Brian Drapeau is founder, GxPFrame and a cell and gene therapy manufacturing and quality systems consultant.

Brian Drapeau

Industry experts from Sanofi, GxP Frame, and Open Biopharma outline how AI is transforming drug discovery, manufacturing, and workforce strategy through biopharma partnerships in the latest episode of Drug Digest.

Drug Digest: How Advanced Analytics Partnerships Enhance the Biopharma Value Chain

, global head of Computational and AI Strategy for R&D at Sanofi, 

, founder and principal consultant of GxP Frame, and 

, senior director of Education, Outreach, and Digital Experience at Open Biopharma Research and Training Institute, discuss artificial intelligence (AI) integration across drug discovery, 

, supply chain operations, and workforce development in this special combined episode of 

Truppo describes AI's primary value as connecting signals across previously siloed data types, including molecular structure, biological assay results, patient demographics, and clinical site performance, to improve scientific decision quality rather than generate autonomous answers. At Sanofi, for example, disease-specific target identification engines integrate genetics, proteomics, transcriptomics, and single-cell data to help researchers determine target druggability, tissue expression, and optimal therapeutic modality, he explains. 

All three experts emphasize that human oversight must be structurally embedded in AI workflows, not applied as an afterthought. 

How should biopharma companies govern AI tools to maintain data integrity and regulatory compliance?

, Drapeau emphasizes that the most valuable AI applications are not headline-grabbing but operational, such as deviation management, batch record review, predictive maintenance, and supply and demand forecasting. In cell and gene therapy (CGT), specifically, he notes that supply chain is patient-specific, making chain of identity and chain of custody the true measures of AI's value. He also notes that no single manufacturer owns the full CGT supply chain, making cross-company partnerships essential.

Drapeau adds that effective partnerships between biopharma and AI tech and/or advanced analytics companies require a partner capable of integrating across multiple domains while maintaining strict data governance. "Completion is not competency," he stresses, cautioning against operators who can execute AI-assisted tasks but lack the independent judgment to troubleshoot or critically evaluate model outputs.

All three experts emphasize that human oversight must be structurally embedded in AI workflows, not applied as an afterthought. Jaenisch states that explainability is non-negotiable, cautioning that AI outputs used in regulated environments must be traceable and defensible to inspectors.

“There are [many] pieces that an explainable AI system is able to retrieve and provide with every output, and if you do not have that, then you are probably not going to have a good time when an audit comes,” he explains.

Truppo adds that data governance, source traceability, and clearly stated model uncertainty are prerequisites for trustworthy AI integration across biopharma R&D and manufacturing operations when dealing with a strategic partner.

Matt Truppo, Global Head of Computational & AI Strategy, R&D, Sanofi

As the global head of Computational and AI Strategy at Sanofi, Dr Truppo oversees platform capabilities across all therapeutic areas and modalities to advance the company’s drug discovery ambitions. He focuses on embedding AI and predictive technologies throughout R&D to accelerate timelines and enable the development of first-in-class therapies. Prior to Sanofi, he led Protein Therapeutics API Development and Biologics Discovery at Johnson & Johnson and spent more than a decade at Merck working across both small- and large-molecule discovery and development. He holds a PhD in Chemistry from the University of Manchester and undergraduate degrees in Chemical Engineering and Engineering Biology from Princeton University. Over a 20+ year career, Truppo has contributed to bringing 13 medicines to market and advanced many more into the clinic, a quest driven by his commitment to improve global health through innovative science.

Richard Jaenisch, Senior Director of Education, Outreach, and Digital Experience, Open Biopharma Research and Training Institute

Jaenisch has spent the past 15 years in patient advocacy since he began and fought off end stage liver disease and more than 15 years in STEM education, marketing, and consulting. Among his duties at Open Biopharma, he is leading several data projects regarding generative AI integration and exploring generative AI in biomanufacturing training. At Prompting Integration and Consulting LLC, he develops training courses and works alongside his brother, a seasoned AI/ML engineer to develop bespoke generative AI integrations for business in life sciences and healthcare and deliver educational programming on generative AI.

