How AI Could Clear Logistics Roadblocks Limiting Patient Access To CGT
By Arnaud Deladeriere, Ph.D., Cell&Gene Consulting Inc.
The April Cell&Gene Foundry roundtable deconstructed a ubiquitous misconception and oversimplification of the patient access debate.
The prevailing logic around delivering ATMPs to more patients can be summed up as a cost issue. When a promising new cell therapy hits the market, big price stickers suck up all the oxygen. The tendency to focus on cost alone ignores other critical and interconnected factors that prevent drugs from reaching the patients who need them.
About the Cell&Gene Foundry
These ideas are shared in collaboration with the Cell&Gene Foundry, an industry group assembled to discuss important topics in cell and gene therapy development, led by Arnaud Deladeriere. This conversation includes insights from: Adrienne Mendoza of BioBridge Global, Rohin Iyer of Marken, Matthew Hewitt of Charles River Laboratories, and Claudia Zylberberg of Akron Bio and Kosten Digital.
To learn more about the Foundry, visit www.cellgeneconsulting.com.
As our group unpacked the issue, a much different narrative emerged. Improving access is, in fact, a complex logistical challenge, and one that AI seems especially poised to solve.
At its core, access can be understood as the ability to match an eligible patient with an available therapy in a timely and reliable manner. That matching process is currently limited by three interdependent dimensions:
- where therapies can be delivered,
- whether patients are identified and referred appropriately, and
- whether the system can support the operational burden of treatment.
The first dimension is geographic and infrastructural. Access is constrained by the number and distribution of treatment centers capable of administering these therapies. Today, these centers remain concentrated in major academic institutions, creating significant gaps for patients outside urban hubs. As efforts expand into community settings, new operational challenges emerge rather than disappear.
The second dimension is knowledge. Awareness of advanced therapies remains uneven across the clinical community. Many providers are still unfamiliar with referral pathways, eligibility criteria, and the operational requirements associated with CGT. This leads to missed opportunities to identify patients who could benefit from treatment and delays in initiating the process once they are identified.
– Adrienne Mendoza
The third dimension is economic, but not in the way it is typically framed. While affordability and reimbursement structures remain relevant, cost of goods is not the primary barrier to access. Commercial adoption trends indicate sustained demand. The more immediate limitation is the system’s ability to deliver therapies consistently and at scale.
This leads to a more fundamental constraint: supply. The industry is currently serving only a fraction of eligible patients outside the United States. Even within the U.S., penetration remains limited relative to the addressable population. The issue is not demand but the ability to operationalize delivery across a fragmented system.
Within that system, logistics plays a central role. The CGT supply chain depends on precise coordination across multiple handoffs, each with strict requirements for timing, temperature control, and traceability. These therapies behave less like conventional products and more like time-sensitive, patient-specific assets. The margin for delay or deviation is minimal.
– Rohin Iyer
At present, the logistics infrastructure is not fully adapted to these constraints. It relies on frameworks that were not designed for this level of time sensitivity or personalization.
Variability across collection sites, differences in scheduling practices, and lack of standardization create inefficiencies that accumulate across the chain. As therapies expand beyond major centers into more distributed geographies, these inefficiencies become more visible.
From a CMC perspective, analytics introduces an additional constraint. While manufacturing timelines have shortened, release testing remains a limiting step. Sterility and potency assays continue to require extended timelines, offsetting gains made upstream.
Taken together, the access problem is not a single bottleneck but a system operating below its potential due to fragmentation, variability, and limited coordination.
The Transplant Model As A Structural Constraint
The structure of today’s delivery model explains much of this fragmentation. Cell therapies have largely been built on the foundation of the bone marrow transplant paradigm. While this model enabled early adoption, it is increasingly misaligned with the requirements of scale.
One of the clearest constraints is the reliance on specialized treatment centers with stringent accreditation requirements. Programs such as FACT accreditation introduce significant cost and operational burden, particularly for community hospitals. Administrative processes such as single case agreements add further complexity, requiring individualized approval for each patient.
The result is a system concentrated in a limited number of institutions. Expansion into community settings is occurring, but more slowly than demand would suggest.
– Adrienne Mendoza
A central question is whether all steps in the treatment pathway need to remain tied to this model. The discussion highlighted several areas where decoupling is both possible and necessary.
Apheresis is one such step. Functionally, it is a manufacturing input rather than a clinical intervention.
Treating it as a hospital-bound procedure constrains scheduling, capacity, and patient access. Alternative approaches, including mobile apheresis units, demonstrate that this step can be decentralized and brought closer to patients. This reduces travel burden and enables participation from regions that would otherwise remain underserved.
From a logistics standpoint, this shift introduces new requirements. Increasing the number of collection points expands the geographic footprint and the number of operational nodes. This creates a need for more standardized processes and tighter coordination across sites. Without that standardization, variability at the point of collection can propagate through the entire supply chain.
Infusion is undergoing a similar transition. As clinical experience grows, there is a movement toward outpatient and community-based administration. However, adoption remains limited. Financial considerations play a role, particularly the need for hospitals to absorb up-front costs while awaiting reimbursement. This creates a constraint for smaller institutions and slows the expansion of treatment capacity..
Here again, logistics becomes a determining factor. Expanding into community settings requires reliable last-mile delivery in regions that are not part of established clinical or manufacturing networks. Ensuring consistent performance requires coordination across transportation providers, clinical sites, and manufacturing facilities under tight timelines.
