Leveraging Synthetic Advisory Boards for Sharper Biopharma Decisions

Synthetic advisory boards (S-ABs) are emerging as a strategic decision-support tool that can help biopharma teams navigate complex decision-making. In an ideal case, key decisions are shaped by carefully curated insights generated by a committee of experts. The reality is that the pace of change and the proliferation of information sources, including congress abstracts, real-world data releases, payer signals and competitor activity, means that decision-makers need new tools to stay ahead of the competition.

S-ABs help close the distance between that ideal and that reality. They are simulated discussions held by a set of artificial intelligence (AI)-enabled “digital twins” — virtual personas configured to represent a health stakeholder whose expertise or point of view is required to inform a decision. Each twin responds to evidence, messages and scenarios differently, reflecting specific expertise, priorities and even biases, all conducted under human oversight. Because S-ABs can run in minutes rather than weeks, teams can use them to explore questions, test hypotheses and sharpen their thinking before requiring experts’ input time.

Where synthetic and live boards earn their places

Live advisory boards remain the gold standard for business-critical moments, where critical decisions require human validation and nuance based on hands-on experience. Their constraint is volume, as a single board can explore only a narrow set of scenarios, and convening one takes time and expert availability that teams rarely have to spare.

S-ABs work differently. Instead of convening on a fixed schedule set months in advance, they can be run in response to data and evidence as they become available. This allows them to iterate on far more scenarios and stakeholder configurations than a live session can reasonably accommodate, making them valuable at many points in the product life cycle (see Figure 1 below).

Running an S-AB first also expands the live board’s contextual role. Teams can stress-test preformed hypotheses and identify where genuine disagreement among experts is likely to surface, then arrive with calibrated questions, tighter stimuli and fewer blind spots. The discussion moves straight to the nuances that only human deliberation can resolve, rather than spending the first hour building context. Anyone who has sat through a live board knows that the richest exchange often begins in its final 30 minutes, sometimes forcing a second session to make up ground. Preparing this way does not negate relationship building. It tends to deepen it, since arriving prepared respects the experts' time.

Inside a synthetic advisory simulation

An S-AB is built by leveraging personas. Each one is a digital twin configured to represent a specific clinical archetype, such as a conservative academic specialist, early-adopter community oncologist or value-focused medical director. Behavioral science supports how each persona is built, with frameworks such as the capability, opportunity and motivation (COM-B) model enhancing the realism of the personas to guide motivations, biases and decision drivers. These personas leverage unique, structured data sources. A physician specialist persona, for instance, might draw on the latest congress proceedings and peer-reviewed publications, while a primary-care persona could lean more on established treatment guidelines and physician survey data that capture real-world prescribing perspectives.¹

An insight-rich dynamic emerges when those personas interact. A multipersona dialogue lets agents respond not just to the original prompt but to each other, surfacing tensions and trade-offs that single-respondent simulations cannot. Personas are configurable as well. A team can direct the system to make a given archetype less conservative, more price-sensitive or more attuned to a specific subspecialty and then observe how the simulated discussion shifts. Outputs can be delivered as fully simulated transcripts for teams that prefer to write the report in their own voice or as templated, actionable insight summaries that match how an organization already presents its advisory work.

Designing for transparency in high-stakes decisions

Hallucination is the first concern people typically raise about AI-generated answers, and the concern is legitimate. In one survey, a large majority of physicians reported encountering erroneous outputs from general-purpose large language models (LLMs), with error rates dependent on whether queries referenced a specific molecule or brand name.² For decisions that touch on evidence strategy or commercial positioning, this type of variance carries real weight, and appropriate safeguards are needed.

An S-AB built for health care should make three things visible in every output: the source data behind each insight; where grounded data ends and agent inference begins; and a way to trace any claim back to its published origin. Purpose-built platforms should be conceived and built from the ground up to turn data into defensible, health care-grade output. This combination gives reviewers an audit trail that holds up under intense regulatory scrutiny, something general-purpose LLMs rarely deliver. When a finding can be traced from claim to inference to source, the conversation between data scientists, medical affairs reviewers and senior decision-makers changes. The debate moves from whether the synthetic output can be trusted to what it is saying and why it is being said.

The cross-functional case for S-ABs

In execution, S-ABs rarely sit inside a single function for long. Medical affairs teams may use them to test communication strategies, refine evidence priorities and pressure-test launch narratives before convening human experts. Market access teams can run scenarios around evolving frameworks, such as the EU’s Joint Clinical Assessment, and model how different payer archetypes are likely to respond to pricing and access positioning. For commercial teams, the same engine helps probe value propositions and competitive scenarios, including those shaped by recent policy shifts. Running cross-functional questions through the same underlying personas and data sources means decisions across the life cycle reflect a coherent base of evidence rather than the fragmented inputs each function tends to gather on its own.

Earlier applications of synthetic personas were essentially reactive. They were used to answer a specific question, generate a transcript and then file the report. The future, however, lies in “always-on” S-ABs, where personas continuously scan published evidence, congress proceedings and emerging signals. They’ll flag the implications of essential decisions before the next strategic review begins, and sometimes even before the next congress takes place. The S-AB itself then becomes one agent inside a larger ecosystem. It works alongside health technology assessment agents, analog-finding agents and market-forecasting agents to determine the returns on investments in evidence generation.

The benefit of striking a balance

The volume of evidence teams must factor in keeps increasing, as does the cost of inaction. Every congress, data readout and payer signal that arrives unexamined can lead to an incomplete or delayed decision. S-ABs do not remove that pressure, but they give teams a way to test more hypotheses, commit to budget with more evidence and reserve scarce expert time for the conversations that genuinely need it. The practical step is a small one. Pick a single decision on the near horizon, simulate the questions worth exploring and bring only the strongest queries into the live room. Teams that build the habit now will turn growing evidence streams into a strategic advantage, while those that wait will defer crucial decisions that come with a hidden price tag.

References

1. Quinones N, O’Loughlin A, Shrivastava A, Bolic J, Arora I. Data-driven decision making through synthetic ad boards: accelerating access to insights. White paper. IQVIA. 2026. https://www.iqvia.com/library/white-papers/data-driven-decision-making-through-synthetic-ad-boards
2. Kim Y, Jeong H, Chen S, et al. Medical hallucinations in foundation models and their impact on healthcare. Preprint. arXiv. 2025. https://arxiv.org/abs/2503.05777