Cell Therapy QC Automation: Getting Integration Right
Authored By:
Félix A. Montero-Julian, PhD, VP Scientific Affairs and R&D, Accellix
Geoffrey Stephens, PhD, MBA, CEO & Founder, AiCella, Inc.
Michelle Duquette, PhD, Chief Scientist, Cell Therapy, Invetech
Matt Yeung, Senior Principal Engineer, Invetech
Kelly Lord, Marketing Manager, Cell Therapy, Invetech
As cell therapy manufacturing scales, quality control (QC) increasingly sits at the center of some of the industry’s hardest challenges. High testing costs, long turnaround times, and limited ability to act on results in time can slow manufacturing and delay release, even as upstream processes continue to develop. Outdated analytical technologies also contribute to these challenges.
Addressing these challenges is not simply a matter of adding automation to existing workflows. It requires rethinking where testing happens, how data flows, and how automation and AI can support faster, more informed decisions throughout manufacturing.
These points were central to our Advanced Therapies Week breakfast session, From Bottleneck to Breakthrough: QC’s Automated Evolution. We brought together perspectives from pharma, engineering, technology, and AI to unpack the bottlenecks keeping us from moving QC forward. The shared conclusion was: automation on its own does not resolve today’s QC challenges. Progress is achieved when QC is designed for rapid decision-making and supported by integrated systems, data, and workflows, while also adopting more advanced analytical methods designed for speed and integration.
The QC Bottleneck Has Changed
Historically, QC bottlenecks were tied to manual execution. Labor-intensive assays, operator variability, and long runtimes slowed progress and limited throughput. Although those challenges still exist, many organizations are now finding that the bigger constraint is not running assays but acting on the results. Decision speed has become a limiting factor, particularly when QC data delays release decisions and ultimately, delivery of therapies to patients.
This is especially true for autologous therapies, for which time is critical, and each lot represents a patient. Even when assays are automated, QC can still fall short if the process requires data to be manually transferred, interpreted, or reconciled across disconnected systems. When QC data cannot inform decisions quickly, the impact of automation remains incremental.
In practice, teams tend to move through this in stages: first automating existing workflows, then reassessing the underlying assay technologies, and finally integrating QC into connected systems that can support faster, more informed decisions.
Know What You Are Solving for Before You Automate
One of the strongest messages from the session was the importance of defining success before introducing automation. For many organizations, the first step begins with layering automation onto existing assays, such as using liquid handling and robotic sample preparation to reduce manual work and improve consistency. This can be a practical starting point, but the session also highlighted the need to plan early for the digital pipeline that sits behind the instruments.
Decisions need to be anchored in clear KPIs from the start, including turnaround time, throughput, COGS, and assay failure rate. The session also made clear that there is no one “right” level of automation. Some teams start with off-the-shelf tools on top of legacy methods because it is the fastest way to reduce manual burden. When the assay itself becomes the constraint, teams begin to consider redesigned platforms that embed automation, standardization, and simplified interpretation more directly into the technology.
In many cases, the most practical path forward is to improve what is already validated by integrating and connecting workflows, rather than replacing entire systems at once. However, this approach doesn’t alleviate the challenges related to turnaround time or the effort required to reconcile all data for product release.
From an engineering perspective, several decisions need to be made early. The goal is to improve consistency, reduce analysis burden and build automation that supports scale from the start. Teams need to decide how many assays should be tackled in parallel and how far end‑to‑end automation should extend. Should automation stop at the batch record, or should it include sample movement throughout the facility? When success criteria are defined upfront, teams can assess these options objectively and focus investment on approaches that will scale effectively.
Efficient Is Not Enough: Rethinking Where QC Happens
Workflow automation is often where organizations begin, and it can meaningfully reduce manual effort and operator variability. However, automating existing QC workflows does not fundamentally change the role QC plays in the manufacturing process.
Much of today’s QC remains an end‑of‑process check. When teams apply automation to current assay technologies, they are often mechanizing methods that were never designed for speed or proximity to manufacturing. This can improve consistency and efficiency, but it does not materially reduce turnaround time. Samples still need to be transported to centralized labs, assays still run for the same duration, and data still needs to be batched, scheduled, reviewed, and interpreted before decisions can be made.
In this model, automation helps teams do the same work more efficiently, but it does not address the structural constraints that delay action. Logistics, centralized infrastructure, large equipment footprints, and dependence on expert interpretation remain in place. As a result, QC data often arrives too late to influence the process, serving primarily as documentation rather than as a real‑time decision input.
