AI inventory with Excel: How it works and what are the alternatives

An AI inventory is a directory which documents all AI use cases within a company. It serves the purpose of creating transparency about all AI assets and their risks and forms the starting point for compliance with the requirements of the AI Act. In times of increasing ethical and legal standards for AI, it is crucial to maintain an overview of all AI systems in use. A well-maintained AI inventory also provides a valuable basis for internal audits and external compliance reviews.

Excel can be a simple and cost-effective method for creating an AI inventory. Structured tables can be used to document and store all relevant information about AI use cases. 

Below, we explain what information should be included in the Excel AI inventory and how you can maintain a comprehensive overview with the structure of the inventory.

For each AI use case, you should record a basic description and the responsible department. The respective contact person responsible for the AI use case should also be named. Use columns in Excel to clearly structure the AI assets and their responsibilities.

To do this, create the following cells in the Excel document: 

• Name of the AI asset
• Description of the AI asset
• Responsible department
• Contact person
• Risk owner

Next, the intended purpose or "use case" of an AI asset must be described. This refers to specific applications of artificial intelligence in which AI technology is used to solve a specific problem or create specific added value in a specific business process or context. It is entirely possible for an AI asset to have several different use cases.

In addition, the specific responsibility for the use case within the company should be documented.

To do this, create the following columns in the Excel document:

• Intended purpose (AI use case)
• Responsible department
• Contact person

Next, it is important to note the role your company plays, taking into account the requirements of the AI Act in Article 3. You should also assign a risk class to your use cases.

To do this, create additional columns with the following information:

• AI role: Selection field with the following possible entries
  • Provider
  • Deployer
  • Authorised representative
  • Importer
  • Distributor
• AI risk class: Selection field with the following possible entries
  • Low
  • Restricted / Limited
  • High
  • Unacceptable

A detailed risk assessment can be understood as part of the AI inventory. Since this requires in-depth technical and legal analysis, it is advisable to carry out this step only after completing the initial AI use case registry.

For risk assessment or evaluation, a simple process is recommended, in which potential risks are first identified and described. Mitigating measures are then documented and, finally, the probability of occurrence and the amount of damage caused by the risk are assessed.

To do this, create the following columns in your Excel list:

• Description of the risk
• Description of mitigating measures
• Probability of occurrence
• Extent of damage

Note: The rating scales for probability of occurrence and severity of damage are not specified for AI risk management. One possible scale is a 4-level classification into "low," "limited," "high," and "very high."

When identifying AI risks, you can use protection goals to perform as comprehensive an assessment as possible. The use of protection goals is already common practice in information security.

Since protection goals cannot be conclusively derived from the AI Act and the technical discourse among experts is still ongoing, we provide you with a selection of possible protection goals for your AI use case registry.

To do this, create a column in the Excel document with the following label:

• Protection goals: Selection field with possible entries:
  • Accuracy
  • Robustness
  • Bias/Discrimination
  • Confidentiality
  • Integrity
  • Availability
  • Purposefulness
  • Authenticity

It is possible to create an AI inventory using Excel, albeit somewhat cumbersome. In fact, there are three specific advantages to implementing an AI Inventory with Excel.

1. Familiarity:

   You don't have to learn how to use new software: Most people have probably used a spreadsheet program like Excel at some point. This means that you don't usually need any training time and can get started right away with an AI inventory in Excel.

2. Cost savings:

   Excel is usually included in Office packages and is generally pre-installed. This saves the company additional licensing costs.

3. Flexibility:

   You have a great deal of flexibility in designing your AI inventory: Excel-based solutions are almost universally applicable and can be tailored to any company, set up in a variety of ways, and adapted at any time.

On the other hand, the advantages of Excel as an AI inventory tool are offset by a number of disadvantages.

1. No links

   Although you can create different tables for different data sets, linking them together is often difficult. This means you end up doing double the work. For example, AI assets must also appear in your general asset list. If something changes in one list, you must also manually change the other list. The same problem arises when AI also processes personal data. It is almost impossible to link to your directory of processing activities with Excel.

2. Manual work and confusion

   An Excel document can be a good start for getting an initial overview of your AI assets. In the long term, however, Excel poses the challenge that every time you change columns and rows, you will have to reformat everything. With an AI inventory, it can be assumed that updates will occur regularly.

3. No automation

   You are wasting the potential of automated processes: These include, for example, automatic notifications when someone makes changes to the document, as well as the automated entry of AI assets.

4. Complicated evaluation

   Another challenge is evaluation: an AI inventory that grows organically in Excel is difficult to simplify in such a way that you have an immediate overview of tasks. Even a dashboard with all the important information as an overview is difficult to implement in Excel. In addition, an evaluation can become incorrect without you noticing if rows or columns change.

5. Lack of quality control and versioning

   An AI inventory provides the basis for compliance with your compliance requirements. The quality of the documentation in this document should be of the highest standard. Excel does not provide you with an overview of the status quo, nor does the program check the accuracy and scope of your entries. It is also difficult to track changes in individual AI use cases in detail.

6. Lack of collaboration

   No single person can have an overview of all AI assets or AI use cases in the company at all times. New AI systems are constantly coming onto the market, and constant input from other areas of the company is needed to stay up to date. With Excel, you make it difficult for other colleagues to participate in the documentation process or to collaborate directly in the document.

To simplify the documentation process and maintain better control, a specialized software solution is recommended. Such tools offer clear structures and automation that reduce the documentation effort while ensuring compliance requirements are met.

You can expect the following features from professional AI compliance software:

• Predefined input fields for your AI inventory
• Preselection of AI roles, AI risk classes, and responsible departments
• Linking of AI assets to your general asset directory

• Prepared questionnaires for identifying AI assets and use cases
• Sending questionnaires to the workforce, including follow-ups
• Notification when software tools or assets that may use AI are documented

• Link AI assets to business processes or processing activities
• Identify AI use cases based on the threshold analysis for Data Protection Impact Assessments
• Use existing risk templates for AI risk management

• Assign responsibilities for each AI asset or use case
• Task and comment function
• Customizable role concept for cross-departmental collaboration
• Automatic email notifications for changes

An AI inventory with Excel provides an initial overview and can be practical for small businesses. However, for more comprehensive and efficient management, specialized AI compliance software is often the better choice, as it not only offers structured documentation but also efficiency gains through automation.

You are welcome to test caralegal's AI compliance software free of charge. Our team will guide you through all the features in a personal appointment and address your challenges. We are confident that our solution will win you over.

If you still want to start your AI compliance journey with Excel, we have prepared a free template for you.