In the pharma industry, the Artificial Intelligence (AI) market is projected to grow from more than $4 billion this year to a whopping $25.7 billion by 2030. However, McKinsey & Company claims that medicine makers have yet to see substantially shorter development timelines or improvements in preclinical or clinical success rates.¹

But several industry gurus believe this new year holds a lot of promise for AI. John Chinnici, CEO of Ledger Run, a next-generation software suite developer, says: “Next year will be a turning point where AI’s promise in life sciences finally meets pragmatic execution.”

“In 2026, expect the transformation of drug development from a predominantly human-driven, sequential process into a continuously learning, agentic AI-supported pipeline,” agrees Andrew Mackinnon, Global Executive General Manager, Medable, whose AI-powered clinical trials platform has been deployed in nearly 400 trials in 70 countries. “Instead of researchers and clinicians manually generating hypotheses, designing studies, reviewing data, and coordinating decisions across long cycle times, agentic AI systems will autonomously propose targets, run virtual experiments, optimize protocols, monitor safety signals, and surface decision-ready recommendations.”

Despite the potential of AI, all industry insiders agree that humans will not be removed from the equation. Instead, they will move into higher-value oversight roles, validating AI-generated options and steering strategy while AI handles the labor-intensive, multi-step analytical and coordination work. “The result will be a funnel that’s faster, more adaptive, less linear, and increasingly self-optimizing, marking the first true structural redesign of the R&D model in decades,” says Mr. Mackinnon.

In this inaugural, exclusive Artificial Intelligence Drug Development & Delivery special feature, leading drug makers, device manufacturers, and contract organizations dispel the myths around AI and Machine Learning (ML) in the pharma industry and share how they are using AI to streamline clinical trials, automate lab tests, optimize resource allocation, enhance development timelines, and improve patient-friendly dosage forms.

Adare Pharma Solutions: AI Is Only As Good As The Support It Receives

“One of the biggest myths in the in­dustry is that AI will instantly transform drug development on its own,” says Tom Sellig, CEO of Adare Pharma Solutions. “In reality, AI is a tool, and just like all tools, it’s only as effective as the systems, people, and practices that support it.”

Successful CDMOs already know that bringing a therapy to market depends on structured processes, rigorous oversight, and clear communication. Those same fundamentals are what AI integration re­quires to be truly effective.

“AI can certainly accelerate insights, but it must operate within the same disci­plined framework CDMOs rely on to de­liver therapies,” he says. “AI will never be a replacement for a team of experienced experts, but AI can enhance those teams’ capabilities and help them make better in­formed decisions, identify risks earlier, and move programs forward with greater con­fidence.”

Mr. Sellig says that AI’s greatest po­tential value will be in reducing friction across the development and manufactur­ing lifecycle. As programs become more complex and timelines shrink, sponsors will depend ever more on integrated, end-to-end partners, like Adare, that can keep projects moving.

“AI’s strength lies in how it can help connect data across development, manu­facturing, and packaging operations, help­ing teams spot risks, uncover opportunities for cost and time savings, and make more informed decisions,” he says. “With AI-dri­ven predictive intelligence, teams can tran­sition from reactive problem-solving to proactively steering operations.”

AI will also provide value by enhanc­ing how patient-friendly dosage forms are designed and optimized. As technologies like 3D screen printing enable more per­sonalized and multi-layered therapies, AI can help predict performance, refine tai­lored formulations, and streamline scale-up, he says.

Bespak: Facilitating Performance Simulations Without Costly Experiments

There is a myth that regulators, such as the FDA and EMA, won’t accept AI models. However, the FDA and the Centre for Research on Complex Generics (CRCG) recently partnered in a workshop on “Modeling and Artificial Intelligence (AI) in Generic Drug Development and Product Lifecycle Management: Regulatory Insights and Future Trends.” During the workshop, the FDA and EMA explained their guidance with regard to AI and their desire to work with generic product devel­opers from the start and bring in novel AI methodologies, explains Alan Harris, Chief Technology Officer of Bespak. The workshop identified many concrete exam­ples of AI being deployed in drug sub­stance and drug product development, including in complex drug-device combi­nation products for simulations of lung deposition and pharmacokinetic perform­ance. The use of AI is also being consid­ered by regulatory authorities like FDA to support submissions reviews and data analysis.

There is also a belief that AI can offer solutions to everything. “The myth of AI providing a silver bullet has been de­bunked,” Mr. Harris says. “While many areas may benefit from AI, scientists re­main indispensable to manage how AI is used in development and to share relevant approaches in regulatory discussions.”

For its part, Bespak is seeing more novel active pharmaceutical ingredients (APIs) coming through the clinic with its drug development partners, but the role of AI in their development is still in its infancy. “Historically, APIs in respiratory medicine – Bespak’s focus area – are small and many variants on the market are modest molecular “redesigns” performed in tradi­tional ways by chemists,” he explains. “As biologics become more popular in respi­ratory medicine, the role of modelling and AI will increase in early development to facilitate simulations of biological, physical, and chemical performance without costly experiments.”

Bespak is already exploring AI for modeling and simulation tools. In collab­oration with the Centre of Excellence in In Silico Regulatory Science and Innovation (CEiRSI) network, Bespak is investigating innovative uses of AI and modelling to support the UK’s ambition to be a power­house in drug development innovation.

