10 rules to find companies with data worth licensing, and to skip the rest.
| Rule | The problem | How we help |
|---|---|---|
| 1Check what data a company holds before you contact it | Most companies have little useful data, and you lose weeks on the wrong ones. | A score from 0 to 100 for any company, in seconds. Score any company |
| 2Older companies have more data than big young ones | Young companies have only a few years of records. AI labs pay most for older records, written before AI tools. | How many years each company has been online. How the score works |
| 3A company's software shows what data it has | You can't see inside a company, so you don't know what data it has. | The software each company uses, like a help desk (support tickets), a CRM (customer and sales records) or call recording (recorded calls). Lists by data type |
| 4Check what the buyer wants before you offer a company | Buyers turn companies down if they are too small, in the wrong country or have the wrong data. | A check against each buyer's rules in about 20 seconds. Check a company |
| 5Companies that are closing down sell their data fastest | When a company closes, its data is often sold or deleted within weeks. | 42,000+ US companies that show signs of winding down, with the systems they ran. Shutdown Radar |
| 6Contact new companies every week, not the same famous names | Everyone calls the same well-known companies. | About 4,200 new US companies every week, 200 for each industry list. Broker Feed |
| 7Find companies that match exactly what the AI lab asks for | Finding, for example, chip design firms that use one method (like UVM) takes days by hand. | 218 methods and certifications tracked, with proof from each company's website. Search by method |
| 8Only contact companies that are still in business | Old databases are full of closed companies, and your emails bounce. | Every company in our lists is checked to be in business, out of 936,578 US companies we score. Company lists |
| 9Start with legal, accounting, software and healthcare companies | Effort goes to industries that labs rarely ask about. | Ready lists for 21 US industries, plus chip design, semiconductor equipment, aerospace, biotech and science labs. Browse the lists |
| 10Start with the companies you already know | You know hundreds of people but not which of their companies matter. | Send us the websites of companies you know and get a score for each, up to 100 at a time. Score your own list |
These rules are for people who find companies for AI data buyers. Some of you work as brokers. Some of you run data companies and need a steady flow of sellers. Some of you simply know a lot of business owners and want to put that network to work.
The job sounds simple. A lab wants records of real work, such as support tickets, sales calls, contracts and code. You find a company that holds those records and will license them. In practice, most of the time goes to companies that were never going to sell. The company is too young. Its records sit in personal inboxes. It is too small for the buyer's rules. Or it closed two years ago, and nobody updated the database.
Each rule below fixes one of those leaks. You can read them in order or jump straight to the one that costs you the most time today.
What goes wrong, and why it costs you time and goodwill.
What to do instead, with steps you can use today for free.
A real figure from our index of US companies, and the page that does the step for you.
Most companies hold little that a lab will license. A lab does not want a website, a brochure or a contact list. It wants years of internal work records, in English, from a company that owns them and can export them. Most firms fall short on at least one of those points.
The trouble is that you find out late. You spend a week on research, intros, calls and follow-ups. Then the owner says the records sit in personal inboxes, or the buyer says the company is too small. The week is gone, and so is some goodwill with the friend who made the intro.
Rank companies before you reach out. Give each one a quick pass or fail before you write a single email. Four questions do most of the work:
If any answer is no, move the company to the bottom of your list. Spend your week on the few that pass all four. Then sort those by strength, and lead with the company that has the longest, most connected history. You will learn faster what buyers say yes to.
Headcount is easy to see, so people lead with it. A startup with hundreds of staff looks like a big catch. But labs pay for years of real work written before AI: tickets, emails, call notes, documents and code, made by people doing their jobs.
A young company has few of those years, however many people it hired last year. Much of its newer writing may also have passed through AI tools. Buyers want the human record, and that record takes years to build.
Ask about years before you ask about size. Put three questions near the top of every first call:
A small firm with many years of unbroken records often beats a large firm with only a few. Size still matters for the buyer's rules, so check it next (rule 4). But history comes first.
Watch for gaps, too. A move to a new help desk can drop years of tickets. A short retention setting can wipe old chat. Ask whether the old data was kept, exported or lost before you tell a buyer how many years there are.
Ask an owner what data the company has, and you get a vague answer. Some say "tons." Some say "nothing special." Neither answer helps you. Owners do not think in the terms a lab uses, and most have never had a reason to.
So you pitch blind. You promise a buyer support tickets, and the company turns out to run support from a shared inbox. Or you skip a company that holds years of recorded sales calls, because nobody mentioned them.
Ask about systems, not data. Every system leaves a kind of record behind, and the map is simple:
So ask one plain question: which tools does your team use every day, and since when? The answer tells you what records exist and how many years they cover. Then ask whether the systems link to each other. A ticket that points to a code change and a reply to the customer is worth more than three separate piles.
