Introducing Elicit Research Agent: Rigor that powers decisions

1 min read

Over the past year, AI has evolved to become a great thought partner. It can help us find information and suggest ideas. But has it become rigorous enough to help us make better decisions? How do we know that more information actually has an impact?

We believe that it’s not enough to find relevant information faster. We want to ensure that better information leads to better decisions. At Elicit, our ultimate mission is to scale up good reasoning for high-stakes decisions. We started by automating evidence synthesis, because good reasoning is evidence-based. It is now time to start working on supporting decisions more directly. Today, we share how we’ve taken the next step toward that vision.

Introducing Research Agent and BioDecisionBench

Elicit Research Agent is an AI environment built for high-stakes decisions. It gathers evidence across millions of scientific sources, public data sources including the web, and internal uploaded data. The agent also creates outputs like fully cited reports, data analyses, tables, visualizations, slides, and documents - because important decisions are shared across many people. It has a proprietary harness trained to avoid reasoning mistakes and weigh tradeoffs for hard decisions in science. The Research Agent is available through elicit.com or via API to bring into your workflows. It is already used for indication prioritization, trial design, laboratory research, scientific and clinical strategy, and patent research.

Here are the key attributes that set it apart from other agents and co-scientists:

A harness with reasoning principles

Research Agent is intentionally designed to reason through high-stakes decisions and a variety of data types with integrated reasoning guidance and process verification. To achieve this, we first had to define what rigorous research and decision support actually require. We looked at existing benchmarks for answers but found them to be insufficient because most measure narrow tasks or only evaluate the final answer. Evaluating drug development presents a good platform to evaluate high stakes decisions due to the complexity and limited information. So, we built BioDecisionBench, a benchmark designed to evaluate decision advice on high-stakes pharma decisions by evaluating the reasoning that goes into those decisions. BioDecisionBench tests for reasoning around preclinical study design, observational study design (e.g. target trial validity), and rigorous reasoning throughout the drug development process. When we evaluate on BioDecisionBench, Elicit highlights more key considerations for decisions (76.7% coverage vs. 68.8% for Claude Opus 5 Max).

Diverse data sources

Elicit Research Agent brings together different perspectives scattered across many sources: the scientific literature, clinical trials, patents, structured biology databases, regulatory databases, conference abstracts, commercial data sources, web sources, and your own uploaded internal data. Investigate multiple angles, compare conflicting findings, and understand how the different sources of evidence support or challenge a conclusion. Science is increasingly interconnected and cross-functional. An agent working across siloes helps base high-stakes decisions on the full picture.

Decision-ready outputs

The Research Agent transforms synthesized evidence into structured, fully cited outputs such as reports, tables, figures, visualizations, bioinformatics analysis, mathematical calculations, slide decks, and documents. Move from an exploratory question to meeting-ready materials that track the evidence and citations collected throughout.

Transparency by default

The Elicit Research Agent prioritizes a systematic and transparent process. Follow the agent's reasoning and methods in real time and inspect every claim with sentence-level citations.

Knowledge that compounds over time

High-stakes decisions build on work that happens over long periods of time. Projects keep sources, context, instructions, and outputs together as they move from one stage of research to the next. Curate one session into a template, and reference it in other sessions in the project. Start a session that summarizes all the work in the project, and evaluates the tradeoffs.

How customers use the Agent to make better decisions

Below are a few examples of how users are already using the Research Agent to make important decisions.

Indication prioritization and
asset strategy

A Commercial Development Leader at a $50B+ global biopharmaceutical company uses the Elicit Research Agent to build business cases and forecasting models that inform indication prioritization and asset strategy. To support these decisions, he researches patient journeys, identifies market segments, estimates disease prevalence and incidence, sizes markets, and analyzes competitive products. He finds Elicit consistent in answering recurring questions and significantly faster than manually searching across Google, ClinicalTrials.gov, and published literature.

Clinical research and
trial design

A Clinical Director at a top 5 global pharmaceutical company uses the Elicit Research Agent to support vaccine-research simulations and inform the choice of clinical outcomes and case definitions. To support these decisions, she gathers and extracts evidence on viral symptomatology, regulatory questions, internal decision making, and trial endpoints. For her, Elicit turns multi-week literature research tasks into a matter of days while avoiding repetitive searches across PubMed.

Analytical method development and
lab research

A Quality Control Analyst at a Swiss pharmaceutical company uses the Elicit Research Agent to develop and validate chromatography methods for drug stability, content, and purity testing with minimal trial and error. To support this work, he finds papers on specific analytical conditions, generates reports on method development approaches, studies degradation mechanisms, and compares method conditions. He finds Elicit’s ability to provide more precise scientific answers than general LLMs and offer more tailored follow-up questions particularly useful.

Scientific strategy and
evidence development

A Managing Director at a strategy advisory firm uses the Elicit Research Agent to help clients respond to competitor claims, develop scientifically defensible positions, and make more informed market-entry decisions. To support these decisions, he builds evidence bases for client recommendations, investigates unfamiliar topics, and evaluates options for complex competitive and market-entry questions. He relies on Elicit to trace every claim back to the original source and consistently produce more reliable results than other AI tools he has used.

Patent research and
scientific prior art

A Founder and Patent Attorney at an intellectual property firm uses the Elicit Research Agent to expand the scientific foundation of patent applications and inform broader patent strategy. To support this work, he builds evidence tables and reports that establish the scientific state of the art, help him understand unfamiliar technologies, and provide cited research for patent drafting. He prefers Elicit’s human-in-the-loop review process and structured outputs, which give him greater confidence in the underlying evidence.

Medical affairs and
clinical strategy

An Executive Director in Global Medical Affairs at a $250B+ global biopharmaceutical company uses the Elicit Research Agent to shape clinical strategy and guide decisions in new therapeutic areas. To support these decisions, he explores clinical trial designs, validates early findings, identifies evidence for deeper review, and prepares deliverables for discussions with external experts. He values Elicit’s ability to extract evidence across publications, congress presentations, and industry communications with minimal prompting.

See it in action

Elicit Research Agent can support full research workflows, from exploring a new scientific area to producing full decision-support materials. Here are some examples to get you started.

Explore complex landscapes

Extract and analyze evidence

Turn research into decision-support deliverables

Try it with your own research at elicit.com or connect with our team to see how Research Agent can support your organization’s workflows. To learn more visit: https://elicit.com/solutions/research-agent.