Bespak is also developing a Digital Engineering Platform to deploy modeling, simulation, and AI across all modalities of its R&D and scale-up services, from formu­lation and process development to device design and inhaler in vivo performance. Mr. Harris explains: “While some of these areas are in early stages, we have several years’ experience in deploying AI and sim­ulation tools for respiratory device and pMDI valve finite element analysis and computational fluid dynamics to de-risk, speed up, and reduce the cost of device development. We can even simulate the filling process of pMDI cans, which has never been done before. By offering these services to customers from a menu of op­tions, we can support their development and regulatory plans and work alongside their own R&D teams.”

The use of AI simulation tools for de­vice design and optimization has gener­ated designs and simulated different use conditions, materials, and other factors without needing to produce pilot tools and parts. This facilitates cost savings and speeds up the development process, which can ultimately benefit patients, he says.

Going forward, Mr. Harris says AI will likely support every aspect of product de­velopment from start to finish, including target APIs, formulation, device develop­ment, clinical study design, and process development through to ex-device per­formance, and PBPK models. Beyond this, AI will have more operational but equally beneficial uses in process control, defect/rejects analysis, real-time monitor­ing of product quality, and data analysis.

Enzene: Distill Complex Data Into Actionable Insights

Today, Enzene’s Fully Connected Continuous Manufacturing™ technology is evolving rapidly to include real-time feed­back that can improve bioprocessing effi­ciency and reduce waste through intelligent, ‘self-adaptive’ systems.

“Today’s large language models demonstrate how AI can distill vast, com­plex information into actionable insights,” says Russell Miller, Vice President, Global Sales and Marketing, Enzene. “Applied to bioprocessing, enormous datasets from sensors, batch records, and quality sys­tems will be processed without the fear of data interpretation overwhelming human operators.”

He adds that AI can not only summa­rize and suggest responses, but can also predict process outcomes, optimize pa­rameters dynamically, and support real-time decision-making.

“In the near future, these capabilities will evolve into autonomous bioprocessing environments powered by digital twins—virtual models that continuously learn and adapt from live data,” he says. “This means faster development cycles, reduced variability, and proactive compliance, ulti­mately accelerating innovation and im­proving patient access to life-saving therapies.”

Mr. Miller sees a transition from AI being viewed as an assistant to one in which AI is the backbone of truly intelligent biomanufacturing ecosystems.

Gerresheimer: Refining Patient Interaction

Artificial intelligence is transforming connected health devices by enabling in­telligent, personalized, and intuitive sup­port. AI has the potential to transform therapy support by shifting from reactive to anticipatory models and by providing access to reliable medical information. Ac­cording to Giacomo Bruno, PhD, Digital Health Platform Lead at Gerresheimer, AI systems can identify risks, such as adher­ence decline or technique errors, before they escalate by analyzing behavioral pat­terns, routines, environmental data, and population trends. These systems deliver timely interventions for maximum impact.

Gerresheimer is integrating AI directly into devices, packaging, digital platforms, and patient services to create seamless, adaptive experiences that fit naturally into daily routines, explains Dr. Bruno. “Instead of requiring additional apps, Ger­resheimer’s smart, connected approach activates AI through simple interactions, such as scanning a QR code or tapping an NFC tag on devices like the Gx Inbeneo® autoinjector,” he says. “This provides im­mediate access to AI-driven support via fa­miliar messaging apps, such as WhatsApp, iMessage, Facebook Messen­ger, and Google Messages, without the need for new downloads or learning curves. Patients receive tailored reminders, electronic Instructions for Use (eIFU) with visuals, and direct answers to questions, resulting in effortless engagement and higher adherence.”

Dr. Bruno says that this integrated AI approach benefits all stakeholders. Pa­tients receive proactive, personalized guid­ance, while healthcare providers access real-time insights. Caregivers stay in­formed, and payers see improved out­comes through better compliance.

“This approach optimizes therapy ef­fectiveness and usability,” he says. “Ger­resheimer aims to redefine patient interaction by combining AI-driven guid­ance with connected devices and low-fric­tion communication channels.”

Patients also benefit from instant, val­idated guidance in plain language through reliable channels, bypassing un­reliable sources. AI extends the reach of healthcare professionals with continuous, context-aware reinforcement, reducing their workload while maintaining clinical oversight.

“Gerresheimer views AI as essential for safer and more adherent patient expe­riences, where reliable support is consis­tently accessible through common tools such as chat apps,” says Dr. Bruno. “This ultimately improves outcomes and effi­ciency throughout the pharmaceutical ecosystem.”

ICON Clinical Research: Predictive Analytics Forecasts Post- Marketing Requirements

One of the biggest myths surrounding AI in drug development is that it will com­pletely replace human decision-making and expertise. AI is not a standalone solu­tion but a strategic enabler. While AI-dri­ven tools offer remarkable capabilities, such as screening millions of compounds rapidly, optimizing clinical trials, and im­proving operational efficiency, they work best when combined with human expert­ise. “Rather than replacing scientists, clini­cians, and decision-makers, AI allows them to focus on higher-value, more com­plex tasks by automating repetitive and data-intensive processes,” says Robert El­lison, Vice President of Data & Applied An­alytics at ICON Clinical Research.

In reality, he says AI is transforming drug delivery by enabling precision and personalization. Machine Learning (ML) models can stratify patient populations using biomarkers and real-world evi­dence, helping to tailor therapies to indi­vidual biological profiles. This ensures that the therapies being developed are both more targeted and effective, he says.

“AI-designed drugs progressing quickly to Phase II trials represents a shift towards therapies that are not only faster to develop, but are also more likely to suc­ceed due to their alignment with specific patient needs,” says Mr. Ellison. “Addition­ally, deep learning in medical imaging im­proves endpoint evaluations, contributing to better data quality for regulatory sub­missions and enhancing targeted drug de­livery mechanisms.”

