Business Insider investigation published on October 26, 2026, the information being sought includes Slack conversations, Google documents, CRM histories and everyday working files. Some businesses could receive between $100,000 and $2 million for their data. The initiative reflects a shift in artificial intelligence development. Having learned from books, websites and publicly available information, AI systems are increasingly being trained to understand how people work inside real organisations, make decisions, manage projects and respond when circumstances do not follow established procedures.
For much of the generative AI boom, the industry’s principal challenge was acquiring enough information to teach models how language, knowledge and human communication work. Vast collections of online material helped systems produce increasingly sophisticated explanations, reports and computer code. Yet an extensive understanding of business terminology does not necessarily translate into the ability to perform business tasks effectively.
A sales team rarely operates according to a perfectly documented sequence. Negotiations stall, budgets change, managers disagree and customers request exceptions. Projects move forward through informal conversations, revised documents and decisions that may never appear in an official procedure. Much of the knowledge required to navigate these situations is embedded in the organisation’s daily activity.
That is the information AI developers increasingly want to access. The market is moving beyond documents describing what professionals should do towards records showing what they actually do, including the mistakes, compromises and unexpected developments that make real work difficult to automate.
The New Commercial Value of Ordinary Workplace Records
On October 9, Micro1 announced a twelve-month commitment to spend $1 billion acquiring and licensing enterprise operational data. The company said the initiative would be financed through capital from Citi and Hercules Capital. The figure represents an announced spending commitment, not a sum already paid to participating businesses.
Business Insider’s October 26 investigation explains how the company is pursuing this market. Micro1 says it generally pays between $100,000 and $2 million for a mid-sized company’s operational data, depending on its scale, complexity and uniqueness. Workplace messaging, customer relationship management records, project documents and design files are among the materials considered valuable.
The investigation describes agreements involving VanHack, a recruitment business that provided information from systems including Google Workspace, Confluence, Slack and its CRM, and MyPortal Marketing, which shared design files and operational checklists. These examples demonstrate that demand is not limited to technology companies or organisations developing artificial intelligence themselves. Ordinary administrative and commercial activities can generate material that model developers consider useful.
What makes this market unusual is that the data was created for another purpose. Employees exchanged messages to resolve problems, maintained CRM records to manage customers and updated documents to coordinate projects. These activities were already creating operational value. The possibility of licensing their historical records adds a separate commercial use that many businesses had never anticipated.
Why AI Needs to Observe Decisions, Not Just Read Procedures
An AI model trained on sales manuals can describe lead qualification, objection handling and contract negotiations. But understanding those principles does not guarantee that it can recognise when a prospect lacks purchasing authority or when a seemingly attractive opportunity should be abandoned. Those judgments often depend on information spread across multiple conversations and systems.
Consider a customer complaint. An official support procedure may specify how to classify the issue and issue a response. The actual resolution might require checking an earlier promise, consulting a logistics provider, obtaining managerial approval and deciding whether an exception is justified. The sequence matters because each action changes what the people involved know and what they can reasonably do next.
Micro1 describes several ways to turn this operational history into AI training material. It can transform internal records for licensing, collect feedback on agents operating within businesses and create simulated working environments based on historical company information. In these environments, a model can be evaluated on how it handles tasks from a given organisational context, including incomplete information and competing priorities.
This approach reflects the growing ambition of agentic AI. Systems designed to perform business activities need more than the ability to produce convincing answers. They need to understand when information is missing, when a task depends on another department and when a decision requires approval. Real operational histories may help researchers train and evaluate these behaviours, although the commercial effectiveness of the resulting agents must still be demonstrated.
The Most Valuable Information May Be the Reason Behind a Decision
A company’s expertise is rarely stored in one perfectly organised knowledge base. Some of it exists in established procedures, but a considerable part emerges through discussions between colleagues, revisions to working documents and responses to unexpected problems. These informal records can reveal how an organisation translates general rules into practical decisions.