Brian Drapeau, Founder and Principal Consultant, GxP Frame

A private consultant to some of the largest biotech companies in the industry, Drapeau brings more than 5 years of experience in workforce development and training across the biotechnology and pharmaceutical sectors. His expertise includes new hire training, instructional program development, contract development and manufacturing organization (CDMO) training requirements, and high-volume ramp-ups. Drapeau also authors 

, a newsletter covering FDA enforcement trends, CDMO training failures, and pharmaceutical manufacturing operations.

® Group editors, who interview industry experts to discuss the emerging opportunities, obstacles, and advances in the pharmaceutical and biopharmaceutical industry for the research, development, formulation, analysis, upstream and downstream processing, manufacturing, supply chain, and packaging of drug products.

Videos

In this episode of Drug Digest, experts from Sanofi, GxP Frame, and Open Biopharma explain how AI is transforming drug discovery, manufacturing, and workforce strategy and how strategic partnerships play a role.

Disclaimer: The views expressed in the article are those of the author and do not necessarily reflect those of MJH Life Sciences or 

A line that stuck with me came out of an FDA inspection in Livonia, Mich. Investigators told a small drug manufacturer it had shipped product without ever running process validation. The company’s explanation, recorded in Warning Letter 320-26-58, was that it hadn’t known validation was required, because the AI agent it had been leaning on to write its procedures never mentioned it.

I’ve spent enough years inside GMP training and quality systems to register a few things at once when I read that. It’s darkly funny and also terrifying. It’s the kind of statement that ends up in a deposition. And it sent me down a longer road than I expected, because the question underneath it is one the industry keeps answering with marketing instead of evidence: can you actually put generative AI inside a regulated quality system and have it survive contact with an inspector, or is “compliant AI” a phrase vendors say because buyers want to hear it?

I wanted a real answer, not the brochure version of one. Here’s where a few weeks of poking at it left me.

Maybe you’ve seen this one go around as FDA’s first standalone citation for AI, the letter that supposedly rewrote pharma’s buying criteria overnight. Slow down there. That framing falls apart the second you actually read the thing.

The firm, Purolea Cosmetics Lab, was a homeopathic and cosmetics operation. I’m using past tense because their operations ceased after the grocery list of citations they received. Among the flagged products were “Dermveda Extra Strength Shingles Relief” and a genital herpes “relief” product carrying disease claims that make them unapproved new drugs.

 Investigators found insects, filth, leaves, and clutter in the facility, plus a docking bay door that opened straight onto the manufacturing area.

 No microbial testing on finished product. No component identity testing. A quality unit that wasn’t reviewing batch records before release.

 This was a firm operating with almost no quality system to speak of.

The AI finding isn’t a separate citation either. It sits inside the quality-unit (QU) failure under 21 CFR 211.22, specifically 211.22(c): if you use AI to help create documents, you have to review those documents to confirm they’re accurate and compliant, and not doing so is the violation.

 The process-validation gap under 211.100 reads like a punchline if it weren’t shipping real product: the owner later told investigators she hadn’t known validation was required because the AI never brought it up.

So a sloppy supplement shop isn’t a stand-in for a real manufacturer, and anyone selling it that way is overreaching. What makes the letter worth your time is the sentence FDA wrote in response, the one that will long be remembered rather than the incident itself. Any output from an AI agent, the agency said, “must be reviewed and cleared by an authorized human representative of your firm’s QU.”

 FDA wasn’t banning AI. It said plainly that AI can be used as an aid. The deficiency was that nobody with authority stood between the machine’s output and the regulated decision.

That’s a low bar. Have a qualified human review it. When I started asking whether that bar is actually enough, the answer got very uncomfortable.