ICON has several proprietary tools that embed AI across the clinical develop­ment continuum to deliver measurable ef­ficiencies and cost savings. The CRO uses platforms to leverage decades of trial per­formance data, real-world evidence, and Machine Learning algorithms, accelerating site identification, reducing start-up timelines and improving enrolment rates. For example, predictive analytics tool forecasts post-marketing requirements early in the development cycle, enabling sponsors to mitigate regulatory risks and avoid costly delays.

Operationally, ICON has a pair of re­view tools that streamline documentation-heavy processes, from site contracts to trial master file management, reducing man­ual effort and improving compliance. “AI technology helps us to identify key opinion leaders in rare disease areas by analyzing millions of publications at speed, cutting timelines from months to days,” he ex­plains. In medical imaging, deep learning algorithms automate segmentation and annotation tasks, reducing processing time from hours to minutes while maintaining accuracy through expert validation.

ICON also deploys AI tools for re­source forecasting and operational met­rics, ensuring optimal resource allocation and faster decision-making. “Together, these solutions have shortened trial time­lines, improved data quality, and reduced operational risk – helping sponsors bring therapies to patients sooner while maxi­mizing return on investment,” he says.

Mr. Ellison says that the future of AI in the pharmaceutical industry lies in gener­ative AI and large language models (LLMs), which are already demonstrating potential in summarizing complex datasets, drafting documents, and analyz­ing vast data lakes. “These tools offer not only efficiency gains but also a significant advantage in knowledge management, decision-making, and operational work­flows,” he says. “When responsibly inte­grated, AI will help the pharmaceutical industry accelerate innovation, reduce risk, and deliver life-changing therapies to pa­tients at a faster pace.”

Lifecore: AI Tools Must Fit Within Regulatory Frameworks

Like many in the industry, Lifecore In­jectables CDMO is investigating the man­ner in which AI tools may fit into and benefit a highly regulated manufacturing environment. While the buzz around AI has been steadily building for quite some time, regulatory bodies like FDA have only just recently drafted industry guidance on AI. With this guidance now in hand, Lifecore’s teams are working to ensure that any AI tools it considers will fit within reg­ulatory and risk-based frameworks prior to testing and adoption. At the same time, Lifecore is working to ensure that its use of these tools does not open the company up to risk regarding the exposure of intellec­tual property or privacy concerns.

At present, Lifecore has taken initial steps to investigate and utilize AI tools aimed at reducing administrative work­load and improving clarity of communica­tions, explains Matt Augustson, Senior Vice President of Information Technology at Lifecore. For example, AI tools are used to help draft quality investigation summaries to ensure that all details are clearly and ef­fectively captured and collated. “The re­duction in hours spent on these tasks has been dramatic, allowing investigators to spend more time working through active investigations, reviewing drafts for accu­racy, and finalizing reports for release,” he says.

For Lifecore, another area of interest for future AI considerations is the automa­tion of laboratory testing. At times, this may involve collaboration with robotic sys­tems, leading to a further reduction in the need for human involvement. “For exam­ple, we’ve seen systems that are designed to help automatically identify and flag agar plate growth,” he says. “Again, these types of systems seem to offer promise, but need to be investigated within our specific environment to ensure that they can reli­ably perform operations over time.”

Lonza: AI-Enabled Toolkit De-risks Development at Every Stage

As drug developers and manufactur­ers face intense pressure to bring more and more complex compounds to market faster, they increasingly rely upon ad­vanced technologies powered by AI and ML to expedite innovative solutions. In that context, AI can be misconstrued as an all-knowing oracle that provides a single cor­rect answer. The reality is that AI provides a ranked list of possibilities with their as­sociated confidence scores. Essentially, AI is a data-driven partner that helps lab chemists explore a wider range of possibilities and focus their own expertise where it can be most impactful, says Aaron John­son, Manager of Cheminformatics and Data Science, Lonza Advanced Synthesis.

To harness the power of technological innovation, Lonza has developed an AI-enabled toolkit to de-risk development at every stage, ensuring phase-appropriate optimization and improving the chances of clinical success. The toolkit is a connected ecosystem of predictive tools designed to facilitate better decision-making earlier in the development cycle. At its core is Lonza’s AI-enabled Route Scouting Serv­ice, which incorporates computer-aided synthesis planning technologies to com­pute the shortest and most viable retrosyn­thetic paths to intermediates or active pharmaceutical ingredients (APIs), lever­aging proprietary informatics, supply chain information, and the expertise and decision-making of experienced chemists.

Another toolkit offering, Solid Form Services (SFS), incorporates Lonza’s predic­tive co-crystal model, which uses ML algo­rithms to screen thousands of molecules in minutes, enabling rapid identification of the optimal solid form for the drug product.

“Together, these tools act as powerful ‘copilots’ for our expert chemists, aug­menting their skill and intuition while al­lowing them to innovate more robust and scalable workflows,” says Mr. Johnson. “The synergy directly boosts efficiency and sustainability in production, facilitating the delivery of high-quality products with lim­ited environmental impact (due to reduced chemical and material waste), a reliable supply chain, and the fewest number of synthetic steps.”

Internal validation studies demon­strate that Lonza’s AI-enabled models achieve predictive accuracy exceeding 86%, allowing the company to accelerate discovery with a high degree of scientific reliability. The toolkit also enables confir­mation of the effectiveness and perform­ance of top-rated candidates while minimizing development time, reducing costs, strengthening intellectual property, and ultimately accelerating the clinical readiness of customers’ compounds. “By integrating these validated AI tools into our daily work, we unlock new levels of inno­vation and operational excellence, creat­ing a distinct competitive edge in drug development and manufacturing,” he says.

