Data storage is an integral aspect of implementing digital solutions such as AI. The system requires a database to access stored data, and this database should be large enough to ensure that the AI solution works optimally and takes all factors into account to propose, predict and perform its activities. On top of the storage itself, there are two other important factors: ensuring that existing data is captured, and that data is governed via lifecycle management.
An AI system cannot work without data, therefore historical data needs to be stored and made easily available. Depending on the process and its optimisation, the data can be already digitised, which simplifies the task. Whatever the case, it is imperative that the data be accurate and consistent. To ensure this, an exercise in data collecting, assessment and sorting should be conducted before it is fed into the digital tool.
Once the system is running, the data grows exponentially and requires effective governance practices to ensure its quality, security, and compliance. This involves assigning responsibility for data management to specific individuals or teams within the organisation. In many companies, this translates into a growth of the IT compliance department, which creates the need for a clear operating model enabling better interactions with other departments, including the quality team as they are responsible for controlling the compliance of systems and the impact on products and patients.
As mentioned, digital systems require proper management and dedicated personnel with sufficient knowledge of these systems.
One of the key concepts that should be discussed here is the Human-in-the-loop (HIL). This concept refers to a process where human intelligence and machine intelligence work together; a combination leading to more accurate, reliable and efficient outcomes. The human aspect may relate to the decision-making process, by providing input and oversight. In the pharmaceutical industry, where the outcome of critical process steps must be tested and controlled, the HIL concept ensures that human expertise is incorporated in quality control and decision-making processes, allowing accountability and error detection.
Personnel is an important part of GxP, nevertheless there is no requirement yet for specific roles having a certain level of knowledge and understanding of digital technologies.
Digital solutions are not predefined, and there is no standard way of working with them. Depending on the system, steps, data, technology and the team, there are different ways to use digitisation in a process. This aspect of industry 4.0 allows a lot of flexibility in searching for the best solution. On the other hand, it implies that there is no standardised guidance available with clear terminology and definitions. Discussions on AI often resort to self-defined terminologies within companies. A common standard to bridge the gap between IT and GxP terminologies and align them with regulators would aid the discussion going forward.
Acknowledging these challenges and fully taking them into consideration are the first steps in the digitisation journey. In the second phase, the opportunities for improvement need to be assessed to define which (Gen) AI solutions might be best implemented to optimise the efficiency and/or quality of processes and/or products. In a quality environment, these solutions might be related to documentation management, predictive quality (e.g. deviations), authorization management, etc.
At PwC, we have put together a digital quality framework to assess a company’s digital maturity, define a digital strategy, select potential (Gen) AI tools and support their implementation.
Morgane Franck