What to have ready before you talk to a data broker: counts, years and a sample

Most first conversations about selling company data start the same way. The owner says what kind of company it is and what kind of records it keeps. The broker asks how many, from which years and whether there is an example. The conversation then waits a week for numbers that were never written down.

You can skip the week. A short list of facts, prepared before the first call, changes what an AI lab does with your data once it reaches them.

Why the numbers matter more than the story

An AI lab sees many more offers than it can read. Most of them say the same thing: years of records, a rich variety, real work. A lab cannot act on that. It can act on a count, a date range and a file it can open.

Two offers for the same kind of data can get opposite results. The first describes everything the company has and gives no counts or samples. The second is smaller, says exactly how many records exist by type, how many years they cover and offers three complete examples. The second one gets read. The first one is often passed over, even when it holds the more interesting records, because the lab has to do the packaging work itself and has plenty of offers where that work is already done.

This is also why a broker is useful. Part of the job is turning a description into a package a lab can evaluate in minutes.

The five facts to write down

1. What the records are. Name the kinds, not the category. "Support tickets with the full reply thread" is useful. "Customer data" is not.
2. How many of each. A rough count is enough: tickets, projects, calls, documents, messages. If you have several kinds, count each kind. Round numbers are fine, and "I am not sure" is better said than guessed.
3. How many years, and from when to when. Long, continuous history is worth more than scattered years, so say whether there are gaps.
4. Where they live. The system or file format they sit in, and whether you can export them whole. A lab thinks about how much work it takes to read your records, so this counts.
5. Who owns them and who is in them. Are they your own records, or do some belong to clients under contracts? Do they contain customer names, health details or voices? You do not need to solve this before the call. You need to know the answer is not yet known.

The sample

If you can prepare one thing, prepare this: one to three complete examples. Complete means the whole story of one unit of work, not a screenshot of a field. For a support team it is one ticket from the first message to the closing note. For an engineering firm it is one finished project with its files. For a finance team it is one month's close with the review notes.

Two cautions about samples:

• Do not send it yet. The sample is a thing you have on hand. Personal details have to be taken out before it leaves the company, and your rights to it have to be checked. We do both together, and that is a conversation, not a form.
• Pick an ordinary one. A hand-picked best case misleads a lab and costs trust later. Choose a typical example, and say it is typical.

What you do not need yet

You do not need a price in mind. No one can price your data without seeing it, and any figure quoted on the first call is a guess. The data value calculator gives a rough range in about a minute if you want orientation.

You also do not need to know whether your data is "AI-ready". The test is whether a person could open one record and understand what happened. If they can, there is something to look at.

What to expect after the call

If the facts above are in hand, the next step is a plain review. We look at what you hold, what has to be removed, who has to agree and which labs might want it. You keep the final decision on every term. The page for sellers describes how a deal runs, and how you get paid explains who pays whom and when.

If you are not sure whether an offer you already received is reasonable, check an offer lists the questions to ask. If the records contain personal information, the de-identification page explains what is removed and how it is recorded. Terms such as data room, sample or license are in the glossary.

When your list is ready, tell us what you have. Even an incomplete list is a good start, because the missing items tell us what to find out first.