Confidential Data AI Utilization Assessment — Organizing how to handle your data together before putting it out there

Challenges

Contracts, meeting minutes, customer information, price quotes, and internal knowledge. We frequently hear from companies that while they have plenty of internal information that could be leveraged with AI, they cannot determine whether it is safe to send to external cloud AIs. Time passes without knowing where to start, and consulting a vendor immediately leads to talks of large-scale development from day one. This assessment is designed for business leaders and administrative department managers at small and medium-sized enterprises facing these exact situations.

Our Approach

At Cor., we value the philosophy of “designing the information to be handled before using AI.” We do not jump straight into building. Information comes with varying levels of confidentiality; it is divided into what can be sent to external cloud AIs and what should be handled by local LLMs running on-premises without letting data leave the company. We begin by organizing this distinction together. If the assessment leads to the conclusion that “it is better not to build yet,” we will communicate that conclusion with complete impartiality.

Implementation

The assessment proceeds according to the following flow:

• Understanding operations and pain points through on-site interviews
• Investigating existing workflows and taking inventory of tasks that can be automated with AI
• Classifying data to be handled (organizing confidentiality levels)
• Segmenting between cloud AI and local LLMs
• Outlining risks (based on four design perspectives: input boundaries, access permissions, operation logs, and human approval)
• Formulating a policy on whether to proceed to a PoC

Rather than starting with technical discussions, we prioritize articulating what should be handled and how.

Deliverables

Upon completion of the assessment, you will receive documentation you can use for subsequent decisions. The contents of this documentation include an inventory of candidate tasks for AI adoption, classification and handling policies for data, risks and mitigation policies, and a proposed direction should you decide to proceed with a PoC.

Following the assessment, there are three options: proceed to a local LLM/AI platform PoC (from 3 million yen for 3 months, varying depending on the challenges and scope), proceed directly to small-scale full implementation, or decide not to build anything for now. Whichever path you choose, the assessment is designed to serve as material that helps you determine your next step. The fee is approximately 100,000 to 300,000 yen (varying depending on the challenges and scope).