Agile in Drug Development: Where It Works and Where It Doesn't

Introduction

Few terms get misapplied in biotech project management as reliably as "Agile." Teams use it to mean flexible. Leaders use it to mean fast. And in some corners of the industry, it has become a catch-all for any process that lacks structure. The problem is that in drug development, the wrong kind of flexibility can delay a program, compromise data integrity, or create regulatory exposure that takes years to resolve.

Agile project management is a legitimate and powerful methodology — but its value depends almost entirely on context. Applied to the right work, it accelerates learning and enables teams to respond to scientific uncertainty. Applied to the wrong work, it creates the illusion of progress while quietly eroding the documentation backbone that regulators require. This post is about making that distinction clearly — and giving biotech and pharma project managers a practical framework for deciding where adaptive methods belong in a drug development program, and where they don't.

Laboratory workstation showing drug development process with data analysis

Agile Is Already Inside the PMBOK Guide — Just Not the Way Most People Think

The first thing to understand is that the Project Body of Knowledge (PMBOK) Guide (link) does not treat agile as a separate or competing methodology. It describes a spectrum of development approaches — predictive, adaptive (agile), and hybrid — and frames the project manager's job as selecting and tailoring the right approach for the specific context. Critically, PMBOK includes "Regulations" as an explicit variable influencing development approach selection, noting that environments with significant regulatory oversight tend to use predictive approaches "due to the required processes, documentation, and demonstration needs." Safety requirements are treated similarly: products with rigorous safety requirements typically demand significant up-front planning to ensure compliance — a description that fits virtually every phase of clinical drug development.

This is an important framing point. The question in biotech is not whether to "go agile" as an organizational philosophy. The question is: which development approach is appropriate for this specific deliverable, on this specific project, within this specific regulatory and scientific context? The answer is almost never a single methodology applied uniformly across an entire program.

Where Agile Genuinely Works: Early Discovery, Data Science, and Digital Platforms

The areas of drug development where adaptive approaches have the most legitimate and productive application share a common characteristic: the deliverables are not yet fixed, feedback is rapid, and change is expected and informative rather than disruptive.

• Early-stage discovery and target research. When a team is screening compound libraries, running structure-activity relationship (SAR) analyses, or testing multiple biological hypotheses in parallel, they are operating in exactly the high-uncertainty, high-feedback environment that Agile was designed for. Sprint-based planning — breaking work into short, time-boxed cycles with defined outcomes and review points — maps naturally onto the iterative hypothesis-test-revise rhythm of discovery science. A two-week sprint reviewing binding assay results against a prioritized backlog of chemical series is not a compliance risk; it is scientifically sound project management.

  
  

• Computational and data science projects. AI-assisted drug design, in silico target modeling, platform development, and bioinformatics workflows are by now mainstream across biotech and pharma (link). These projects generate rapid feedback, evolve continuously, and rarely have fully specified requirements at initiation. Kanban boards, sprint reviews, and iterative backlog refinement are genuinely effective tools for this work.

  
  

• Internal software and technology platforms. Clinical Trial Management Systems (CTMS), laboratory informatics tools, and data visualization dashboards are another appropriate home for agile principles, particularly during requirements development and user interface design. Here, the PMBOK hybrid model applies well: adaptive approaches for the design and iteration of the product, with predictive processes governing Computer System Validation (CSV) and go-live activities downstream.

  
  

In all of these settings, what makes agile viable is the absence of binding external documentation requirements during the adaptive phase, the low cost of changing course, and the high value of fast-cycle learning.

Where Agile Does Not Belong: Regulated Trials, Submissions, and GMP Operations

It is equally important to be direct about where adaptive approaches are inappropriate — not because the methodology is flawed, but because the regulatory and scientific environment overrides its assumptions.

• Regulated clinical trials are the most consequential example. A Phase II or Phase III clinical trial operates under an approved protocol that constitutes a regulatory commitment. Sponsor obligations to IRBs, ethics committees, and health authorities are anchored to that protocol. Subjects are enrolled under informed consent documents tied to specific procedures and endpoints. Unilateral "iterations" to scope or methodology — the currency of agile — are not sprints in this context. They are protocol amendments, and each one triggers formal review, regulatory notification, and potential timeline impact. The pharmaceutical industry has been slower to adopt agile than the medical device sector, largely due to compliance concerns and fundamental misunderstandings about what regulatory expectations look like in a GxP environment. That hesitation is not entirely unfounded. GxP compliance — spanning Good Manufacturing Practice, Good Clinical Practice, and Good Laboratory Practice — exists precisely to ensure that changes to validated processes and controlled procedures follow a structured, documented path. An agile sprint that modifies a data collection procedure mid-trial is not a feature; it is a protocol deviation.

