Starting on the AI journey, some things will be familiar, and some will be different. As with any major program or investment/ project, the objective needs to be clearly defined. In other words, describe what will be different when the new capability is launched. Next you need to identify the data that will be needed. Here the same considerations as for any other project are applicable including a) what type of data, customer or system data; b) who owns the data; c) does it fall under any regulation (e.g., privacy such as HIPAA and GDPR, Health Authority); d) will the data be transferred; e) how often will it be refreshed; f) who will have access etc.  If a third party is engaged in this effort (which is likely) what are the respective roles and responsibilities.

None of this is different from what we are used to. However, preparing the organization and to ensure that they have the skills and support that can provide the required oversight will be different from a “legacy” project.

There are other pitfalls and limitations such as oversimplifying the objective, interpretability or a black box system where the desired transparency is not available, data quality and availability, ethical concerns, the so-called hallucinations and the importance of separating the source of input from the source of truth and the role of the human in doing so. The need to monitor, refine and reassess is an ongoing need. It will be a transition for team members to learn that new mindset.

Generative AI does not get us off the hook. It just became more challenging ensuring we have the right controls and understanding. We need to make sure our QMS evolves accordingly. But if we can get this right, then we can realize outcomes that in the past seemed impossible.