The Trouble with “A Human Reviews It”

The reflex fix, once you accept AI into the workflow, is to put a competent reviewer at the end of it. Problem solved, supposedly. The model drafts, the expert checks, the expert signs.

The research on how people behave around automated output doesn’t support that confidence. A systematic review of automation bias in clinical decision support found that good automated advice lifts performance, but incomplete or incorrect advice drags it down, because reviewers stop scrutinizing and start deferring to the system.

 study of radiologists working with AI assistance found the effect isn’t even reliably positive. Whether the AI helped or hurt depended on the case and the individual, and in some scenarios inaccurate predictions pulled skilled readers toward worse calls than they’d have made alone.

Translate that into a quality organization. A reviewer facing 40 AI-drafted deviation rationales in a week, each one reading clean and authoritative, is not the neutral error-catcher the control assumes. Volume erodes the scrutiny. So does any process that treats the signature as a formality rather than a decision. “A human reviewed it” can be technically true and functionally empty, and an inspector who understands that won’t be impressed that a name appears in the approval field.

This is the point where I stopped believing the problem was about better models or better prompts. If the human checkpoint is weaker than everyone assumes, then making the AI more fluent only makes its mistakes more convincing.

The Governance Problem Hiding Underneath AI

The sharpest version of this I heard came from two people I’ll get to in a moment, the founders of Counterpoint AI, and it reframed the whole thing for me. The industry keeps framing regulated AI around content accuracy and hallucination: the fear that a generative system will produce something false that slips into a regulated record. That risk is real, but they argue it may not be the defining governance problem as AI adoption expands in pharma. I went in unsure they were right about that and came out mostly persuaded.

Their harder problem, in their words, is keeping a regulated operation ready while change comes faster than it used to. Quality has never just produced accurate documents. It keeps track of what applies to whom, what got authorized, who is qualified for what, and whether all of that still holds as procedures and protocols move underneath it. That work has always been a grind. It used to be a manageable one.

What makes that manageability run out is volume and velocity. The share of trial protocols carrying at least one substantial amendment has climbed from 57% to 76% since 2015, and the mean number of amendments per protocol is up ~60%, to 3.3 from 2.1.

 The parallel rise of generative AI makes it tempting to believe these capabilities will let an organization keep up by generating and propagating the content around that change. The catch is that AI scales content generation far faster than it scales governance, which still runs on slower, human-scale coordination. Without mechanisms that control downstream operational activity, ungoverned document generation outruns the humans who are supposed to keep it in check.

Their framing gave me the language for something I had felt but could not name, and building on it, I think the failure mode has a more concrete shape than “readiness” suggests. Call it operational drift. The exposure I keep coming back to isn’t the wrong document, it’s the right one landing in the wrong place. A change that propagates incompletely. A qualification state that quietly diverges from its governing source. A procedure that goes operationally live before anyone has governed its downstream consequence. Run that forward and you get a second, worse problem: the gradual loss of authoritative visibility itself, where quality leaders can no longer say with confidence which people, sites, procedures, or qualifications are operating in a governed state. That is a different and harder class of failure than a hallucinated sentence, and it’s the one most quality systems were never built to catch, because they assume change moves at human speed and AI has changed the speed.

Change control is the clean example. A machine can determine which sites and roles a change touches and hold it until training and validation are done, and that is real work taken off a human plate. But assessing the quality impact and deciding to approve or reject is judgment that stays human, and it gets sharper when the routine reconciliation underneath it is handled. Counterpoint maps the whole quality operation this way, and once you look closely most functions do split along those lines. Some of the work is pure administration a system can absorb, some shrinks to a smaller human footprint, and some is judgment that does not go anywhere. The trouble starts when a tool quietly takes on the judgment-shaped parts under cover of doing the administrative ones.

The founders believe this challenge lands hardest on contract research organizations (CROs), and the logic held up for me. A CRO juggling several sponsors runs each program as a discrete governance context, with separate procedures, separate qualification matrices, and overlapping staff, then absorbs the coordination cost as overhead that scales faster than its revenue. A single qualification gap at one site doesn’t stay at that site, because sponsors are obligated to look at what it means across every program they run with that CRO. That’s where the manual model strains first and where the operational and commercial pressure become impossible to ignore, because what’s at stake is a book of business.

