Life Sciences & Healthcare
Agentic AI as Autonomous Innovation Director
In the life sciences sector, agentic AI may be evolving from a predictive tool into an autonomous driver of research and development. Acting as a virtual Research & Development Director, the AI identifies promising research areas, sets priorities, designs experiments, allocates resources, and analyses results - often without human prompts. In the pharmaceutical industry, Agentic AI could significantly reduce development costs by automating laboratory processes, simulating preclinical tests more efficiently, thus lowering the number of costly failures in early development stages.
Drawing from scientific literature, clinical trial data, patents, market insights, and lab results, the system dynamically adapts research strategies in real time. It may independently suggest follow-up studies, simulate trials, or recommend assigning tasks to internal teams or external CROs. Outcomes are evaluated against regulatory benchmarks and strategic goals, with breakthrough signals flagged for escalation.
In effect, the AI may become a Co-Head of Innovation - steering pipelines, spotting market white space, and proposing new programs, e.g. novel drug formulations or diagnostic tools. This frees human experts to focus on strategic validation, ethical review, and regulatory navigation, while the AI helps to accelerate cycles, sharpens investment decisions, and drives the translation of science into product.
Regulatory Compliance
When agentic AI systems assume co-project management roles in research and development for medical products, they enter a highly regulated space. Unlike prompt-based generative AI tools that assist with discrete tasks, Agentic AI operates autonomously - suggesting initiating workflows, prioritise decisions, and coordinating across internal and external teams with minimal human intervention.
For example, such systems might compile technical documentation for a Class IIa medical device, pre-structure interactions with notified bodies, or generate risk assessments under the Medical Devices Regulation. They may also produce drafts for labelling content, clinical evaluation reports, or post-market surveillance plans.
In GxP-regulated environments (e.g. GMP), the use of agentic AI is subject to additional constraints. Documentation created or processed by such systems must comply with specific requirements for documentation, version control, attribution, traceability and auditability (see EU GMP Guidelines).
Health advertising and regulatory approval: Internal autonomy meets external constraints
The use of agentic AI in pharmaceutical research & development and commercialisation also has significant implications for advertising compliance, particularly in highly regulated product categories such as medicinal products, medical devices, and cannabis-based therapies.
In these sectors, healthcare advertising is subject to strict legal constraints. For example, under German law the Heilmittelwerbegesetz (HWG) imposes additional requirements beyond the general Unfair Commercial Practices framework. Promotional claims must be scientifically substantiated - typically by Gold Standard evidence, such as randomised, placebo-controlled clinical trials. Misleading suggestions of effectiveness, exaggeration of study results, or comparisons lacking proper scientific basis are prohibited. Commonly cited HWG violations include:
- Referring to “risk-free” therapies,
- Implying guaranteed success,
- Advertising to unauthorised target groups (e.g. laypersons for prescription-only medicines),
- Using emotive patient testimonials, and
- Presenting unverifiable “breakthroughs” without proper contextualisation.
If an agentic AI system acts as an outward-facing recommender or communicator e.g. by generating promotional messages or explaining treatment advantages on a website or digital sales interface, it is directly bound by these advertising laws. This mirrors the concerns discussed in the sales use case above: autonomy does not exempt from compliance.
Autonomous generation of claims, if not properly reviewed and validated, can undermine advertising compliance and expose companies to legal and reputational risks. For this reason, the outputs of Agentic AI must be embedded in robust review loops, ensuring that no claim - direct or implied - enters public communication without meeting all formal advertising standards
Agentic AI and patent law: Human inventorship in machine-led innovation
The increasing use of agentic AI in research and development also affects patent law.
WIPO distinguishes between four categories: AI models (new algorithms), AI-assisted inventions (human uses AI as tool), AI-based inventions (AI integrated into a product), and AI-generated inventions (autonomously created by AI without human input).
The decisive legal line is clear: only natural persons can be named as inventors. Court and office practice worldwide, most prominently in the DABUS cases, confirm: no human inventor, no patent. It is acceptable to note that a human inventor used AI in the inventive process, but the AI system itself cannot be listed as inventor. The inventor designation exists to allocate rights, ensure entitlement claims, and safeguard compensation: functions a machine cannot fulfil.
Importantly, for patentability it does not matter how the invention was made. Both European and international practice make clear that the origin of the inventive idea - whether by human reasoning, accident, or with the help of AI - is irrelevant. What counts is whether the invention itself is novel and inventive. Thus, inventions developed with the assistance of AI can be patentable, provided that a natural person is designated as inventor.
For companies deploying agentic AI, the challenge is to keep processes compliant:
- A natural person must always be identifiable as inventor, typically through problem definition, supervision, or evaluation of AI outputs.
- Statements that the invention was generated “by AI alone” risk formal rejection.
So, whilst agentic AI may redefine how inventions are made, under current law inventorship and rights remain firmly anchored in humans.
What to consider next:
- Life sciences companies should establish strict governance for autonomous R&D workflows. This includes defining how far the AI may drive research decisions, enforcing GxP-compliant documentation and traceability, and ensuring that all advertising-relevant outputs undergo human medical - legal review.
- For patent matters, maintain clear records of human contribution so that a natural person can always be named as inventor, even where AI plays a major role in generating ideas.