Looking to the future, Mr. Johnson en­visions potentially valuable applications of AI in areas such as formulation develop­ment, where the technology can be used to predict excipient compatibility, and in process chemistry, in which AI can opti­mize reaction conditions for yield and pu­rity. Additionally, as demand grows for highly potent APIs, he expects increased use of AI-driven simulations to predict re­action outcomes, optimize process param­eters, and minimize waste, leading to more efficient and sustainable production. On the manufacturing side, AI can aid in optimizing facility design and containment strategies, which can minimize worker ex­posure to hazardous materials and the en­vironmental impact of such materials. Johnson also expects increasing use of AI-powered tools to streamline supply chain management, improving the predictability and security of raw material sourcing.

“All these possibilities inform Lonza’s long-term vision, which is to create a con­nected ecosystem of AI tools that bridge the entire development cycle, allowing in­sights from one stage to accelerate the next,” he summarizes.

MedPharm: Support Predictions of Drug/Excipient Interactions

“Artificial Intelligence is a hot topic across all industries now, and drug prod­uct development is no exception,” says Charles Evans, PhD, Senior Vice President of Pharmaceutical Development at Med­Pharm. “However, depending on who you speak with, there is often a gap between what people believe it can do, what it can do now, and what it can do in the future.”

He says a common misconception or myth is of AI replacing scientists, practical experiments, and independently predicting clinical outcomes. In reality, the use of AI in guiding direction and planning, accelerat­ing insight and context, and supporting de­cisions when paired with strong scientific judgment is where it brings most value. AI systems can learn patterns, simulate findings from data, generate new options, and predict outcomes, but they still require human interpretation and validation.

For example, in-silico modelling AI can be extremely useful for narrowing down large molecular libraries, exploring theoretical mechanisms, or predicting how a drug might behave in different excipi­ents, but it cannot replace empirical sci­ence, Dr. Evans stresses.

“A study by Lenn et al. (2018) on RNA aptamer delivery through intact human skin found that, although modelling would likely have suggested the molecule was too large to penetrate the skin, laboratory test­ing demonstrated that it could cross the skin barrier,” he explains. “Likewise, while AI may be able to predict certain formula­tion behaviors, it is unlikely to capture every chemical or physical interaction within a complex formulation that is com­posed of not only the drug(s) but numer­ous excipients with differing functionalities, which may have never been looked at be­fore. As such, robust preformulation, for­mulation development, and in-vitro/in-vivo testing is still required to check and vali­date any AI predictions, though of course, such data may prove useful in any further work. It is also important to note that any AI model is limited by the data it has been trained on, and for new chemical or bio­logical entities there may be little-to-no prior data available.”

A major benefit of AI is cost and time saving, which is multi-faceted, both in sup­portive functions and in core R&D activi­ties. AI provides decision support by highlighting potential risks, predicting likely outcomes, and helping teams focus resources on experiments most likely to succeed. Such an approach, he says, re­duces wasted effort and material costs. “In formulation development, AI has the po­tential to support prediction of drug/excip­ient interactions, stability trends, and likely drug-release profiles, which although cur­rently need to be checked, would still trun­cate the process itself and help inform the next development steps,” he continues.

Even without fully implemented pre­dictive models, AI is already useful for an­alyzing experimental data, highlighting patterns, and identifying areas where ad­ditional testing may be most impactful. For Dr. Evans, the future for AI lies in support­ing faster and more informed decision making while helping to improve the prob­ability of success across the development pipeline. “We have already seen this in Al­phaFold, an AI system developed by Google DeepMind, which predicts the 3D structure of proteins with high accuracy and significantly accelerates biological re­search.”

In addition, large language models (LLMs) have been shown to speed up the approval of clinical trials; and clinical study reports, which typically take 9-10 weeks to summarize clinical results, can be reduced to around a week, he says. “As more systems are developed and the cur­rent systems mature, their ability to further reduce, de-risk, and importantly support scientific decisions should offer significant value to the pharmaceutical industry.”

Phillips Medisize: Delivering Predictive Insights & Streamlining Workflows

In the discovery and development stages at Phillips Medisize, AI helps elimi­nate low-value administrative tasks by automating data management and pro­cessing. This shift allows scientists and en­gineers to devote more time to analyzing data, uncovering insights, and focusing on innovations that directly benefit patients and customers, explains Dave Thoreson, Vice President, Global Operations, Phillips Medisize.

“Strategically deploying AI — espe­cially Machine Learning (ML) — in our manufacturing processes has further driven significant cost and time savings,” he says. “AI delivers predictive insights and consol­idates complex datasets that were previ­ously inaccessible, empowering our teams to make faster, more informed decisions.”

ML accelerates data analysis by effi­ciently identifying hidden patterns and po­tential risks that traditional methods might miss. By augmenting human expertise, AI enables Phillips Medisize’s workforce to concentrate on higher-value activities and reduce manual, repetitive tasks. Addition­ally, AI helps mitigate unconscious human biases through objective, data-driven in­sights. This improves the accuracy of deci­sion-making, leading to more effective project prioritization and better allocation of resources, he says. “Collectively, these advantages streamline workflows, cut down trial-and-error cycles, and reduce costs — enabling us to help our customers bring therapies to market faster and with greater efficiency.”