For example, a CRM entry recording a lost deal provides a commercial outcome. A sequence of messages explaining the customer’s hesitation, the internal pricing discussion and the final decision provides a much richer account of the process. Similarly, the final version of a project proposal reveals what was approved, while its revision history may explain which assumptions were questioned or rejected.
This distinction affects how data is valued. A large collection of routine messages may be less informative than a smaller but coherent record of complex work. Historical continuity, links between systems and the ability to understand outcomes are potentially more useful than sheer storage volume. Micro1 explicitly highlights workflow complexity, data quality and distinctive operational knowledge when explaining its pricing.
The economic implication is significant. Businesses have traditionally treated internal data as an asset for improving their own processes, customer relationships and performance. Now, some of that accumulated experience may have commercial value outside the organisation. The new asset is not simply information about what the company knows, but evidence of how its people use that knowledge.
The Right to License Data Is Not the Same as Possessing It
Workplace archives do not contain only information created by a company’s owners. They may include employees’ conversations, communications with customers, confidential supplier details and documents subject to contractual restrictions. Having access to these systems does not automatically establish an unrestricted right to license their contents for external AI development.
Micro1 states that it removes or transforms identifying information, segregates company datasets and applies agreed retention periods. These safeguards deserve consideration, but they must be assessed against actual processing methods and contractual terms. Removing names does not necessarily eliminate the possibility of identifying individuals through contextual details, especially when messages describe distinctive roles, events or relationships.
The distinction matters particularly under European data protection law. Truly anonymised data and pseudonymised personal data are not legally equivalent. The latter may remain subject to GDPR obligations, including requirements concerning lawful processing, transparency and international transfers. Commercial confidentiality and intellectual property introduce additional questions beyond personal information.
Employees may also have contributed to records whose commercial value they were never told to expect. The Business Insider investigation raises questions about their awareness and participation in these transactions. This is not necessarily evidence that every arrangement is unlawful or inappropriate. It does mean that an agreement signed by management is only one part of a broader assessment involving workplace relationships, contractual rights and the interests of third parties.
Selling Data May Generate Revenue Without Being a Simple Business Decision
For eligible companies, a substantial licensing payment can be attractive. Years of accumulated files and messages may suddenly represent a potential source of revenue. Yet the economics of a transaction extend beyond the headline amount. Businesses must consider the cost of extracting records, reviewing their contents, establishing the rights to share them and implementing adequate safeguards.
Contractual structure is equally important. Who may use the transformed records, for which purposes and for how long? Is the license exclusive? Can the information be used to develop general-purpose models or only designated evaluation environments? What happens when data must be deleted, and how are derived materials handled? The answers can influence both the financial value and the long-term consequences of an agreement.
This development also connects to the governance challenges discussed in our analysis of AI agents whose actions can become difficult even for their developers to reconstruct. As AI systems gain access to more operational context, the need to define permissions, responsibilities and oversight becomes more concrete. Similar considerations now apply at the earlier stage of determining which workplace information should be made available for their development.
The decision to license operational data consequently belongs within a wider framework of AI business governance. Companies must distinguish information that can be shared responsibly from knowledge whose competitive or confidential character makes external licensing inappropriate. The most profitable-looking transaction may not necessarily be the most advantageous once these considerations are included.
The importance of Micro1’s initiative lies in what it reveals about the direction of the AI industry. Publicly available content has helped models learn to describe the world and explain how work should be performed. Internal business records offer something different: examples of people turning imperfect information into decisions under real organisational constraints.
There is no guarantee that spending a billion dollars on such material will produce agents capable of matching human judgment, nor that every participating business will benefit equally. But the existence of a market for operational histories already changes how businesses may view their digital archives.
For years, organisations have been told that their data was valuable because it could help them make better decisions. They are now being offered money because it may help artificial intelligence learn how those decisions are made. The next question is not simply how much that information is worth, but how much of their accumulated expertise companies should be prepared to share.
Written by Julien Ricciarelli-Bonnal
26 October 2026
