By Jason Campagna, biotech executive, as part of the From The Trenches feature of LifeSciVC
I recently joined Inventiva Pharma, a late-stage company developing a therapeutic for metabolic disease and MASH, a field that is both scientifically complex and once again visible. It is an interesting moment to return to late-stage work after several years in early-stage biotech. Early-stage life is oriented toward possibility and invention. Late-stage life is oriented toward consequence and delivery, and that difference becomes apparent almost immediately.
In earlier essays I described the capability stack, the collection of technologies and tools beneath a therapeutic, and the organizational stack, the system that must carry those tools into practice. The widening space between them is new. The capability stack is compounding faster than the organizational stack can absorb. They no longer move together. Their separation is where the hard problems live, where time reverses its logic and each day brings less room to explore and more urgency to decide. When I describe these as hard problems, I mean something specific. Early-stage development is demanding, but its demands are exploratory as much as procedural: documenting biology, mapping uncertainty, designing early trials, understanding exposure, and building an initial picture of behavior. The work is structured around learning. The system expects uncertainty and tolerates failure because its purpose is knowledge accumulation. Even setbacks add information.
Late-stage development is different, and returning to it makes that difference clear. What once felt open now feels bounded by commitments already made. Systems that once made the work seem effortless now reveal their limits. Late-stage development rewards precision and continuity. The cadence is set not by what teams hope to discover but by what must hold to bring a therapy to patients. Much of what matters is quiet and rarely noticed until it is tested. Early-stage environments can absorb iteration. Late-stage environments tolerate very little of it. The problems become hard because at some point you must assume you understand enough biology to scale. The phase 3 program must work because the phase 2 signal held. The safety profile must remain stable because the aggregate data suggest it will. Entire medical affairs and commercial plans begin on foundations that feel familiar, but organizational sediment can obscure how fragile those foundations are. Commitments expand. Timelines narrow. Consequences sharpen. Planning shifts from kilograms for trials to tons for global supply. The questions shift from whether the biology is plausible to whether the system can withstand the scrutiny of regulators, health systems, and payers across regions. Scientific uncertainty becomes operational and financial.
That is the point at which the work becomes hard in the way I mean it. You are no longer learning. You are committing.
In my earlier essay on biotech’s strategic infrastructure moment, I argued that value was no longer defined solely by the molecule but by the system that surrounded it. The geopolitical shifts of the past several years made this easier to see. Supply chain fragility, manufacturing sovereignty, and emerging federal infrastructure programs placed deployment on equal footing with discovery. Early-stage groups responded by building new layers. The capability stack thickened. What I did not fully appreciate then, but feel now, is how slowly the organizational stack adapts in late-stage life. Processes harden for reasons that are legitimate — regulatory expectations, patient safety, data integrity, commercial rigor — but those same reasons create friction when new capability tries to enter the system. A better tool does not immediately translate to a better outcome. Late-stage systems must absorb the cost of change before they benefit from it.
It is easy to dismiss this as conservatism, but I believe that much of it is structural. A later-stage organization is shaped by commitments already made. It holds inspection timelines, quality milestones, reimbursement strategies, evidence plans, and global sequencing decisions that cannot be rewritten because a new capability has appeared. The early-stage instinct is to improve the toolset. The late-stage reality is that improvement is only valuable if the system can tolerate the change.
What is new is the velocity gap. Early-stage teams build with assumptions about how the system will respond downstream — rapid CMC transitions, flexible manufacturing, resilient supply chains, and trial infrastructure that can pivot as needed. These assumptions rarely survive into late stage. What remains is obligation. And obligation without context is where systems fail. Late-stage organizations confront regulatory shifts, payer expectations, regional constraints, and timelines that compress rather than flex. Stated differently, early stage imagines what could be. Late-stage bears what must hold.
This difference in accountability resonates with me in ways that have been unexpectedly sharp and even a little painful. Before biotech I spent years in academic anesthesia and then in private practice as part of a twenty-eight-physician group responsible for the daily execution of care. The rhythm was precise and unyielding. Cases had to run on time. Surgeons needed to return to their offices to see patients. Every encounter carried both clinical and financial implications. At the same time, I served as the chief medical quality officer for the hospital system, responsible for moving the institution toward evidence-based practice, pay-for-performance expectations, and reimbursement tied to outcomes. The challenges in that environment mirror the challenges here: systems that depend on routine, coordination, and consequence.
But there is a critical difference. In medicine the work continues regardless of how difficult the day or week has been. Operating rooms open again. More patients arrive. The system stretches because the clinical imperative demands it. Late-stage drug development carries a different form of consequence. A difficult day or week may begin with the same type of event — a safety signal or a quality issue — but the effects do not dissipate in the same way. In clinical care the system resets. In drug development the event can redefine the program, and in late-stage work it often does, because the timeline does not forgive. That difference helped me see more clearly why the gap between the capability stack and the organizational stack matters so much now.