Looking ahead, Mr. Thoreson believes that AI holds transformative potential across many facets of the industry, with the greatest value in enhancing data-driven decision-making and operational effi­ciency. “Ultimately, the benefit to the pa­tient or healthcare provider will be delivering solutions faster,” he says. “With data insights from ML enabling better, faster decision making, innovation can happen faster.”

Predictive AI, while promising, re­mains challenging to fully validate and can sometimes underperform due to the complexity and variability of real-world conditions. However, ML applications fo­cused on sorting, categorizing, and ana­lyzing large datasets tend to deliver more consistent and actionable results.

In manufacturing specifically, ML can accelerate operational efficiency and op­timize cycle times by automating routine tasks, improving predictive maintenance, and enabling smarter resource allocation. “These gains hold promise to not only im­prove internal efficiency but also have downstream impacts — speeding up time to market for new products and ultimately benefiting patients by providing faster ac­cess to innovative medicines,” he says. “In summary, AI’s most valuable contribution will be its ability to augment human ex­pertise with faster, more precise data analysis and decision support, driving both operational excellence and improved cus­tomer outcomes.”

Portal Instruments, Inc.: Broaden What Science Can Create & Enable What Patients Can Receive

The myth is that AI will instantly dis­cover new drugs. The reality is that it does not replace biology, clinical validation or manufacturability constraints, etc. What AI truly does is collapse cycle times by elimi­nating thousands of dead ends early. “The winners will be those who integrate AI tightly with experimental data, device en­gineering, and real-world constraints, not those who expect AI to “solve” drug deliv­ery development in isolation,” says Patrick Anquetil, CEO, Portal Instruments, Inc.

As AI platforms expand the universe of new biologics (e.g. higher concentra­tions, higher viscosities), they will create therapies that will be increasingly difficult to deliver through traditional devices. For Portal, this accelerating trend aligns di­rectly with what PRIME NEXUS (a closed-loop, computer-controlled drug delivery system) is built to solve:

• Higher-viscosity formulations: AI-de­signed molecules frequently optimize potency at the expense of injectability; NEXUS is engineered specifically to de­liver viscous or large-volume drugs that spring-based pens cannot handle.
• More precise PK/PD requirements: As AI optimizes molecular behavior, dosing precision, and delivery rate control be­come far more important; Portal’s closed-loop, software-defined delivery architecture meets that need.
• Rapid iteration: When drug designs change quickly, pharma partners can­not wait 9 to 18 months for new hard­ware. A software-configurable platform like NEXUS allows device parameters to be tuned in minutes rather than through hardware redesigns.

“Today we use AI to design our adap­tive delivery algorithms: AI helps us opti­mize motor control and flow rate in real time,” he says. “In short, AI is broadening what science can create and Portal is en­abling what patients can actually receive.”

In the future, Mr. Anquetil believes that AI will assist in:

• Predictive maintenance: Forecast device or cassette issues before they occur.
• Adherence intelligence: As patients struggle with chronic injections, NEXUS will help analyze use patterns, predict drop-off risk, and prompt targeted in­terventions.
• Clinical insight generation: Aggregated, de-identified data supports payers, providers, and pharma with evidence on real-world use and persistence.

“The real opportunity is seamless drug-device integration powered by AI,” he says. “Instead of treating the molecule and the delivery system as separate prob­lems, AI could match a biologic’s physical properties with the optimal delivery pa­rameters in-silico at the moment the mol­ecule is designed. This would allow pharma to anticipate viscosity, volume, and delivery constraints upfront, thus en­abling faster development, smoother launches, and ultimately more personal­ized dosing for patients.”

Quotient Sciences: 50% Reduction In Formulation Development Time

Artificial Intelligence has firmly estab­lished itself as a transformative force in drug discovery. Over the past decade, AI-driven platforms have demonstrated clear benefits in target identification, lead opti­mization, and molecule design. By lever­aging vast biological and chemical datasets, AI can predict drug-target inter­actions, assess toxicity, and even generate novel molecular structures through gener­ative algorithms. These capabilities have significantly reduced timelines and costs compared to traditional trial-and-error ap­proaches, enabling pharmaceutical com­panies to accelerate the journey from concept to candidate, explains John Mc­Dermott, Vice President, Scientific Consult­ing, Quotient Sciences.

While discovery remains the most ma­ture application of AI, new use cases are emerging across other stages of drug devel­opment, particularly in formulation design and optimization. Historically, formulation development has relied on human experi­ence and labor-intensive experimentation to address drug delivery challenges relative to a target product profile.

AI and Machine Learning are now being deployed to streamline these processes. “Techniques such as Bayesian optimization and active learning allow it­erative refinement of models using mini­mal initial data, guiding scientists toward the most informative experiments,” he says. “This ‘decision-support’ tool also helps improve the quality and robustness of the formulation development process.”

Quotient Sciences’ own experiences in evaluating such approaches show re­duced formulation development time by up to 50%, while simultaneously providing greater understanding of the relationships between composition and laboratory end­points. If the model is developed further to enable predictions of clinical product per­formance, it will give clients a streamlined pathway to their next milestone with greater confidence of clinical success, says Mr. McDermott.

“However, despite these advances, AI will never replace human expertise. These models are only as good as the data they are trained on and the prompts they re­ceive,” he says. “In pharmaceutical devel­opment — where regulatory compliance, patient safety, and nuanced scientific judg­ment are paramount — human-in-the-loop approaches remain essential. Scientists guide AI systems by framing the right questions, validating outputs, and in­terpreting predictions. This ensures that AI augments rather than overrides human decision-making.”