One thing I owe you up front. I’m not the first to read Purolea this way. In the weeks after the letter, several quality and medical-device writers reached the same conclusion: a company with no functioning quality unit had reached for AI to plug the hole, and everything in the letter about AI was downstream of that. Greenlight Guru, Clarkston, and others made versions of the point, and they were right, so the diagnosis isn’t mine to claim.

The question I couldn’t find anyone answering was the one I cared most about. If the real problem is governance rather than accuracy, does anyone have a fix that would survive an inspector? That’s what sent me looking for someone building for the hard problem instead of the easy one.

One company, at least, believes the answer is yes.

In Part II, I'll examine one company's attempt to answer that question: can governance-by-design solve a problem that better content generation alone cannot?

U.S. Food and Drug Administration. Warning Letter 320-26-58, Purolea Cosmetics Lab (FEI 3011669383). Center for Drug Evaluation and Research, Office of Compliance. April 2, 2026. 

Goddard K, Roudsari A, Wyatt JC. Automation bias: a systematic review of frequency, effect mediators, and mitigators. 

Yu F, Moehring A, Banerjee O, et al. Heterogeneity and predictors of the effects of AI assistance on radiologists. 

 2024;30:837-849. doi:10.1038/s41591-024-02850-w

Getz KA, Smith Z, Botto E, et al. New benchmarks on protocol amendment practices, trends and their impact on clinical trial performance. 

Getz KA, Stergiopoulos S, Short M, et al. The impact of protocol amendments on clinical trial performance and cost. 

 2016;50(4):436-441. doi:10.1177/2168479016632271

Articles

An FDA warning letter often cited as a landmark “AI compliance” moment instead reflects profound GMP breakdowns, with AI-generated procedures left unreviewed by an empowered quality unit. The enduring regulatory signal is simple: AI may assist, but authorized human QU review and clearance remain mandatory. That safeguard is challenged by automation bias, where fluent machine output degrades scrutiny under volume. The larger risk shifts from hallucinated text to governance failures—operational drift, incomplete change propagation, and loss of authoritative visibility as AI accelerates document throughput beyond human-scale oversight.

A warning letter, a hard look at why “a human reviews it” doesn’t settle anything, and one company’s argument that accuracy was never the real problem.

Is Compliant AI in Pharma Even Possible? The Search for an Honest Answer

Women, scientist and discussion with tablet by glass, reading and review with results at pharma company. People, mentor and feedback with notes, application and medical research for mpox vaccine | Image Credit: ©Malambo20/peopleimages.com -stock.adobe.com

The cell and gene therapy (CGT) contract development and manufacturing organization (CDMO) market is projected to grow 10x from $6.41 billion to roughly $75 billion by 2034 (see 

 However, FDA Form 483 observation trends

 tell a different story than one would gather from investment headlines. Recurring observations include incomplete training documentation, absence of written training procedures, failure to qualify personnel for assigned tasks, and the absence of a structured program for ongoing competency (detailed in 

). This discrepancy stems from a workforce being asked to expand on deliverables while remaining undertrained, under retained, and inadequately transitioned into the very roles they are hired to fill and execute. What follows is a cost-based argument for why training infrastructure is the variable that will separate the organizations still standing in 2034 from the ones that did not make it.

Figure 1. CGT CDMO Market Projection, 2024-2034. The global CGT CDMO market is projected to grow from $6.41 billion in 2024 to $75.32 billion by 2034, representing a compound annual growth rate of 27.94%.

 Source: Towards Healthcare, citing Precedence Research (2025).

