Panel Discussion
Redesigning Clinical Workflows in the AI Era
Format: Moderated panel with structured questions
Duration: 90 minutes
Panelists: up to 4 (ideally 2 large pharma, 2 midsize pharma)
Moderator: Aga Rasinska
Goal: Clinical data workflows were engineered around constraints that no longer exist - paper records, batch processing, siloed teams, and end-of-study evidence assembly. AI, continuous data streams, and adaptive models don’t just improve those workflows. They make their underlying logic questionable. This panel asks whether what we’re all doing is still the effective thing to be doing and what we would build if we started from scratch today. The goal is productive discomfort.
Panelists Profiles:
Four panelists. Chosen for role diversity, company size contrast, and experience.
Panelist A — VP / Head Biometrics, Large Pharma
The senior voice. Has lived through multiple paradigm shifts. Knows what “regulatory-ready” means under pressure. Not opposed to change, but aware of what breaks when you move fast. Brings the weight of leadership, experience, and the question nobody wants to ask out loud.
Panelist B — Head of Clinical Data Science & AI, Large Pharma
Driving enterprise AI adoption from the inside. Building governance frameworks, validation approaches, and the infrastructure needed to get models into regulated workflows. Brings the gap between pilot and production, and what that gap costs.
Panelist C — Head of Biostatistics & Data, Midsize Biotech
Building without the legacy. Closer to the data, fewer layers between decision and action. Brings the agility perspective.
Panelist D — Clinical Data & Analytics Lead, Midsize Pharma
The hands-on builder. The person who actually makes the pipelines work - R, Python, cloud infrastructure, and increasingly agentic tooling. Brings ground-level reality to abstract frameworks. If the panel gets too strategic, Panelist D brings it back to: what does that mean for someone writing the code on Monday morning?
Structure & Timing:
Moderator Intro 5 min
Opening Round 10 min
Segment 1 — Where are we now 20 min
Segment 2 — AI in Regulated Practice 20 min
Segment 3 — The Updated Operating Model 20 min
Closing Round 10 min
TOTAL 90 min
Flow with example questions:
- Opening (5 min)
Set the frame.
- Opening Round (10 min)
Moderator addresses each panelist directly. No discussion yet - just position-setting.
- When you hear people talk about AI-driven clinical analytics — what is the gap between the vision being described and what you know would have to be true for it to meet the reality with a regulator?
- "You are building AI into regulated workflows at scale. What is the thing you wish understood about how hard that actually is - the thing that doesn't appear in talks?"
- "You are building without the legacy. What assumption do you see large organisations carrying that you have simply chosen not to inherit — and has that turned out to be the right call?"
- "When the conversation in a room like this turns to redesigning clinical workflows from first principles — what do you want the strategists to understand before they start drawing new org charts?"
SEGMENT 1 — Where are we now (20 min)
Purpose: Before anyone reaches for solutions, this segment demands a shared, honest reckoning with what is actually happening - not what could be better, but what is structurally challenging. Underlying tension: Is what's challenging a technology problem, a process problem, a design problem, or other?
Opening question:
"Before we talk about where we're going, let's be honest about where we are — not the aspirational present, the actual one. What is structurally challenging with how clinical data workflows operate today?
SEGMENT 2 — AI in Regulated Practice (20 min)
Purpose: Move the AI conversation from possibility to operational reality. This segment demands specificity: what is actually running in production, what broke on the way there, and where the regulatory ceiling actually sits right now. Underlying tension: There are two fundamentally different categories of AI use — AI that improves speed without changing the evidence, and AI that shapes the evidence itself. Most governance frameworks treat these as a spectrum. They may be two entirely different things.
Opening question:
"Let's make this specific. Not AI as a direction — AI as something your team is actually running, or tried to run, inside a regulated clinical workflow. What does that look like in practice? Not the demo. The production reality."
SEGMENT 3 — The Updated Operating Model (20 min)
Purpose: Make the organisational consequences of Segments 1 and 2 explicit and uncomfortable. Not a general discussion about change management. Underlying tension: The functional silos — statistics, data management, programming — were designed for sequential handoffs with clearly bounded accountability. An AI-assisted pipeline that spans all three in a single run doesn't respect those boundaries.
