By: Dr. Tamara Patzer
A company’s how-to documents, templates and checklists are not usually considered valuable outside the building. micro1, an AI data company in San Francisco, is betting otherwise. It announced on October 9 that it will spend $1 billion over 12 months on Company Data Partnerships. If your company runs on procedures, playbooks, saved message threads and shared documents, you may hold more of that material than you realize, and it could be worth more than you expect.
Key Takeaways
- The documents in question are procedures, templates, workflows and knowledge bases.
- micro1 uses them to build practice environments for AI agents.
- The program lists a minimum of 30 full-time employees.
- About 654,501 U.S. firms have between 20 and 499 employees.
What Could Count: Your Messages, Procedures and Documents
Paperwork is where most of this starts. The permit checklist, the intake form, the training manual in a ring binder and the string of emails that explains an exception can all be part of the picture. A binder that was scanned once for an audit may already be most of the way to a usable file.
Owners often have their own words for it. You may call it “the way we do things here.” In this world, it is called a standard operating procedure. The binder in the back office is a playbook. How your best manager handles an unhappy customer is a workflow. The email everyone copies is a template. The shared drive nobody can search is a knowledge base. The filing cabinet of paper manuals is a document archive, and it can be scanned.
That last point matters. Paper binders, printed manuals and signed forms can be scanned, and files that already exist digitally can usually be exported. micro1’s page lists procedures, process documents, knowledge bases, workflows and templates. Message and email archives are a separate question, so ask the program which record types it accepts before assuming either way.
A short test: ask a few people who run things what a new hire needs to learn in the first year. Whatever they point to may be what a buyer wants to see. If you are near the 30 full-time employee line, count every location and run the online estimator. It takes minutes, and you may qualify.
What Happens to a Procedure
Take a three-page checklist for handling a refund. A person reads it and processes the refund. An AI agent can be trained on the same checklist by practicing in a simulated workplace, where its attempts are scored and corrected. Micro1 calls these reinforcement learning environments.
The company says it builds them from de-identified company data. Its page lists standard operating procedures, process documentation, knowledge bases, workflows and templates. It also lists CRM records, support tickets and project histories, which carry more risk because they may contain personal details.
Why Now
AI models have read most of the public internet. What they lack is the inside view of how work gets done. Business Insider, as summarized by Let’s Data Science, reported that Gartner expects AI data spending to rise from under $1 billion last year to nearly $6.5 billion in 2027. That is a forecast.
Who Can Take Part
Micro1’s page describes established companies with at least 30 full-time employees and several years of records across multiple business systems. The data should be primarily in English. The Small Business Administration counts 654,501 U.S. firms with 20 to 499 employees, so the eligible group is a small share of the roughly 36 million small businesses in the country.
What Companies Should Check
Rajesh Beri, writing at Beri.net on October 10, cautioned that owning files is not the same as owning the right to license them. Customer contracts, privacy rules and third-party materials can all limit what a company may share.
micro1 says companies keep ownership of their data, do not need to anonymize it themselves, and can expect raw data to be deleted on an agreed schedule. The terms of any license determine what that means in practice.
The Steps
The application process runs from an online estimator to an indicative estimate, a short application, pricing, a contract and delivery. micro1 has 30 days to review delivered data, and payment follows within 30 days of invoicing after approval. The page lists advertised ranges from $100,000 upward, but calls its estimate “indicative only.”
What Stays Out
A company does not have to hand over everything it owns. A cautious approach separates the pile into three groups. The first holds generic procedures, templates and playbooks that name no one. The second holds material that mentions customers, employees or partners. The third holds anything built by a vendor, a franchisor or a software provider, which may not be the company’s to license.
Only the first group is simple. The second needs privacy review. The third needs permission from whoever owns it. Micro1’s page says personal and identifying information is removed or transformed before data goes to third parties, but companies still need to decide what they are comfortable sending.
Questions to Ask Before Sharing a Sample
Who is the end buyer? Is the license exclusive? What happens to the data if the agreement ends? Who pays if something goes wrong? Beri noted that micro1’s public terms do not answer these. They are contract questions, and a lawyer should answer them.
What the Money Looks Like on Paper
As of November 7, micro1’s page lists three advertised tiers: $100,000 to $500,000, $500,000 to $3 million, and $3 million and up. The company calls its first estimate “indicative only,” and Magica noted in October that advertised tiers are not a record of what companies have received. Readers should treat the figures as the start of a conversation, not a forecast.
A Plain-Language Glossary
Reinforcement learning is a method in which a program tries a task, receives a score and adjusts. A reinforcement learning environment is the simulated setting where that happens, such as a mock customer-service desk. De-identified data is data from which names and other direct identifiers have been removed or transformed. A license is permission to use data under specific terms, which is not the same as selling it outright.
Knowing these four terms makes the program’s page easier to read, and it helps owners ask better questions when they talk to their lawyer.
Frequently Asked Questions
Is an offer guaranteed?
No. Each partnership is evaluated individually.
Does the program take customer data?
The page lists CRM records and support tickets among what it accepts, and says personal information is removed or transformed. Many advisers suggest starting with procedures and templates.
Where can a company start?
Eligible companies can see the eligibility requirements or visit micro1’s company partnership page.
Does micro1 build AI agents for companies?
The page describes an optional agent program in which micro1 can build and deploy agents in a company’s workflows using the company’s own systems and permissions, run continuous evaluations, and license qualifying agent trajectories for $1 million to $2 million a year with company approval. It is a listed option, and a lawyer should explain it before anyone agrees.
Who can I ask at micro1?
Start with micro1’s application page, linked in this article.
Dr. Tamara Patzer is a publisher and former newspaper editor who writes about AI and business data.
Sources (all read November 7, 2026)
- micro1, “micro1 commits $1B to enterprise data acquisition,” October 9, 2026: micro1.ai/blog/micro1-commits-1b-to-enterprise-data-acquisition
- micro1, Company Data Partnerships estimator and FAQ: micro1.ai/data-estimate
- U.S. Small Business Administration Office of Advocacy, 2025 Small Business Profile for the United States (Census Bureau 2022 data)
- Let’s Data Science, “Micro1 Commits $1B to Buy and License Enterprise Data,” citing Business Insider and Gartner
- Rajesh Beri, “micro1 commits $1B to buy enterprise data for AI training,” Beri.net, October 10, 2026
- micro1, Data Partnerships page, micro1.ai/data-partnerships, read November 7, 2026
Disclaimer: The author is a referral partner of micro1. This article is information, not a promise of money, and not legal advice.