Mr. McDermott points to one industry perspective (ValenceAI), which notes that explainable AI and interactive workflows are critical to building trust and ensuring ethical, accurate outcomes in high-stakes environments like drug development.

In December, Quotient Sciences forged a partnership with Intrepid Labs to advance the use of AI in early drug devel­opment. Under the terms of the agree­ment, Quotient Sciences will have access to Intrepid’s Machine Learning model, An­dromeda, an AI platform for development and optimizing clinical performance of drug products. Andromeda supports rapid exploration of formulation options, reduc­ing experimental burden, minimizing drug substance demands, and enhancing data-driven decision making. The partnership builds on the companies’ existing collab­oration where Intrepid’s AI model was in­corporated into Quotient’s Translational Pharmaceutics® platform to accelerate the identification of optimal formulation com­positions and reduce time to transition new drug products into clinical development.

“In summary, AI has moved beyond its early promise in discovery to become a versatile enabler across the drug develop­ment continuum,” he says. “Its integration into formulation development represents a paradigm shift — one that accelerates timelines, reduces costs, and improves product quality. Through careful experi­mental design and human-in-loop over­sight, AI will help us deliver better medicines, faster.”

Sapio Sciences: Multiplies Scientific Throughput & De-risks Decisions

The greatest myth is that AI will au­tonomously replace the scientist. That view understates the challenges that drug de­velopment must overcome to yield positive results.

“The reality is partnership: scientists lead the work and remain in control, and AI amplifies and accelerates its impact,” says Mike Hampton, Chief Commercial Officer, Sapio Sciences. Furthermore, wet lab verification remains essential because models can predict, but only experimenta­tion can confirm. “Used well, AI multiplies scientific throughput and de-risks decisions by collaborating with accountable human experts, helping the industry to deliver treatments faster,” he says.

The biggest near-term opportunity in the field of AI is finding ways to democra­tize its access for the scientific community. “Most labs now run multiple specialized tools for modeling, screening, and analy­sis, but without integration they create more friction than progress,” he explains. “Scientists spend more time figuring out where trustworthy AI models are, who can access them, and how to utilize them rather than interpreting the results.”

The breakthroughs, Mr. Hampton ex­plains, will come from orchestrating deep, science-centric AI tools, creating a single, governed environment where models connect, data flows cleanly, and every action is traceable. “AI only delivers value when it is built into an integrated workflow with clear ownership and high-quality data, and these workflows are accessible to sci­entists,” he says. “The teams that focus on unifying the AI environment in the lab will move faster and gain a lasting edge.”

Simtra BioPharma Solutions: Laying The Foundation For Meaningful AI Integration

Artificial intelligence is transforming drug development, but misconceptions persist. One of the greatest myths is that AI can already deliver a fully automated, end-to-end drug program. In reality, AI’s strengths are still within early phases of the development process, namely in the dis­covery phase where it is used in the screening and identification of the right molecular entity to provide the desired therapeutic benefit within an acceptable toxicity level. “Once programs advance into clinical development and large-scale manufacturing, AI’s role becomes limited as data complexity rises and process records remain less digitized,” says Luis Mustafa Perez, Vice President of Opera­tional Execution at Simtra BioPharma So­lutions. “While AI is revolutionizing discovery and adding value in isolated use cases, the notion of end-to-end au­tonomous drug development is far from today’s reality.”

Generative AI, however, is pushing boundaries. Acting as a full-stack molec­ular design engine, it accelerates hit iden­tification and early-stage development, enabling AI-designed drugs to reach Phase II trials faster. This progress sharp­ens the focus on precision and targeted delivery. “Yet, as early-stage acceleration shifts bottlenecks downstream, the industry must modernize and digitize delivery mod­els,” he says. “Creating a ‘digital space’ for rapid experimentation and hypothesis-driven candidate selection is critical. CDMOs are beginning to implement AI-driven systems for process development, leveraging proprietary data to enhance learning and adaptability.”

Simtra BioPharma Solutions is laying the foundation for meaningful AI integra­tion. While AI has supported process de­velopment through simulation models and clinical experiment design, the gains have primarily been in productivity and cost avoidance rather than direct reductions in R&D timelines or manufacturing expenses, Mr. Perez explains. “True transformation will come as data systems and operational workflows become fully digitalized, en­abling AI to influence decision-making at scale,” he says.

Looking ahead, AI’s greatest potential lies in unlocking insights into previously in­accessible processes, he says. By connect­ing disparate systems, datasets, and models, AI can surface trends and risks earlier, guide optimization, and empower smarter, faster decisions. For CDMOs, this means not only improving speed, cost, and reliability but also helping clients de­sign better products that achieve their in­tended therapeutic purpose. Mr. Perez concludes: “The future of AI in drug devel­opment is about amplifying human expert­ise through connected intelligence.”

Stevanato Group: Improve Defect Detection Accuracy Up To 99.9%

Stevanato Group has realized signifi­cant cost and time savings by integrating Artificial Intelligence into automatic visual inspection lines used in one of its client’s production processes. The AI platform uses deep learning models to improve defect detection accuracy up to 99.9% for cos­metic and particle inspection, while reduc­ing false rejects by a factor of ten. “This means fewer good products are discarded and less time spent on costly re-inspec­tion,” says Federico Scattolin, System Owner AI, Stevanato Group.

AI also eliminates the need for fre­quent machine reprogramming and opti­mization, which traditionally requires manual intervention and long setup times. By learning from real production data, the system is robust to variations, speeding up changeovers and reducing downtime, he explains.