Table 1. Recurring Training-Related FDA 483 Observation Categories in Biopharmaceutical Manufacturing (FY 2024)

The financial influx surrounding the investment in CGT manufacturing right now is genuinely striking. Johnson and Johnson committed $1 billion to a new facility in Montgomery County, PA, in February 2026, projecting more than 500 hires and committing to a production-ready date of 2031.

 Towards Healthcare, citing Precedence Research data, puts the global CGT CDMO market at a compound annual growth rate of roughly 28%, projecting a near 12-fold expansion over the present decade.

However, what those figures fail to show is this harsh reality: between 2020 and 2024, 74% of Complete Response Letters issued to CGT developers by the FDA cited quality or manufacturing deficiencies as the primary cause.

 In the same period, approximately 40% of investigational new drug applications were either rejected or outright stopped before reaching clinical evaluation as a result of inadequate chemistry, manufacturing, and controls.

 These are not the numbers of an industry that has its execution problem solved, or even close to being so. In fact, they are the numbers of an industry that has accelerated past its own operational readiness. The question is, why and how? A quick analysis of the pattern in FDA 483 observations from the same period shows exactly where these breakdowns are taking place.

A review of FDA Form 483 observations issued to biopharmaceutical manufacturers during fiscal year 2024 identified several recurring categories, all related to training deficiency: incomplete training documentation, absence of written training procedures, failure to qualify personnel for their assigned tasks, and no structured training program to maintain ongoing competency (see 

 These observations were not limited to startups or first-time facilities being inspected. Established CGT manufacturers received them as well.

The reason this pattern repeats itself is worth examining carefully. Training gaps in good manufacturing practice (GMP) environments are rarely fixed by uploading additional content to a learning management system (LMS). The LMS takes on the role of becoming an electronic repository of written procedures, but a completion report is not the same as training infrastructure. It is just a document repository with associated quizzes that serve to demonstrate a required procedure has been read. However, knowledge of all material read during a session is rarely retained, and the practical application of said knowledge is an entirely separate skill that requires repetition, failure, and troubleshooting. So, when the inspectors come with their questions, they rarely ask whether the course was completed, as it is assumed it is. Rather, they ask if the person doing the work can demonstrate competency in each area, and those are two very different questions.

) 211.25, personnel engaged in manufacture, processing, packing, or holding of a drug product must possess the education, training, and experience required for their assigned functions.

 Training must be conducted by qualified individuals, must cover the specific tasks the employee performs, and must include current GMP requirements. In a CGT facility, those tasks routinely include lentiviral vector handling, autologous cell processing, chain-of-identity management, and closed-system bioreactor operation. Applying the 211.25 standard to those environments is not a matter of checking a training completion box. An inspector standing in front of an operator who cannot explain why chain-of-identity controls exist, or what the consequence of a deviation in that step is, has found a training deficiency regardless of what the LMS record claims.

What is the Burn Rate in High Turnover GMP Environments?

The workforce problem strengthening this compliance picture is well-documented. A 2025 survey by the International Society for Cell and Gene Therapy found manufacturing and process development as the largest gap in the entire CGT workforce, with shortages greatest in technical laboratory, manufacturing, and quality control (QC) roles.

 A 2024 BioPlan Associates survey, as reported by Locwin, found that vacancy rates in critical CDMO positions, specifically manufacturing science and technology (MSAT), quality assurance, and automation engineering, exceeded 15% at many sites. The result was seen in major delays during technology transfer timelines and a significant increase in deviation frequency.

The industry response has been unsurprising. To mitigate these serious concerns, increased money has been thrown at the problem: larger salaries, elevated offers, title inflation, aggressive counteroffers, and extreme poaching. Not to say this approach is not understandable, as it does address immediate hiring pressure, but it fails to address what happens 12 months after you have placed new talent in a broken system.

The Society for Human Resource Management reports that replacing an employee costs between 50% and 200% of their annual salary when both direct and indirect expenses are fully accounted for.

