• Written by Azhar Aliyar, Director

The weekly sales report on the dashboard of a large retail organisation's intranet is often left unopened, not because the data is unimportant, but because the figures are presented without any context or source, making them unreliable for decision-making. What is citation in AI answers is a direct reference to the original data source that the AI system used to generate the answer, in this case, the sales data from the enterprise resource planning system.

Definition of Citation in AI Answers

Citation in AI answers is the process of providing a reference to the original data source that the AI system used to generate the answer. This can be a link to a database table, a document, or even a specific cell in a spreadsheet. The citation provides transparency and accountability, allowing users to verify the accuracy of the answer and understand the context in which it was generated.

For example, a category manager comparing two suppliers in the procurement system may ask the AI system which supplier has the best delivery track record. The AI system may respond with a percentage, but without citation, this is simply a claim. With citation, the AI system can provide a link to the purchase order table, allowing the category manager to see the actual data and verify the answer.

Difference from Related Concepts

Many people confuse citation in AI answers with data provenance, which refers to the history of how the data was created, processed, and transformed. While related, these concepts are distinct. Data provenance is about understanding the lineage of the data, whereas citation is about providing a direct reference to the original data source.

Others may think that citation is the same as data validation, which is the process of checking the accuracy and consistency of the data. However, citation is not about validating the data, but rather about providing transparency and accountability. A cited answer can still be incorrect, but at least the user can understand where the answer came from and verify it themselves.

Applications in Large Organisations

In a large organisation, citation in AI answers is particularly important when dealing with complex data sets and multiple stakeholders. For instance, a shift supervisor checking the maintenance log may ask the AI system when the last maintenance was performed on a specific machine. The AI system can provide a cited answer, referencing the maintenance log database, allowing the shift supervisor to verify the answer and take appropriate action.

Citation is also crucial in regulated industries, such as finance and healthcare, where transparency and accountability are paramount. By providing cited answers, organisations can demonstrate compliance with regulatory requirements and reduce the risk of errors or misinterpretation.

Limitations of Citation in AI Answers

While citation in AI answers provides transparency and accountability, it does not guarantee accuracy or completeness. A cited answer can still be incorrect or outdated, and users must still verify the answer themselves. Additionally, citation may not always be possible, such as when the data source is proprietary or confidential.

Organisations should also be aware that citation in AI answers can add complexity to the system, requiring additional infrastructure and maintenance. For example, the AI system may need to be integrated with multiple data sources, each with its own authentication and authorisation mechanisms. To learn more about how AsscherAi can help with citation in AI answers, visit our website.

When Citation is Not the Right Choice

In some cases, citation in AI answers may not be the right choice, such as when the data is highly sensitive or confidential. In these cases, organisations may need to implement additional security measures, such as encryption or access controls, to protect the data. Alternatively, organisations may need to consider alternative approaches, such as using aggregated or anonymised data, to balance the need for transparency with the need for security.

For organisations that are unsure about how to implement citation in AI answers, or have questions about how AsscherAi can help, please contact us to discuss your specific needs and requirements. By understanding the limitations and potential applications of citation in AI answers, organisations can make informed decisions about how to use this technology to improve their operations and decision-making processes.