• Written by Sonali Subhadarsini, Product Manager

A category manager in a pharmaceutical company is likely to have experienced the frustration of trying to get timely answers from their organisation's data, only to be held back by the strict documentation rules and validated processes that govern their industry. The shift supervisor on the production line may need to check the maintenance log for a specific piece of equipment, but the process of doing so is slowed down by the need to ensure that all changes are properly documented and approved. This is just one example of how the pharmaceutical industry's unique constraints can make it harder to implement new technologies, including enterprise AI governance.

Pharmaceutical industry constraints

The pharmaceutical industry is subject to a range of strict regulations and guidelines, including those related to change control, batch record review, and deviation investigation. These regulations are in place to ensure the quality and safety of pharmaceutical products, but they can also make it more difficult to implement new technologies. For example, the overhead of change control can make it harder to update systems and processes, while the need for batch record review can slow down the process of getting products to market. Deviation investigation time can also be lengthy, as all deviations from standard procedures must be thoroughly investigated and documented.

These constraints can make it harder for pharmaceutical companies to adopt new technologies, including enterprise AI governance. The need for strict documentation and validation can make it more difficult to implement AI systems, and the risk of non-compliance can be high. However, this does not mean that pharmaceutical companies should avoid using AI altogether. Instead, they need to find ways to implement AI governance in a way that meets the unique constraints of their industry.

Enforcing permission at retrieval

One way to address the constraints of the pharmaceutical industry is to enforce permission at retrieval, rather than by instructing the model. This means that users can only access the data and information that they are authorised to see, and that all access is properly documented and audited. This approach can help to reduce the risk of non-compliance, while also making it easier to implement AI governance. For example, a shift supervisor may need to access the maintenance log for a specific piece of equipment, but they should only be able to see the information that is relevant to their role.

This approach requires a deep understanding of the organisation's data and systems, as well as the roles and responsibilities of different users. It also requires a robust system for enforcing permissions and auditing access, to ensure that all access is properly documented and compliant with regulatory requirements. By enforcing permission at retrieval, pharmaceutical companies can reduce the risk of non-compliance and make it easier to implement AI governance.

Auditability

Auditability is a critical component of enterprise AI governance in pharmaceuticals. All access to data and information must be properly documented and audited, to ensure that the organisation can demonstrate compliance with regulatory requirements. This includes tracking what was asked, what was returned, and on what data. For example, if a category manager uses an AI system to analyse sales data, the system should track what questions were asked, what answers were provided, and what data was used to generate those answers.

This information can be used to demonstrate compliance with regulatory requirements, and to identify any potential issues or risks. It can also be used to improve the performance of the AI system, by identifying areas where the system may not be providing accurate or relevant answers. By tracking all access to data and information, pharmaceutical companies can ensure that their AI systems are transparent, auditable, and compliant with regulatory requirements. For more information on how AsscherAi can support auditability, please visit our website.

Routing to a human

There are some categories of questions that should always be routed to a human, rather than being answered by an AI system. These include questions that require a high level of judgement or expertise, or that involve sensitive or confidential information. For example, a question about a specific patient's medical history should always be routed to a human, rather than being answered by an AI system. This is because the answer to this question requires a high level of judgement and expertise, and involves sensitive and confidential information.

Other categories of questions that may need to be routed to a human include those that involve complex or nuanced issues, or that require a deep understanding of the organisation's systems and processes. By routing these questions to a human, pharmaceutical companies can ensure that they are answered accurately and appropriately, and that the organisation is not exposed to unnecessary risk. If you would like to discuss how AsscherAi can support routing to a human, please contact us.

Limitations of enterprise AI governance

While enterprise AI governance can be a powerful tool for pharmaceutical companies, it is not a solution to all problems. There are some situations where AI governance may not be the best choice, or where it may not be effective. For example, AI governance may not be suitable for situations where the data is highly unstructured or variable, or where the questions being asked are highly complex or nuanced. In these situations, other approaches may be more effective, such as using human analysts or experts to answer questions and provide insights.

Additionally, AI governance is not a replacement for human judgement and expertise. While AI systems can provide fast and accurate answers to many questions, they are not a substitute for human analysis and decision-making. Pharmaceutical companies should always use AI governance in conjunction with human judgement and expertise, to ensure that they are getting the most accurate and relevant answers to their questions. By understanding the limitations of enterprise AI governance, pharmaceutical companies can use it more effectively, and avoid potential pitfalls and risks.