Confidential Data AI Assessment — Organizing how your data is handled, before you hand it over

Challenge

Contracts, meeting minutes, customer records, quotations, internal knowledge. We often hear that a company holds plenty of information that looks like it could be put to work with AI, yet nobody can judge whether it is acceptable to hand that material to an external cloud AI as it stands. Time passes without a clear place to start, and consulting a vendor tends to turn into a conversation about a large development project from the outset. This assessment is for the executives and back-office teams of small and mid-sized companies in exactly that position.

Approach

What Cor. Inc. values is the idea of designing how information is handled before using AI. We do not start by building. Every piece of information carries its own level of sensitivity, and it divides into material that may go to an external cloud AI and material that should be handled by a local LLM running on your own side, without sending data outside the company. We begin by working through that distinction together. Sometimes the assessment arrives at the conclusion that it is better not to build anything yet, and we tell you so plainly, that outcome included.

Implementation

The assessment proceeds as follows.

• Interviews on site to understand the work and where the difficulties lie
• A survey of existing operations and an inventory of tasks that could be handled by AI
• Classification of the data involved (organizing it by sensitivity)
• Separating what belongs on cloud AI from what belongs on a local LLM
• Organizing the risks (four design perspectives: input boundaries, access permissions, operation logs, and human approval)
• Setting a direction on whether to proceed to a PoC

Rather than opening with a conversation about technology, we give priority to putting into words what should be handled and how.

Results

When the assessment ends, you are left with material you can use for your next decision: the inventory of candidate tasks for AI, the data classification and how each class should be handled, the risks and the policies for addressing them, and a draft plan for a PoC if you go on to one.

From there, three paths are open. You can move to a local LLM / AI platform PoC (from ¥3M over three months, varying with the problem and the scope), move to a small production implementation, or decide that now is not the time to build. Whichever you choose, the assessment itself is designed to serve as material for deciding your next step. The fee is roughly ¥100k–¥300k (varying with the problem and the scope). Start with a conversation.