• Regulatory submissions — INDs, NDAs, BLAs, and equivalent filings — are predictive work by nature. The content of a regulatory submission is defined by FDA and EMA expectations, each section governed by guidance documents, technical dossier standards, and common technical document (CTD) format requirements. There is no meaningful backlog to prioritize here, no sprint review that modifies scope. The discipline required is completeness, traceability, and version control — all of which are hallmarks of predictive project management.

• GMP manufacturing projects — process development, tech transfer, facility buildout, process validation — are similarly incompatible with core agile principles. Manufacturing processes in a GxP environment are validated, and validation is inherently a predictive, sequential activity. Iterating through design changes in sprints is appropriate during formulation development; it is not appropriate once a process is under validation, where change control procedures govern every modification.

The risk of applying agile to regulated work is not just procedural. It creates documentation gaps that auditors find, data integrity questions that regulators raise, and, in the worst cases, patient safety exposures that no project velocity metric can justify.

The Hybrid Model: Where Most Drug Development Programs Actually Live

Understanding the limits of both pure agile and pure predictive approaches leads to a more useful conclusion: most drug development programs are best managed as hybrids, with methodology selection driven by the nature of each work stream rather than an organization-wide philosophical commitment.

This is precisely the model the PMBOK describes when it defines hybrid approaches as combining adaptive and predictive elements based on deliverable characteristics, regulatory requirements, and the degree of scope stability. In practice, a small molecule development program might run discovery compound prioritization on sprint-based cycles while simultaneously managing IND-enabling studies under a fully predictive plan with stage-gated milestones and formal change control. These two workstreams can coexist within the same program provided the governance structure is clear about which methodology applies where — and why.

This is also a natural extension of the Ganvion 6-Stage Small Molecule Development Framework, in which stages are defined as deliverables, decisions as work packages, and criteria as activities. As a program advances from discovery (where adaptive methods are appropriate) through to IND-enabling and clinical stages (where predictive governance is non-negotiable), the development approach should evolve accordingly. The stage-gate decision points become the natural governance boundary between adaptive and predictive work — a place where the program pauses, evaluates, and formally transitions to a more controlled execution mode.

ICH E6(R3): Regulatory Modernization Is Not the Same as Regulatory Flexibility

It is worth addressing a common misreading of the recently finalized ICH E6(R3) Good Clinical Practice guideline. ICH E6(R3) offers modernized recommendations for designing and conducting clinical trials efficiently, safely, and innovatively, incorporating risk-based monitoring, decentralized trial elements, and a proportionate approach to oversight. The updated guideline introduces a new Risk Proportionality principle, emphasizing that risk management strategies should be tailored to the level of risk posed by the trial.

Some have interpreted these updates as a signal that clinical operations are becoming more "agile." That reading conflates two different things. ICH E6(R3) modernizes how clinical trials are overseen and monitored — it does not alter the fundamental requirement that trials be conducted according to a pre-specified, IRB-approved, and regulatory-submitted protocol. Risk-based monitoring is still monitoring. Centralized data review is still review. A decentralized trial design must still be fully documented and prospectively justified. The principles of Good Clinical Practice — informed consent, data integrity, investigator accountability — remain unchanged. Project managers who treat E6(R3) as a green light for less structured clinical execution misunderstand both the guideline and the risk it was designed to manage.

Conclusion

The question for biotech and pharma project managers is not whether to embrace agile — it is whether they are applying it with the contextual discipline that drug development demands. Agile in discovery is an asset. Agile in a Phase III trial is a liability. Getting that distinction right requires not just methodology knowledge, but a clear understanding of what each development stage requires, what regulators expect, and what risks uncontrolled iteration introduces. At Ganvion Biotech Solutions, we help clients build program governance frameworks that match the right development approach to the right work — protecting regulatory commitments where it matters most while preserving the speed and adaptability that early-stage science requires. If your organization is navigating the agile-versus-predictive question, we can help you build the framework that answers it correctly.