“Using Stevanato Group cloud-based SG Vision AI platform — a secure solution featuring deep learning models — cus­tomers work in a GMP-compliant environ­ment,” Mr. Scattolin continues. “SG Vision AI helps deliver enhanced inspection per­formance by increasing detection rates and minimizing false rejection rates, while reducing costly re-inspection. Customers can upload, label, and manage data eas­ily through a user-friendly interface, while benefiting from continuous expert support for model development and qualification.”

Overall, he says these improvements translate into faster production cycles, lower operational costs, and more efficient resource use — helping Stevanato Group and its partners deliver reliable quality control while saving time and money.

Thermo Fisher Scientific: AI Model For Vial Particle Inspection Reduces Rejection Rates By 84%

There is a myth that Artificial Intelli­gence (AI) will eliminate the need for sci­entists, but the reality is that AI will only enhance – not replace – human expertise. Throughout drug discovery and develop­ment, scientists are already leveraging AI/Machine Learning (ML) models as part of a digital toolkit to help them process and analyze vast datasets, as well as pre­dict outcomes. This means that they can spend more time on smart, targeted ex­perimentation to get to the heart of com­plex challenges, says Sanjay Konagurthu, Senior Director, Science and Innovation, Pharma Services, Thermo Fisher Scientific. “As a CDMO and CRO, Thermo Fisher sees AI as a powerful enabler that helps teams make faster, more informed deci­sions that translate into better patient out­comes,” he says.

AI/ML technologies can help drug de­velopers address common formulation challenges that impact efficacy with greater speed and precision than previ­ously possible via traditional trial-and-error experimentation. Whether drug developers are tackling solubility, perme­ability or bioavailability, modeling First-in-Human (FIH) dosing, identifying the right packaging to ensure stability, gathering data for an Investigational New Drug (IND) application or preparing to scale up for production, data-driven insights will improve outcomes and accelerate progress, he says.

“In this way, AI/ML-enabled platforms help drug developers to better understand molecule behavior, save valuable active pharmaceutical ingredients, shorten time­lines, and mitigate risks,” says Mr. Kon­agurthu. “Furthermore, by predicting formulation behavior, scientists can lever­age AI-enabled insights to address chal­lenges in real time before they derail programs.”

From using predictive modeling to helping its customers design experiments more efficiently to optimizing manufactur­ing processes that minimize batch failures, Thermo Fisher’s deployment of AI has re­sulted in measurable cost savings at nu­merous stages of drug development. To date, Thermo Fisher has used AI/ML and predictive modelling to support early de­velopment, formulation development, and process development, as well as stability predictions for more than 400 com­pounds.

“In some of our manufacturing processes, we have leveraged the combi­nation of human expertise with an AI model for vial particle inspection to reduce rejection rates by 84%,” he says. “We also rolled out a manufacturing operating sys­tem that enables intelligent alarm analysis and performance modeling beyond stan­dard analytics. Most recently, we began work to integrate OpenAI’s advanced in­terface and Application Programming In­terface (API) technology across operations, particularly in our clinical research busi­ness to reduce clinical trial cycle times to help identify therapies less likely to suc­ceed, enabling customers to reallocate in­vestments toward more promising drug candidates. The impact is tangible – we’ve supported the development and commer­cialization of small molecule blockbusters, solved hundreds of solubility issues, and been involved in the development of nu­merous innovative therapies.”

As the pharmaceutical and biotech in­dustries continue to prioritize bringing safe and effective therapies to market, many are exploring how to make treatments even more accessible, especially delivering novel therapies (often large molecules) in formats such as oral solid dose drugs. Once deemed suitable for injectable deliv­ery only, large molecules face new possi­bilities as AI-enabled technologies enable drug developers to experiment with formu­lations to deliver them orally. GLP-1 med­ications are a great example, with several candidates showing promising results in clinical trials.

“AI/ML technologies bring an extraor­dinary amount of value to the biopharma industry, and that value will only continue to grow,” says Mr. Konagurthu. “These technologies can transform how data in­forms key decisions throughout the drug development journey, which, in turn, ac­celerates the process that gets life-chang­ing therapies to patients.”

Veridix: Reshaping Clinical Research

The greatest myth is that AI is a magic button that will completely replace people or fully automate trials end-to-end. The re­ality is that AI’s impact is maximized when embedded into structured processes, uni­fied data models, and multidisciplinary teams, says Fareed Melhem, President of Veridix, an Emmes Group company. Clin­ical trials involve thousands of context-spe­cific judgments. AI augments human expertise — orchestrating data, docu­ments, and analysis. This accelerates cycle time and removes operational drag so teams can focus on science and patient outcomes.

The acceleration of AI-designed mol­ecules changes the entire downstream op­erational model. Veridix’s approach to embedding AI has been to connect the en­tire clinical workflow rather than introduce isolated tools. “When assets reach Phase I/II with less historical data and more un­certainty, we need a trial engine that can learn and adapt in real time,” says Mr. Melhem. “With our Document Authoring Agent, we can generate protocol drafts, SAPs, and downstream study documents in days instead of weeks. More importantly, the protocol becomes machine-readable, enabling our downstream data, biostats, and monitoring agents to orchestrate workflows automatically.”

On the analytics side, Veridix’s biostats and analytics agents ingest the SAP directly, generate TFL shells, configure listings, and produce analysis-ready TFLs with real-time validation. “This doesn’t just make the work easier, but it allows us to run analysis as often as needed to con­stantly re-evaluate dose exploration, re­cruitment patterns, safety signals, and operational risks with a speed that matches the pace of AI-driven drug devel­opment,” he says. “The shift isn’t just speed — it’s tighter integration between documents, data, and analytics, enabling trials that truly adapt as evidence accumu­lates – and analytics feed instantly into Clinical Study Reports.”