 Direct costs are familiar: recruiting fees, time-to-fill, onboarding administration, and formal training. The indirect costs are where GMP environments diverge sharply from general industry. An operator less than six months into their role represents a documentation exposure in every batch record they sign. A quality associate who has not yet internalized investigation methodology represents a deviation-handling vulnerability that can surface in the next FDA inspection. A trainer who was never formally qualified is generating training records that an investigator can challenge on their face. These are not theoretical risks. They are the conditions described repeatedly, in the existing 483 observations.

The math is simple. From personal experience, when the chimeric antigen receptor (CAR)-T therapy boom first began in Northern NJ around 2013-2014, starting with Celgene and Novartis, the average cost to train a new operator was roughly equal to 1/3 to 1/4 of their starting salary, culminating in a roughly $75,000-$85,000 investment per operator once materials and trainer time were accounted for. Using the SHRM midpoint estimate, total replacement cost, including training investment, recruitment, and productivity loss, falls in the range of $75,000 for a manufacturing-level role.

 At a facility of 100 employees running 40% annual turnover, that generates 40 replacement events per year at a cost of approximately $3 million. At 20% turnover, the same facility spends roughly $1.5 million. That $1.5 million annual gap is not the result of paying people more (see 

). It is the result of whether the organization gave employees a reason to stay.

Table 2. Annual Turnover Cost Comparison: 100-Person GMP Manufacturing Facility

What Will Separate the Operations That Survive?

There is a recognizable profile to the organizations most at risk in the current expansion cycle. They have invested heavily in equipment and facilities. They carry significant overhead from paying competitive salaries. Many have implemented an LMS and consider the training question largely resolved. Unfortunately, reading and understanding only go so far. What is missing, consistently, is a sufficient training architecture built to create repeatable outcomes with minimal variance regardless of the individual working through it. A structured system connects role-specific competency requirements to a progression pathway from hire to fully qualified and does so in a way that enables the operator to understand not just the how, but the why. Implement documented checkpoints along that pathway and a visible career development model that answers the question every new employee is asking: What will my future here look like?

The career development component is the one most frequently absent, yet it is the main driver of unnecessary turnover.

When a CGT manufacturing operator leaves six months after completing onboarding, the departure is rarely about salary alone. In most cases, no competitor offered them a dramatically better compensation package. They received a marginally higher offer accompanied by a title that suggested forward movement. The incumbent employer lost that person not because they were outbid, but because they never communicated a development trajectory. The competitor did not offer a better job. They offered a story about the future.

Building that story is not a significant capital investment. A tiered qualification program, a structured mentorship arrangement, and a documented promotion framework require organizational design time and a committed training budget. They are operational decisions, not capital expenditures. Organizations that have made them consistently report lower voluntary turnover, more predictable inspection performance, and a workforce that can respond to FDA questions with documented evidence rather than a completion certificate.

The distinction the FDA draws during an inspection is the same one that determines whether an organization can scale. A training program designed to generate records is a compliance theater. A training architecture designed to produce competent operators is the actual product.

Regional Investment and the Workforce Readiness Gap

The concentration of CGT manufacturing investment along the mid-Atlantic corridor right now is not subtle. J&J's $1 billion Montgomery County commitment requires 500 qualified employees, and they are not waiting until 2031 to start hiring. Validation work starts well before product runs, which means they are going to be qualifying operators in a facility with no production history, on a regulatory timeline that cannot slip, while every other CGT site in the Philadelphia corridor is fishing in the same talent pool simultaneously.

So where does that workforce come from? The honest answer is that nobody really knows yet, and that is the problem. The ISCT survey from 2025 flagged that the US Department of Education has acknowledged that biomanufacturing workforce development is lagging current market demand.

 Community college partnerships are valuable. Regional workforce initiatives are a start. But enrollment to GMP-qualified operator is a multi-year process, not a quarterly one. Any organization banking on hiring a ready-made workforce from the existing market is going to run headfirst into a 15% vacancy rate and realize the market was never going to supply what they needed.