– Matthew Hewitt
More broadly, the transplant model embeds assumptions that do not hold at scale. It centralizes activities that could be distributed, ties operational steps to clinical environments where they may not be required, and introduces layers of administrative complexity that limit throughput.
Digital Orchestration As The Enabler
If fragmentation is the underlying constraint, then integration becomes the central requirement for scale.
The CGT value chain is composed of multiple specialized components. Patient identification, cell collection, logistics, manufacturing, testing, and clinical administration each operate within separate systems. Data is generated at every step, but it is rarely connected in a way that supports end-to-end visibility.
This lack of integration limits both efficiency and learning. Decisions are made locally rather than system-wide, and data that could inform improvement remains isolated.
Digital orchestration, connecting existing systems, addresses this coordination gap. A platform-based approach could integrate electronic health records, manufacturing systems, logistics tracking, and analytical data into a unified framework. This would create a continuous digital thread across the vein-to-vein journey, enabling real-time visibility into patient status, product location, and process performance.
– Rohin Iyer
The implications extend beyond operational efficiency. Linking post-infusion outcomes to upstream variables would create a feedback loop for continuous improvement. Manufacturing parameters, logistics conditions, and patient characteristics could be analyzed together, enabling a more data-driven approach to optimization.
This type of integration also has strategic implications. Real-world data generated through connected systems could support regulatory discussions, inform reimbursement models, and strengthen the evidence base for these therapies. In that sense, digital infrastructure becomes a foundation not only for operations but also for long-term adoption.
Near-Term Opportunities For AI And Digital Tools
While full-scale orchestration is a longer-term objective, several areas offer more immediate opportunities for impact.
AI-optimized distribution networks
– Claudia Zylberberg
In logistics, AI is already being applied to route optimization and risk management. By analyzing historical performance data and external variables, systems can identify routes that balance speed, reliability, and cost. This is particularly relevant as the network expands beyond major hubs into more distributed geographies.
AI-triggered process interventions
Real-time monitoring is a complementary capability. Sensors tracking temperature, location, and handling conditions generate continuous data streams. AI can detect deviations and trigger corrective actions before product integrity is compromised, enabling a more predictive approach to logistics operations.
In manufacturing, process data can support earlier intervention. Growth curves and metabolic indicators can be analyzed to identify deviations before they lead to batch failure, improving consistency and reducing reactive troubleshooting.
AI-enabled toolkits for rapid site deployment
Site activation is another area where digital tools can deliver immediate value. Establishing new treatment centers currently requires significant effort in documentation, training, and coordination. AI-enabled toolkits can streamline this process by generating site-specific materials and training resources, reducing the time required to bring new centers online and supporting geographic expansion.
Small, strategic steps with measured expectations deliver immediate effects
Data integration, even at a limited scale, can also provide near-term benefits. Connecting a subset of systems within a defined region or network can serve as a proof of concept. These “digital units” can be replicated over time, building connectivity without requiring full system transformation up front.
It is important to recognize the role of AI within this context. These tools are most effective when applied to structured problems with clear inputs and outputs. They do not replace human expertise, particularly in clinical or complex operational decisions. Their value lies in augmenting existing processes and enabling better use of available data.
Five Key Takeaways
1. Patient access is already constrained, and scaling will amplify the gap
The industry is currently serving only a fraction of eligible patients, despite strong demand and commercial traction. As therapies move earlier in treatment lines and into new indications, this gap will widen significantly. Coordinating this increase in volume across a fragmented system will exceed human operational capacity. AI will be required to manage this complexity and maintain throughput.
2. A distributed delivery model is inevitable but requires coordination
Decoupling apheresis and expanding into community settings is necessary to increase access. However, a more distributed network introduces variability across sites, schedules, and processes. Without a digital backbone to coordinate these nodes, the system becomes harder to manage, not easier. AI-enabled orchestration will be needed to make distributed models operational at scale.
3. Logistics complexity will increase non-linearly with scale
As the number of patients, sites, and routes grows, logistics becomes significantly more complex, particularly for time- and temperature-sensitive therapies. Current models rely heavily on semi-automated coordination and predefined pathways. This approach will not hold as the network expands. AI-driven optimization and real-time decision-making will be necessary to maintain reliability and prevent logistics from becoming the primary bottleneck.
4. The upstream funnel will remain constrained without better use of data
Many eligible patients are still not identified or referred in time due to fragmented clinical pathways and limited awareness. As demand grows, this inefficiency will persist unless patient identification and referral become more systematic. AI can support this by leveraging existing data to improve patient matching, pathway navigation, and decision support at the clinical level.
5. Integration is the prerequisite for both scale and learning
The CGT ecosystem remains highly fragmented, with data and decision-making siloed across the value chain. As volumes increase, this lack of integration will limit both operational efficiency and the ability to learn from outcomes. AI has the potential to connect these components and enable continuous improvement, but only if data flows across the entire vein-to-vein journey. Without that integration, scaling the industry will remain constrained.
About The Author:
Arnaud Deladeriere, Ph.D., is principal consultant at Cell&Gene Consulting Inc. Previously, he was head of MSAT and Manufacturing at Triumvira Immunologics, and before that, manufacturing manager at C3i. He received his Ph.D. in biochemistry from the University of Cambridge.