Across the Veridix–Emmes ecosystem, the impact of AI has beyond cost savings; it has fundamentally reshaped cycle times. The Document Authoring Agent has re­duced protocol, SAP, and medical-writing timelines by 60% or more, routinely deliv­ering protocol drafts in three days instead of six weeks and automatically generating downstream documents that used to re­quire multiple teams and handoffs, he ex­plains.

The Data Management Agent also automates study builds directly from the protocol to drive cleaner, standardized data throughout a trial, which enables faster mid-study changes, significantly re­ducing queries and accelerating down­stream analysis. Mr. Melhem says: “As a result, we are eliminating weeks to months of latency across the study timeline. The result for sponsors is a dramatic reduction in cycle time and higher predictability; for investigators, fewer manual queries and administrative burdens; and for patients, cleaner data, and faster recognition of safety and efficacy signals.”

Mr. Melhem says AI’s greatest poten­tial going forward lies in creating a more connected and anticipatory clinical ecosys­tem. As agents continue to mature, they will enable trials where design, data cap­ture, and analysis are inherently aligned — reducing friction and allowing insights to surface far earlier than they do today.

“The next frontier is real-time adapt­ability: trials that can forecast operational risks, adjust based on emerging data, and integrate multimodal evidence without manual intervention,” he envisions. “Rather than automating isolated tasks, AI will increasingly function as the fabric that links decisions, systems, and stakeholders across the study lifecycle. That shift — to­ward intelligent orchestration and contin­uous learning — is where we see the most profound potential to reshape clinical re­search.”

Vetter: Navigate Disruption, Build Resilience & Enhance Competitiveness

“Artificial Intelligence is undoubtedly one of the most transformative technologies of our time,” says Titus Ottinger, Man­aging Director, Vetter. “However, for a CDMO like Vetter, AI creates real value only when it is applied in a concrete, trans­parent, and secure way.”

In an enterprise context, AI’s applica­tions can broadly be categorized into three areas:

• AI for Everyday Efficiency – Tools that simplify daily tasks such as writing, planning, or summarizing. The goal: make work easier, faster, and more ef­ficient.
• AI for Process Transformation – Solu­tions that rethink workflows and signifi­cantly improve efficiency. An example would be robotic process automation.
• AI for Strategic Differentiation – Initia­tives that strengthen the business model and secure its future relevance.

“In short, AI can deliver impact on multiple levels — from incremental effi­ciency gains to strategic realignment,” says Mr. Ottinger. “Whether it’s analyzing large datasets, optimizing processes or enabling predictive maintenance, AI adds value wherever patterns can be detected and risks are identified early.”

However, he says, AI does not replace experience, judgment, or accountability — especially in a highly regulated environ­ment like the biopharmaceutical industry. “One of the greatest myths surrounding AI is that technology alone can think, decide, or lead,” he says. “In reality, AI is a powerful enabler, but its success depends on the people who guide and apply it responsibly.”

Vetter embeds appropriate technol­ogy features in its corporate strategy to navigate disruption, build resilience, and enhance competitiveness. As a CDMO, the company leverages AI technologies, such as Natural Language Generation and Ro­botic Process Automation, to streamline quality and controlling processes, mini­mize risks of human errors, and accelerate reliable decision-making.

“These tools can optimize workflows and free up resources, ultimately leading to improved precision and reliability,” says Mr. Ottinger. “Looking ahead, we see the greatest potential of AI in predictive main­tenance, remote services, and intelligent assistance systems — solutions that not only can boost operational efficiency but also contribute to better patient outcomes. The key lies in balance: using technology where it strengthens us — while maintain­ing a clear human compass. In the end, it’s not about Artificial Intelligence, but about real responsibility.”

VectorSeek: Conversational Search for All Your Documentation

A persistent myth in the pharmaceuti­cal industry is that organizations need be­spoke AI systems before they can meaningfully benefit from the technology. In reality, many of the bottlenecks slowing drug development stem not from a lack of advanced modeling, but from the inability of teams to find and use the data they al­ready have.

This is where VectorSeek, a private, domain-specific AI search platform, deliv­ers immediate value. Rather than replac­ing existing infrastructure or forcing teams into rigid new workflows, VectorSeek in­dexes an organization’s entire internal knowledge base or external website and transforms it into a secure, conversational search engine. Scientists and operational leads can ask questions in natural lan­guage and receive citation-backed an­swers sourced directly from their own controlled content.

The result is dramatically faster access to institutional knowledge, which directly supports the very AI-driven acceleration the industry is striving toward.

“I like to think of it as ‘un-generative AI,’ because it’s not making up new stuff, it’s pointing you to data that you already have,” says Mike Walker, Co-founder. “In drug development, where precision and compliance matter, AI shouldn’t replace your expertise, it should be used as a tool to make your company more efficient.”

VectorSeek emphasizes privacy and governance and the model can operate on a customer’s proprietary data without using it to train any external systems, ad­dressing a core concern among CDMOs, CROs, and pharma organizations evalu­ating AI adoption.

In a landscape where AI promises un­precedented speed, VectorSeek enables the clarity, traceability, and informed deci­sion-making that allows that speed to translate into real-world impact, without sacrificing scientific rigor or regulatory confidence. Domain knowledge specific to your company and the products you’re cre­ating may be buried in various documents, VectorSeek makes that expertise available to everyone in your organization.