PAT in Practice:
Why Timing, Data Integration and Mindset Are Decisive
Process Analytical Technology in Development and Manufacturing
Technology is rarely the problem; decisive factors are timing, integration, and organizational anchoring
Experience from chemical active pharmaceutical ingredient development and pharmaceutical manufacturing
Interview with Dr. Matthias Balsam
Bayer AG, conducted on Jan. 9, 2026
Process Analytical Technology (PAT) is once again moving into sharper focus in the manufacture of solid oral dosage forms. However, the step from technically robust measurement to standardized industrial application remains demanding. In an interview with the Technology Training Center at Glatt, Dr. Matthias Balsam, PAT expert at Bayer AG, explains the role that the right timing, integration into the data landscape, and a truly lived PAT mindset play in successful projects.
Technically Feasible, Operationally Demanding
“In my view, measurement-related questions can in many cases be solved with reasonable effort,” says Balsam. “Complexity arises where a PAT measurement cannot stand alone as an isolated solution.” As soon as data connectivity, data harmonization, and GxP compliance come into play, project effort increases noticeably. This also increases the need for thorough planning, particularly at the interfaces between the PAT system, the process control system, and the data architecture.
The technical hurdle is therefore shaped less by sensor technology and analytics, and more by integration. A measurement only creates full potential when process data and PAT data can be used together, for control, release, or at least for a consistent understanding of the process.
Timing Is Often the Decisive Factor
Balsam sees the crucial gap between technical feasibility and established industrial practice primarily in choosing the right point in time for implementation: If a PAT initiative only starts once a process has been fully validated, almost everything becomes more difficult. The immediate mechanical question arises: Can a probe still be integrated without modifying qualified equipment? At the same time, an established at-line is often already included as in process control (IPC) in the dossier. A later change to online or inline measurement must then be justified not only technically, but also from a regulatory and economic perspective.
With spectroscopic methods, an additional aspect comes into play. Calibration models require variation and suitable data conditions. In a stable, frozen process, these degrees of freedom are often lacking. As a result, model development becomes either lengthy or remains unsatisfactory. Consistent early integration, already during process development, reverses this logic. Measurement feasibility becomes visible early on. At the same time, data flows directly back into development and accelerates decision making.
From Pilot Project to Routine: Organization and Mindset
According to Balsam, the difference between large corporations and smaller companies lies less in willingness and more in structure. In large organizations, specialized PAT groups often exist that carry projects and support them throughout the entire lifecycle. Especially for more complex applications, PAT is rarely plug and play. It requires application development, adaptation to the process, and continuous support.
In smaller companies, this know how is often purchased externally. This creates hurdles such as limited availability, interface effort, and reduced continuity. At the same time, even in large companies, PAT remains a special project if there is no clear PAT mindset, one that is questioned at the first budget review.
“We have a saying: PAT is not a cheap hobby. It all costs a bit, and that’s not just time and resources, but also money in the end.” (Dr. Matthias Balsam, Bayer AG)
Balsam links mindset to a clear understanding of value. The focus is not solely on short term cost savings, but on robust process understanding. This process understanding forms the basis for optimization, troubleshooting, and stable production. As a result, the discussion shifts away from a pure one-year ROI toward the ability to control processes in a targeted manner and detect deviations early.
Looking Ahead: AI and Competitive Pressure
For the coming years, Balsam sees two main drivers. First, global competition is increasing, along with pressure on costs, personnel, and resource utilization. PAT can help shift laboratory analytics into the process, reduce production times, and enable earlier decision making.
Second, the use of AI in the process industry will continue to grow. Balsam warns against the assumption that arbitrary amounts of data automatically lead to good models. AI requires meaningful, process relevant data. PAT delivers exactly these data and thus becomes a key building block, both as a data source and, prospectively, through simplification of model development and data preparation.
A Practical Example from API Development
“Working in chemical drug development had a formative influence on me,” says Balsam. “I have been supporting development projects there with PAT for several years. Initially, the goal was to accelerate process development based on data.” Over time, a recurring pattern emerged: the measurement initially intended for process understanding evolves along the scale up path into IPC. Later, the use of this IPC via PAT is no longer questioned in principle, because the measurement is well established and familiar. This makes it clear: an early start and an actively exemplified PAT mindset are crucial. An additional practical advantage emerges. Models, for example in NIR analytics, grow together with the process. In operation, this avoids having to start from scratch under tight constraints.
OSD and Biologics: Different Necessity, Same Standard
For classical solid oral dosage forms, a stable process can often be achieved with little PAT. In this case, PAT primarily delivers efficiency gains, for example through faster in process controls or shortened process times. Comprehensive control based on process models plays a smaller role in many cases, apart from continuous processes.
In biological processes, the situation is fundamentally different. Here, operation is often based on a process model built on critical process parameters and requiring real time monitoring. In this context, Balsam describes PAT as a prerequisite for operation. Without this monitoring, the process can practically not be run.
What a Conference Can Contribute
What options are there for making PAT expertise from specialist circles accessible to a broader audience? Balsam emphasizes that traditional PAT conferences provide valuable opportunities for exchange among experts, but notes that these events are often only relevant for insiders. Things become truly interesting when plant operators, equipment manufacturers and users work together and present concrete, practical solutions.
These act as a catalyst for new ideas. They demonstrate that technical problems are usually solvable, provided that the required effort is planned in advance. It also becomes clear why it is helpful to integrate PAT into development as early as possible and to focus on a solid data strategy right from the start. This significantly facilitates subsequent operation.
Conclusion
PAT has rarely failed because of sensor technology. The decisive stumbling blocks lie in integration, compliance, and above all in the starting point. Those who integrate PAT early into development build data foundations, models, and trust. In this way, a measurement becomes a tool that makes processes more stable, faster, and more transparent.