The longer‑term vision shifts QC closer to the manufacturing floor. Cartridge‑based systems, minimal manual intervention, simplified analysis workflows, and timely results that can inform the next steps in manufacturing point toward a more responsive QC model. Instead of bringing samples to the lab, the test moves closer to the process.
When evaluating legacy platforms versus new technologies, revalidation cost often dominates the conversation. However, the more complete comparison must also consider the cost and effort required to continue scaling manual and centralized processes. Once teams recognize that some level of revalidation is inevitable either way, it opens the door to more forward‑leaning automation decisions that can truly improve turnaround time and process responsiveness.
There is a growing importance of moving from retrospective QC evaluation toward proactive manufacturing intelligence.
As decentralized and near-process analytical technologies continue to emerge, AI-driven platforms may help connect in-process manufacturing data with downstream QC outcomes instantly. This creates the potential for earlier intervention, adaptive process control, and more efficient resource allocation.
By combining automation, digital integration, and machine learning, manufacturers may eventually shift from static release testing models toward dynamic quality frameworks capable of continuously assessing process health throughout manufacturing. This evolution could be particularly impactful for autologous cell therapies, where manufacturing timelines are compressed, and rapid operational decisions directly influence patient treatment schedules.
Data as the Foundation for Predictive QC
Data strategy must be addressed early. AI readiness depends on structured, traceable data, making it essential to define chain of identity, chain of custody, sample IDs, test IDs, and information architecture before systems are scaled. QC cannot be siloed from quality and digital teams.
Once data is structured and flowing reliably, QC can move closer to the manufacturing process itself. Process Analytical Technology (PAT) enables real‑time insight into manufacturing, allowing teams to react to in‑process data and adjust downstream parameters as conditions change. When implemented with the right automation and data integration, PAT moves quality closer to the process enabling more timely and actionable insights.
Over time, as high‑quality data accumulates, this foundation enables predictive QC, using models to anticipate outcomes and proactively tailor process parameters based on input material characteristics. The shift from reactive control to predictive decision‑making depends on designing data, systems, and workflows with this end state in mind.
From an AI perspective, predictive QC requires more than simply applying AI models to manufacturing datasets. The real challenge lies in creating harmonized, analysis-ready data environments capable of integrating process parameters, analytical outputs, operator actions, equipment metadata, and imaging data into a unified framework. In practice, much of the industry’s effort is still spent organizing fragmented datasets, reconciling inconsistent naming conventions, and standardizing historical records before advanced analytics can generate meaningful insights.
Successful implementation of AI-enabled QC depends on building scalable digital infrastructure that supports explainable and actionable decision-making. This includes integrating machine learning with process analytical technologies (PAT), enabling models to identify patterns associated with potency, process drift, assay variability, and manufacturing risk earlier in the workflow. Rather than replacing existing QC systems outright, AI can augment operational decision-making by helping teams prioritize investigations, identify emerging trends, and support faster release readiness assessments.
Ultimately, models linking product attributes to patient response are emerging to move manufacturing analytics to clinical efficacy.
Designing QC for Scale
Teams that succeed at scale are making intentional choices now. Platform selection, data architecture, and decisions about where testing truly needs to happen will define future performance.
Scalability in cell therapy manufacturing will increasingly depend on the industry’s ability to operationalize data across the entire manufacturing ecosystem. Future-ready QC systems must be designed not only for assay execution, but also for interoperability between instruments, MES platforms, electronic batch records, LIMS environments, automation platforms, and AI-enabled analytics tools.
AI adoption in advanced therapies is not simply a software initiative. It requires organizational alignment across data governance, workflow standardization, and cross-functional collaboration amongst manufacturing, quality, engineering, and informatics teams. Organizations that establish these digital foundations early will be better positioned to implement predictive analytics, reduce operational variability, and accelerate manufacturing scale-up over time.
QC is no longer just about meeting release criteria. It plays an increasingly important role in how manufacturing teams make decisions, manage risk, and scale reliably, while supporting more accurate prediction of clinical efficacy.
As the industry continues evolving toward smarter manufacturing paradigms, forward-thinking companies are helping demonstrate how AI, automation, and integrated data strategies can work together to transform QC from a retrospective checkpoint into a proactive manufacturing capability. The long-term opportunity lies not simply in automating individual assays, but in enabling connected decision-making environments that improve responsiveness, reduce risk, and support scalable commercialization of advanced therapies.
Looking ahead, success will depend less on individual tools and more on how well systems, data, and workflows are integrated to support decision-making across manufacturing.
Panel (L-R) Michelle Duquette, PhD Invetech, Félix Alejandro Montero Julian, PhD Accellix, Edith Chang, PhD Bristol Myers Squibb, Geoffrey Stephens, PhD AiCella & Matt Yeung Invetech
