Data licensing for US companies with 20+ employeeshello@getdataoffer.com

Getting started

What happens after you accept a data offer

What happens between accepting an AI lab's offer and getting paid: contract, export, de-identification, your review, delivery and deletion.

A hand signing a printed agreement with a fountain pen on a walnut desk

Getting an offer is the exciting part. What comes next is mostly careful, unglamorous work: finalizing the contract, exporting and preparing the data, checking it, and handing it over securely. Knowing the sequence in advance helps you plan your team's time and avoid surprises.

This guide walks through a typical post-offer process. Every deal is different, and the agreement you sign is what actually governs the steps, but most follow a similar shape.

Step 1: Choose the offer and confirm the scope

Accepting an offer usually starts with a non-binding acceptance or a short term sheet: buyer, price, structure (license or transfer, exclusive or not), permitted uses and a high-level description of the dataset.

Before anything is signed, the scope should be written down precisely:

  • Which systems, and which parts of each (channels, mailboxes, folders, record types)
  • Date ranges
  • Exclusions, named explicitly
  • Format and structure of the delivered data
  • The de-identification standard to be applied

A precise scope document protects both sides. Disputes about "what was promised" almost always trace back to a vague one.

Step 2: Negotiate and sign the agreement

This is where your counsel earns their fee. Key terms include the grant of rights, security and access requirements, prohibitions on re-identification, retention and deletion obligations, representations and warranties, liability caps and indemnities, and payment terms. Our licensing agreement checklist goes through each clause.

Two things deserve particular attention at this stage:

  • What "delivered" and "accepted" mean. If payment depends on acceptance, the agreement should define the acceptance criteria and a time window, so the buyer can't hold payment open indefinitely.
  • Your representations. You'll usually confirm you have the right to license the data. Make sure the scope and exclusions support that.

This isn't legal advice. Have counsel review the agreement before you sign.

Step 3: Export

With a signed agreement, the export begins. Depending on the arrangement, your team runs it, the partner's engineers run it with scoped access you grant, or both work together.

Good practice at this stage:

  • Use scoped, temporary access. Create dedicated admin accounts or tokens limited to what's needed, and revoke them when the work is done.
  • Export into a controlled environment, not someone's laptop.
  • Log what was exported: systems, date ranges, counts.
  • Apply exclusions at export time where possible, so excluded data never leaves its source system.

How long this takes depends on the volume and the systems. API-based exports of large histories can take days to run in the background, but they rarely require much of your team's attention.

Step 4: Clean, structure and de-identify

Raw exports get turned into a consistent dataset:

  1. Cleaning: removing duplicates, bot and automated content, broken records and empty files
  2. Structuring: consistent formats across systems, threads and relationships preserved, metadata normalized
  3. De-identification: replacing people, companies and identifiers with consistent pseudonyms, scanning free text and handling attachments, according to the agreed standard
  4. Documentation: a data card or description of what's included, how it was processed, and known limitations

See our de-identification guide for the techniques.

Step 5: You review a sample

Before anything goes to the buyer, you should see a representative de-identified sample, ideally drawn at random rather than hand-picked. Things to look for:

  • Can you recognize any employee, customer or vendor?
  • Did any excluded content slip through?
  • Is anything sensitive in a place you didn't expect (attachments, signatures, file names)?
  • Does the data still make sense as a record of work?

If something's wrong, the processing is adjusted and you review again. Nothing should move until you're satisfied.

Step 6: Secure delivery

Delivery typically happens by secure transfer into the buyer's environment, for example an encrypted storage bucket with restricted access, or a secure file transfer. The agreement may specify encryption, access controls and who at the buyer can access the data. You should receive confirmation of what was delivered and when.

Step 7: Buyer acceptance and payment

Buyers usually run their own checks: format, completeness and quality against the agreed spec. Payment then follows the agreed terms, which may be a single payment on delivery or acceptance, or tranches tied to milestones. Fees to any broker are typically handled as described in your agreement with them, so confirm how and when they're paid.

Step 8: Cleanup and deletion

After delivery, the working copies created during processing should be deleted according to the agreement, and access granted for the project should be revoked. Ask for written confirmation. The buyer's own retention and deletion obligations continue under the license terms.

Step 9: Future deliveries (if any)

Some agreements include options for additional data later, for example another year of history. If yours does, the same scope, processing and review steps apply each time, and you should still approve each delivery.

How long does it all take?

It depends on the volume of data, the number of systems, how quickly contracts get negotiated and how fast your team reviews samples. Contract negotiation is often the longest step. A focused, well-scoped dataset moves faster than a sprawling one.

What your team actually has to do

For most companies, the internal effort is modest:

  • An executive to approve the offer and sign
  • Counsel to review the agreement
  • An IT admin to grant scoped access
  • Someone who knows the business to review the de-identified sample

Getting started

If you haven't received an offer yet, the first step is simple: share rough estimates of which systems you use, how much data is in each, years of history, headcount and years in business.

DataOffer handles the process end to end, from packaging and de-identification (or supporting your team) to competing offers and delivery. Our engineers can do the export and cleanup, and nothing is shared until you approve the buyer, price and terms.

Ready to see what your data is worth?

Share rough estimates (systems, approximate volume, years of history, headcount) and we'll come back with competing offers from AI labs. No upfront cost, no commitment, and nothing is shared until you approve.

This guide is general information, not legal, tax or financial advice. Figures and ranges are illustrative; talk to qualified advisors about your situation.