I have found that the most useful models for this work come from fields where timing, clarity, and consequence converge. As I wrote previously, special operations and medicine operate this way, and their principles translate more directly than anything I have seen in management literature. Business theory often reaches similar conclusions, but from a greater distance. In practice the rule is simple: authority should move to the person with the clearest context. It is a functional requirement. The statistician inside the data flow of an ongoing study. The pharmacovigilance lead who senses a pattern before it appears in the signaling data. The regional medical lead who understands why a label phrase will block access in one country and not another. On many days those are the people who should lead the decision. This only works when the organization has done the slow work of building shared context.
This is where the stack disconnect becomes an organizational problem rather than a technical one.
Early-stage capabilities often promise acceleration, and early-stage teams expect downstream systems to absorb advances with the same agility. Assumptions accumulate quietly: manufacturing can pivot without penalty; quality systems can absorb new data flows; regulatory pathways will accommodate new methods; medical affairs will translate emerging complexity into something clinicians can use. Each assumption is reasonable on its own. Together they form an implicit model of a system that can stretch indefinitely. Late-stage organizations do not have that latitude. They carry the accumulated commitments of the program. They cannot accept acceleration as a free parameter. Every gain in speed demands assurance that the system will remain stable under inspection, reproducible in manufacturing, defensible to regulators, and credible to payers. This requires a clear articulation of where the system can stretch and where it cannot. Without that framework, claims of acceleration become noise.
The real work shows up in decisions that cannot be abstracted away. When to file an NDA and how much data is enough. Whether to dual-source drug substance or accept the risk of a single supplier. How far to extend long-term safety exposure, not only as a scientific or ethical matter but as a regulatory and reimbursement requirement. These are not theoretical tradeoffs. They commit an organization to a specific course of action. They determine whether a therapeutic can reach patients and whether the company can return value to the long line of inventors, operators, and investors who have already committed time and capital. They define what the system must now sustain.
So what to do about it.
The gap will not close by asking late-stage teams to behave like early-stage groups or by assuming new capability will carry itself into practice. Most of the work lies in creating deliberate connections between the two stacks so expectations and constraints are visible at the same time. One answer is to create explicit handshakes between the stacks. Someone has to translate a capability into operational terms and surface its consequences early. A small integration group working beside a pivotal program can remove months of delay by capturing the cost of adoption before it becomes a crisis. When implications are visible early, much of the cynicism that accompanies late change begins to recede. This is also where familiar business vocabulary tends to surface — change management, alignment frameworks, integration models. The labels are secondary. The function is what matters: make consequences visible early enough that the system can absorb them without destabilizing itself.
Narrative is another form of handshake. Early-stage groups speak the language of capability and potential. Late-stage groups speak the language of credibility and durability. If a capability cannot be expressed in late-stage terms, it is not ready. It remains aspiration rather than practice. A common example is a pharmacovigilance vendor promising faster triage or improved signal detection. In early-stage settings that may sound sufficient. In late-stage settings it is not. The questions become concrete: Can they meet inspection standards across regions. Can they maintain validated audit trails. Can their system remain consistent across thousands of cases when a label is under review. Can safety leadership explain the basis of a signal if part of the logic is automated. If those translations do not exist, the capability is not ready.
Medical Affairs often becomes the definitive handshake here. Their work is sometimes reduced to perjoratives — messaging, unprompted recognition, field alignment — but those are partial descriptions of a much harder job. They set the external frame: how clinicians, thought leaders, healthcare systems, and payers will understand the therapy, what evidence will matter, and where the claims must hold. If the narrative remains rooted in capability and potential, the organization will fail at the point of consequence because none of the external stakeholders operate in that register. Medical Affairs must translate innovation into a story that is scientifically coherent, clinically relevant, and durable under challenge. When the capability stack outruns the organizational stack, this becomes defining work.
It is easy to treat the gap I’m describing as a failure of intent and to assign blame to policy shifts, risk committees, or cultures shaped by different tempos. Each plays a role, but the gap, as I stated above, is structural. Late-stage constraints exist because obligations are real. Patients, regulators, and payers are not theoretical. The system carries commitments that cannot be undone simply because a new capability has appeared or because acceleration seems desirable.
The challenge is to help the organization absorb new capability without destabilizing itself. That is not a matter of speed. It is a matter of design, and the work does not resolve neatly. Early stage opens into possibility while late stage narrows into consequence, and the space between them now carries most of the real difficulty. These problems are not solved by accelerating one side or restraining the other. They are solved by making the system explicit: where it can stretch, where it cannot, and what each decision will displace downstream. Organizations that learn to operate with this level of clarity will carry programs across the gap. Those that do not will continue to mistake speed for progress. The space between the stacks is no longer a transition. It is the operating environment for modern late-stage biotech.