Buyer programs publish hard rules, and they reject companies that miss them. A company can hold great records and still fail on headcount, country or language. Pitch it anyway, and the buyer says no. The owner feels let down, and your name takes the hit on both sides.
Read each buyer's published rules, and check every company against them before anyone gets your email. Two examples: micro1 asks for 30+ employees, and Mode asks for 20+ US staff, as published by those programs.
Keep a short checklist for each buyer you work with, and run every company through it. It takes less time than one wasted call. You can see the published rules side by side on our page of buyer programs compared.
A company that is winding down runs on a short clock. Its tickets, email and code are often sold or deleted within weeks. Subscriptions lapse. Servers get wiped. The staff who knew the systems leave. If you hear about the closure a month late, the records may already be gone.
That is also why these companies sell fastest. The records will not be used again inside the company, and a license can turn them into one last payment.
Treat every closing company as urgent. Speed matters more here than anywhere else.
Be kind and be direct. The owner is having a hard month. A short, honest note that explains the value of the records lands better than a sales pitch. Our guide on selling data when shutting down covers the owner's side of the process.
Everyone calls the same famous names. The best known firms in each industry hear from brokers and data companies all the time. Their owners stop answering. The deals are in companies nobody has called yet.
Lists go stale, too. If you work from the same list for a year, you end up calling the same people again, and they remember.
Add fresh companies to your pipeline every week, not once a quarter.
A steady flow of new names beats a big list worked once. It also gives you more chances to learn which companies say yes, so each week gets a little easier.
Labs often ask for something narrow: "chip design firms that use UVM," "defense suppliers that work under ITAR," or "labs with CLIA certification." A request like that takes days by hand. You open site after site, read page after page, and guess.
Then you send a list that is close, and close does not sell. The lab checks the first few names, finds one that does not fit, and stops trusting the rest.
Treat the request as a checklist, not a theme.
Learn the words of your niche, too. UVM is a standard method for testing chip designs. ITAR is the US rule set for defense exports. CLIA is the US certification for clinical labs. When you know what a term means, you spot weak matches fast, and the lab sees that you know its field.
Old company databases are full of closed companies. Firms shut down, merge or change names, and the old records stay. You find out when the third email bounces. Worse, you spend an hour of research, or a friend's intro, on a company that stopped trading years ago.
Check that a company is alive before you spend time on it.
Note the difference from rule 5. A company that is closing still has people and records, so it is a good lead. A company that closed years ago usually has neither. The first needs speed. The second needs to leave your list.
Labs ask about some industries again and again, and others almost never. If you spread your effort across every sector, most of it lands where demand is thin. Each industry also has its own words, rules and risks. Nobody learns them all at once.
Go where labs ask for data: legal, accounting, software, healthcare and science labs. Pick one or two, and learn them well.
Learn the job titles, the systems and the legal limits in the field you choose. Read the buyer program pages often, and move early when a buyer names a new industry or data type.
You already know a lot of people. Say you know 500. Maybe 20 of their companies hold useful records. The rest are too young, too small or in the wrong field.
Without a filter, you face a bad choice. You pitch everyone and wear out your network, or you pitch no one and waste it.
Sort your network before you ask anyone for anything.
A warm intro from a friend beats a cold email, so use it where it counts. Repeat the exercise every few months. People change jobs, and a contact at a weak company today may join a strong one next year.
You can follow every rule on this page with a notebook and some patience. If you want to move faster, each tool below does one of the steps for you.
Ready lists of US companies by industry, data type and niche. Every company checked as active.
See the lists New companies every weekAbout 4,200 new US companies every week, 200 for each of 21 industry lists, each one scored.
See the Broker Feed Companies closing down42,000+ US companies showing signs of winding down, with the systems they ran.
See the Radar Score your own listSend up to 100 company domains per request and get a 0 to 100 score for each.
See the API Sell your own company's dataSee your own company's score and how it fits the published rules of the buyer programs.
Check your companyNo. They work for anyone who finds companies for AI data buyers: brokers, data companies, consultants and people with a large business network. Company owners can use them too. Read them the other way around, and they tell you whether your own company is a good fit before you talk to a buyer.
No. Every rule comes with advice you can use with no tools at all: questions to ask, checks to run and habits to keep. Our products make each step faster, and the company check is free.
From our own index of US companies, as of October 2026. They describe the companies in our lists and the US company domains we score. They will change as the index grows.
The AI Data Value Score rates a company from 0 to 100 for how much AI labs would want its data. A high score is a reason to look closer. It is not a price or an offer. Only a buyer that reviews the data can say what it is worth.
Labs pay for years of real work written before AI: tickets, emails, call notes, documents and code. A young company has few of those years, however large it is. Size still matters for buyer rules. micro1 asks for 30+ employees and Mode for 20+ US staff, as published by those programs.
Yes. The rules apply anywhere. The numbers on this page cover US companies only. Buyer programs also set country rules of their own, so check those first (rule 4).