Opening question:
"The technology is moving. The workflow is moving. The operating model conversation — who owns what, how functions relate, how accountability is structured — is largely not moving. And if it is — why is it the last thing to change?"
Closing Round (10 min)
Moderator closes with a direct ask. Approximately 2 minutes per panelist. No cross-discussion.
Panel Discussion
Redesigning Clinical Workflows in the AI Era
Format: Moderated panel with structured questions
Duration: 90 minutes
Panelists: up to 4 (ideally 2 large pharma, 2 midsize pharma)
Moderator: Aga Rasinska
Goal: Clinical data workflows were engineered around constraints that no longer exist - paper records, batch processing, siloed teams, and end-of-study evidence assembly. AI, continuous data streams, and adaptive models don’t just improve those workflows. They make their underlying logic questionable. This panel asks whether what we’re all doing is still the effective thing to be doing and what we would build if we started from scratch today. The goal is productive discomfort.
Panelists Profiles:
Four panelists. Chosen for role diversity, company size contrast, and experience.
Panelist A — VP / Head Biometrics, Large Pharma
The senior voice. Has lived through multiple paradigm shifts. Knows what “regulatory-ready” means under pressure. Not opposed to change, but aware of what breaks when you move fast. Brings the weight of leadership, experience, and the question nobody wants to ask out loud.
Panelist B — Head of Clinical Data Science & AI, Large Pharma
Driving enterprise AI adoption from the inside. Building governance frameworks, validation approaches, and the infrastructure needed to get models into regulated workflows. Brings the gap between pilot and production, and what that gap costs.
Panelist C — Head of Biostatistics & Data, Midsize Biotech
Building without the legacy. Closer to the data, fewer layers between decision and action. Brings the agility perspective.
Panelist D — Clinical Data & Analytics Lead, Midsize Pharma
The hands-on builder. The person who actually makes the pipelines work - R, Python, cloud infrastructure, and increasingly agentic tooling. Brings ground-level reality to abstract frameworks. If the panel gets too strategic, Panelist D brings it back to: what does that mean for someone writing the code on Monday morning?
Structure & Timing:
Moderator Intro 5 min
Opening Round 10 min
Segment 1 — Where are we now 20 min
Segment 2 — AI in Regulated Practice 20 min
Segment 3 — The Updated Operating Model 20 min
Closing Round 10 min
TOTAL 90 min
Flow with example questions:
Set the frame.
Moderator addresses each panelist directly. No discussion yet - just position-setting.
SEGMENT 1 — Where are we now (20 min)
Purpose: Before anyone reaches for solutions, this segment demands a shared, honest reckoning with what is actually happening - not what could be better, but what is structurally challenging. Underlying tension: Is what's challenging a technology problem, a process problem, a design problem, or other?
Opening question:
"Before we talk about where we're going, let's be honest about where we are — not the aspirational present, the actual one. What is structurally challenging with how clinical data workflows operate today?
SEGMENT 2 — AI in Regulated Practice (20 min)
Purpose: Move the AI conversation from possibility to operational reality. This segment demands specificity: what is actually running in production, what broke on the way there, and where the regulatory ceiling actually sits right now. Underlying tension: There are two fundamentally different categories of AI use — AI that improves speed without changing the evidence, and AI that shapes the evidence itself. Most governance frameworks treat these as a spectrum. They may be two entirely different things.
Opening question:
"Let's make this specific. Not AI as a direction — AI as something your team is actually running, or tried to run, inside a regulated clinical workflow. What does that look like in practice? Not the demo. The production reality."
SEGMENT 3 — The Updated Operating Model (20 min)
Purpose: Make the organisational consequences of Segments 1 and 2 explicit and uncomfortable. Not a general discussion about change management. Underlying tension: The functional silos — statistics, data management, programming — were designed for sequential handoffs with clearly bounded accountability. An AI-assisted pipeline that spans all three in a single run doesn't respect those boundaries.
Opening question:
"The technology is moving. The workflow is moving. The operating model conversation — who owns what, how functions relate, how accountability is structured — is largely not moving. And if it is — why is it the last thing to change?"
Closing Round (10 min)
Moderator closes with a direct ask. Approximately 2 minutes per panelist. No cross-discussion.