At PAT Connect 2026, Markus Gruell from Autolus offered a clear view of what this reality looks like inside a commercial CAR T organisation. Autolus manufactures clinical and commercial autologous Drug Products in the United States, the United Kingdom and the European Union and treats patients with acute lymphoblastic leukaemia, many of whom have no therapeutic alternatives left.

How PAT shapes manufacturing decisions

Autolus’ manufacturing process takes six to seven days on average, followed by roughly another week of quality control testing. When cryogenic transport between continents is included, the full turnaround time reaches about nineteen days. Since each batch corresponds to one individual patient, every day represents clinical risk rather than operational inefficiency.

In this context, PAT is not simply an optimisation tool. It changes when decisions can be made. As Markus explains, “PAT would allow us to manufacture quicker and certify the batches quicker and ship them back to the patients in the hospital.” Autolus is continuously working on reducing the length of  its manufacturing process, currently aiming for  four days. Earlier insight into process performance would not only shorten timelines but also reduce the risk of failed batches, which cannot be recovered in personalised medicine.

Working with regulators – not just for them

The EMA’s regulatory sandbox, a structured programme for early dialogue between innovators and regulators on novel technologies, has created significant interest in the ATMP sector. For Gruell, its usefulness depends on how actively it is maintained. “The regulatory sandbox is a good idea, but it has to be a living tool,” he said. Continuous exchange with regulators is essential, particularly when adopting new methods such as model‑based control or real‑time release.

His experience with the FDA highlights the difference that scientific engagement can make. During a pre‑licensing inspection, FDA inspectors spent extensive time reviewing Autolus’ potency assay, asking detailed questions about methods and data. “They were open to learn more about our scientific data,” Gruell recalls. In contrast, interactions within Europe were described as more conservative and less conducive to constructive exchange.

For PAT adoption, this difference matters. The validation of continuous analytics and control models depends not only on technical capability but also on a shared understanding of the scientific basis for decisions.

Why point-of-care manufacturing is still out of reach

Point‑of‑care (POC) manufacturing is frequently presented as a future direction for personalised therapies. By producing therapies directly at the hospital, it could eliminate cold‑chain logistics and reduce vein‑to‑vein time to days instead of weeks. Autolus closely examined the UK’s decentralised manufacturing guidance after its publication.

The concept is attractive, yet current processes are not ready. “Our process is too complex at the moment to really be at the bedside of the patient,” Gruell explained. Multi‑step cleanroom operations and extensive QC testing make decentralisation impractical today. However, if manufacturing times can be reduced to one day and cleanroom operations simplified, POC manufacturing becomes a realistic long‑term goal. In such a scenario, PAT would be central to ensuring consistency, traceability and regulatory confidence across distributed sites.

The barriers are organisational, not technological

PAT adoption has not kept pace with its recognised potential in ATMP manufacturing. Gruell described several interconnected reasons. A major factor is organisational readiness. Autolus’ current manufacturing process was developed over more than a decade and is considered mature and deeply embedded. “Our manufacturing process has been developed over twelve years now, and it’s too complex and difficult to radically  change it at this point,” he said.

Practical hurdles also play a role. Limited availability of GMP‑grade materials, the need for extensive incoming testing and resource constraints all compete with strategic improvement initiatives. As Gruell noted, “We still have several operational opportunities to improve upon  so that we may not have the required capacity in the company to really implement PAT yet.”

In many ATMP companies, process designs become fixed during early‑phase development. Introducing PAT later requires significant technical, regulatory and organisational effort, which often slows down adoption.

AI: useful, but not in charge

Artificial intelligence is another topic attracting attention, but Autolus is taking a measured approach. “I am open-minded but also sceptical to a certain degree,” Gruell said. The company has started using AI tools within its quality management system to support deviation handling and impact assessments. However, AI has not been applied to manufacturing control or analytical decision-making.

This selective use reflects a clear boundary: automation should support structured tasks where consistency can be improved without shifting responsibility for critical decisions. High‑impact manufacturing actions must remain under human oversight.

Conclusion

What Gruell’s experience at Autolus makes clear is that the obstacles to PAT adoption are rarely technical. Timing matters, organisational capacity matters, and so does the confidence of regulators. These are slower-moving constraints than any sensor or software.

As personalised medicine scales, the ability to understand a process in real time will be as important as the process itself. Gruell puts it directly: „Our patients are extremely ill and we are constantly in a race against the clock.” That urgency is ultimately what makes PAT a clinical question as much as a manufacturing one.