How to Prepare Your Company's Data for AI Licensing

Step 1: Take an inventory

Start with a simple list of the systems your business has used over the years, including ones you no longer actively use. For each, note roughly what it contains, how far back the history goes and who in the company knows it best. You do not need to open or export anything yet.

Include systems you have stopped using. An old project tracker or a retired support tool may hold years of history that nobody has looked at since a migration, and older records are sometimes more complete than newer ones.

• Code: GitHub, GitLab, Bitbucket or self-hosted repositories, plus CI systems.
• Planning and documentation: Jira, Linear, Confluence, Notion, Google Drive, SharePoint.
• Support and communication: Zendesk, Intercom, Freshdesk, Slack, Teams, shared mailboxes.
• Business systems: CRM, ERP, accounting software, databases and data warehouses.
• Process material: SOPs, training guides, templates and checklists.

Step 2: Check ownership and restrictions

For each item, ask whether your company owns it outright. Client deliverables, client-supplied data, partner materials and third-party content may come with contractual limits. Personal data brings privacy obligations that depend on where you and the people in the data are located.

Mark anything uncertain for review. This is general information rather than legal advice; a qualified lawyer should look at your specific contracts and the privacy laws that apply to you.

A simple traffic-light system works well at this stage. Green for material you clearly own with little personal data, amber for material that needs a closer look, and red for anything you already know you will not include. This gives you and any partner a quick view of where to focus.

Step 3: Have an early conversation

It helps to talk to a partner before you invest in exports, because not everything in your inventory will be equally useful. With Momentra, first contact needs only contact details, company, industry and interests. The initial value assessment then discusses data type, completeness, scarcity and the tasks your data could support, which helps you focus effort on the material that matters.

This step also helps you understand the likely partnership model, whether licensing, purchasing or a custom project, before you commit internal time. That can affect how much of your history you choose to prepare.

The most useful data usually captures a complete workflow: the objective, the tools used, the decisions made and a verifiable outcome. Use that as a lens when you decide where to focus.

Step 4: Agree boundaries before sharing samples

Before any sample leaves your hands, sign an NDA that defines confidentiality, permitted use, retention and security. Then pick samples that are representative but low-risk, so the review can judge quality without exposing your most sensitive material. Keep a note of exactly what was shared as a sample.

A sample does not need to be large. A handful of complete examples, such as a few tickets with their full history or one repository with its linked issues, usually says more about quality than a large volume of disconnected records.

Step 5: Export with history and links intact

When the scope is agreed, export in a way that keeps context. A repository snapshot without its commit history loses much of its value. Tickets exported without their comments, status changes or links to code are much less informative. Where possible, keep identifiers that connect records across systems, such as ticket numbers referenced in commit messages.

Record what you export: the system, the date range, the filters applied and the format. This becomes part of the documentation that goes with the dataset.

If a system only lets you export current data rather than history, note that limitation rather than working around it in a hurry. Some tools offer separate audit or activity exports that fill the gap, and it is better to discuss the options than to deliver something incomplete without saying so.

Step 6: De-identify and review

Sensitive information must be detected and removed or replaced before further use. Momentra supports this with professional anonymization using purpose-built tools, covering things like personal details, credentials, customer identifiers and internal infrastructure references. Your team should still review samples of the output, because you know which names, codes and project terms are sensitive in your business better than any tool.

• Give the de-identification process a list of terms you know are sensitive: client names, product codenames, internal domains.
• Spot-check output from every source system, not just the largest.
• Confirm that links between records still work after identifiers are replaced consistently.

Step 7: Document and hand over

Finish with a short written description of the dataset: what systems it came from, date ranges, what was excluded and why, and any known gaps. This helps the receiving side use the data correctly and gives you a clear record that matches the signed agreement, which governs scope, permitted use, transfer and processing. Transfer itself should follow a secure, agreed method with a manifest of what was sent.

Keep a copy of this description with your signed agreement and the transfer manifest. If questions come up months later about what was included, these three documents together give a clear answer, and they make any future partnership quicker to set up because much of the preparation can be reused.