The workforce must be built internally. That means onboarding programs designed to compress qualification time without creating the documentation gaps that show up in 483 observations. It means training systems where inspection-ready records come out the other end of normal operations, not out of a remediation sprint two weeks before an audit. And it means giving people a reason to still be there in year two.

The $75 billion projection is real. The pipeline behind it is real. Nobody is disputing that this market is going where analysts say it is going.

What is not modeled in any of those projections is what it costs to deliver on them with an undertrained, high-turnover workforce. A $60,000 salary and $60,000 in training costs represent $120,000 walking out the door every time someone leaves before they have returned that investment to the organization. Run that math at 40% annual turnover across a hundred-person facility, and there is a $3 million annual burn rate that does not show up on any capital expenditure line, does not depreciate on a schedule, and compounds every year the underlying problem is not fixed.

The CDMOs that are still standing at the end of this decade are not necessarily going to be the best funded or the most scientifically sophisticated. They are going to be the ones that figured out earlier than their competitors that they cannot inspect their way to a competent workforce and they cannot salary their way to a retained one. Infrastructure must be built that produces both, and it must be done before the FDA performs an inspection.

The floor is not the LMS completion report. It never was. It is whether the person on the manufacturing floor can tell an investigator what they are doing, why it matters, and what happens when it goes wrong. Everything else is compliance theater.

Cell and gene therapy CDMO market size, share, trends, growth rate to 2035. Towards Healthcare. Published November 1, 2025. Accessed March 2026. 

https://www.towardshealthcare.com/insights/cell-and-gene-therapy-cdmo-market-sizing

Top FDA 483 observations from 2024 and strategies to mitigate them in 2025. PRP Compliance Group. Published July 12, 2025. Accessed March 2026. 

https://www.prp-compliance.com/post/understanding-the-top-7-fda-483-observations-from-2024-and-strategies-to-mitigate-them-in-2025

J&J plans $1B investment in PA cell therapy manufacturing facility. BioSpace. Published February 19, 2026. Accessed March 2026. 

https://www.biospace.com/drug-delivery/j-j-plans-1b-investment-in-pa-cell-therapy-manufacturing-facility

Why gene and cell therapies are stalling at the FDA. Drug Discovery News. Published August 3, 2025. Accessed March 2026. 

https://www.drugdiscoverynews.com/why-gene-and-cell-therapies-are-stalling-at-the-fda-16527

https://www.ecfr.gov/current/title-21/chapter-I/subchapter-C/part-211/subpart-B/section-211.25

International Society for Cell and Gene Therapy Laboratory Practices Committee. Shaping the future of cell and gene therapy workforce development: training of cell therapy processing personnel. 

. Published online April 10, 2025. Accessed March 2026. 

https://www.sciencedirect.com/science/article/abs/pii/S1465324925006802

Locwin B. Science, scale, and skills: what will define the CGT CDMO champions. BioProcess International / Workforce Genetics. Published April 30, 2025. Accessed March 2026. 

https://workforcegenetics.com/science-scale-cell-gene-therapy-cdmo-market/

Society for Human Resource Management. Turnover cost calculation spreadsheet and benchmarking data. Updated 2025. Accessed March 2026. 

https://www.shrm.org/topics-tools/tools/forms/turnover-cost-calculation-spreadsheet

CGT CDMOs are on a steep growth trajectory, yet FDA quality signals show persistent execution risk driven by workforce readiness. Form 483 trends repeatedly cite training-system failures—missing procedures, incomplete documentation, unqualified personnel, and absent competency maintenance—reflecting gaps between LMS “completion” and demonstrable GMP capability. High vacancy and turnover amplify deviations, tech-transfer delays, and inspection exposure, creating a material, recurring burn rate. Durable advantage will come from training architecture that ties role-specific competencies to qualification pathways, mentorship, and career progression to retain talent and sustain inspection-ready performance.

The cell and gene therapy manufacturing workforce is having to expand on deliverables while remaining under trained.