AI in Pharmacokinetics: From Faster Analyses to Better Decisions

Artificial intelligence is often discussed in terms of speed, faster analyses, faster workflows, faster outputs. But speed alone does not transform drug development. The real impact of AI in pharmacokinetics and clinical pharmacology lies in how it changes decision-making, not just how quickly calculations are performed.

It is also worth acknowledging that any answer about AI today may not be the same answer in a month or two. The pace of change is extraordinary, particularly in how AI is being adopted across the pharmaceutical industry. Still, several clear patterns are emerging that point to where real value is being created, and where it is not.

Three Ways AI Is Changing Pharmacokinetics Today

There are currently two well-established ways AI is influencing pharmacokinetics and clinical pharmacology, with a third beginning to emerge.

1. Machine Learning for Data Analysis

Machine learning has been used in pharmacokinetic and pharmacodynamic analysis for several years. Early ambitions focused on replacing traditional population PK modeling with neural networks as a faster, more efficient way to describe PK/PD relationships.

Progress in this area has been meaningful but slow. Adoption remains limited, largely because these approaches are still exploratory and can be challenging to implement in regulated environments. While technically promising, they have not yet displaced established modeling frameworks in routine practice.

2. Large Language Models as Regulatory Intelligence Engines

The second and fastest growing area is the use of large language models (LLMs) to support daily scientific and regulatory activities. Many companies now offer custom-built LLMs designed specifically for regulatory intelligence.

These systems allow users to ask questions such as: Which drugs have been approved without a CYP3A4 DDI study? The AI can search across literature, regulatory guidances, and prior approvals to return relevant examples. Some tools can even propose entire clinical development strategies based on historical precedent.

While general-purpose tools such as ChatGPT, Gemini, or others can perform similar functions, curated, domain-specific systems often provide higher-quality and more reliable outputs. This represents a meaningful shift in how regulatory strategy is informed, moving from manual research to AI-assisted synthesis.

3. Automation and Execution of Analyses

The third area, and arguably the most transformative, is the automation and execution of pharmacokinetic analyses themselves.

There is an important distinction between using AI to recommend what analysis should be done and enabling AI to actually perform that analysis. Today, LLMs primarily function as fast encyclopedias, summarizing information, providing guidance, and collating content. The next stage is building AI agents that can take a dataset, apply the correct analytical methods, run calculations, and produce results using integrated tools. These AI agents will need cloud-based tools that can be operated through APIs to accomplish this work.

This is the frontier. When analytical platforms can be directly leveraged by AI systems, scientists will no longer need to master every software package. Instead, they will focus on defining the scientific question and interpreting the outcome.

Where AI Is Ready, and Where Humans Still Lead

One area that is particularly ready for AI-driven transformation today is regulatory strategy in clinical pharmacology. Large language models excel at reading, interpreting, and synthesizing vast amounts of documentation far more efficiently than any individual human could.

By providing context, such as drug class, molecule type, or development goals, scientists can use AI to generate informed starting points based on prior approvals. Human expertise remains essential to assess whether those recommendations are appropriate for a specific compound or patient population. However, if AI can complete 80% of the groundwork, it accelerates decision-making dramatically.

In contrast, AI systems today are not yet capable of reliably conducting numerical analyses independently. Humans are still responsible for running analyses, reviewing diagnostic plots, and validating outputs. That will change. When AI systems can evaluate multiple models, select optimal approaches, and summarize results, the productivity gains for the field will be substantial.

The Biggest Misconception About AI in Pharmacokinetics

The most common misconception about AI in pharmacokinetics and clinical pharmacology is that it will replace human workers. This misunderstanding exists on both sides.

From a management perspective, AI is sometimes viewed as a cost-saving tool that reduces the need for staff. From the employee perspective, there is fear of job loss. Both views miss the point.

AI does not replace scientists. It replaces busy work.

Scientific work is the application of the scientific method: generating hypotheses, designing experiments, and evaluating results to confirm or refute those hypotheses. Humans will always be required for this work because it demands creativity, judgment, and the ability to imagine possibilities that do not yet exist. 

AI enables scientists to spend more time on this core scientific work by handling repetitive, routine tasks, such as generating tables or executing standard analyses. The result is not fewer scientists, but scientists doing more impactful work.

A useful analogy is the transition from horse-drawn carriages to automobiles. While cars replaced horse-drawn transport, they ultimately enabled people to travel faster and more efficiently. Skills adapted. Roles evolved. The same will happen in pharmacokinetics and clinical pharmacology.

Standardization Is the Real Opportunity

Pharmacokinetics has historically lacked conformity. Analyses often vary based on individual expertise, preferences, and subjective decisions. While expertise is valuable, this variability has not always served the discipline well. It has contributed to the perception that PK is more complex than it truly is and has limited efficiency and standardization.

This lack of standardization often reveals itself too late. Issues such as incorrect sampling times or protocol misunderstandings are frequently identified only after a study has concluded, when nothing can be corrected. These problems typically arise because experts were not sufficiently involved during study design and execution.

AI creates an opportunity to shift expertise upstream. By freeing scientists from routine analytical tasks, their knowledge can be applied earlier, during protocol development, site training, and study execution. This leads to better-designed experiments, higher-quality data, and more consistent outcomes.

Perfect agreement on metrics such as half-life estimates is not required. One half-life estimate does not determine the success or failure of a drug. What matters is consistency, robustness, and experimental quality.

New Questions Become Possible

As AI removes traditional analytical bottlenecks, the nature of scientific inquiry changes. Researchers can spend more time on meta-analyses, cross-study evaluations, and cross-program insights.

Large pharmaceutical companies have data from hundreds or even thousands of development programs. AI can mine this information to identify trends, uncover overlooked opportunities, and support drug repurposing efforts. Entire classes of questions that were previously impractical become feasible.

The Breakthrough Ahead: Simulation at Scale

Looking five to ten years ahead, the defining breakthrough in AI for pharmacokinetics and clinical pharmacology will be AI’s ability to conduct analyses end-to-end.

Providing AI with a dataset and protocol and having it generate a complete analytical report will be transformative, but that is only the first step. Once AI can link analyses together, it can enable sophisticated simulations, from preclinical dose optimization to full clinical trial simulations across multiple designs.

This capability represents a fundamental shift in drug development. Extensive in silico exploration before dosing a single patient will allow teams to run fewer, more targeted trials with higher probabilities of success, ultimately accelerating patient access to effective therapies.

Where Aplos Fits

At Aplos Analytics, the focus is on enabling repeatable and reproducible pharmacokinetic analyses through both manual interaction and automation. The goal is not to build AI for its own sake, but to provide reliable analytical tools that AI agents can leverage.

AI agents, like skilled professionals, require tools to do their work. Aplos NCA serves as one of those tools, an execution layer that enables standardization, automation, and scalability. By combining AI-driven decision support with robust analytical execution, human expertise can be redirected toward hypothesis generation, experimental design, and interpretation.

Conclusion

AI is not the scientist. It is the force multiplier.

The teams that benefit most will not be those that simply claim to use AI, but those that meaningfully integrate it into how decisions are made. By embracing standardization, shifting expertise upstream, and freeing scientists from routine tasks, the field of pharmacokinetics and clinical pharmacology can evolve toward better science, and better outcomes for patients.