By Deborah Borfitz 

November 5, 2026 | Flatiron Health is building a global oncology intelligence engine that is poised to change how research gets done by scaling data collection and speeding up how researchers extract evidence, according to Kate Estep, the company’s chief product officer. The platform, known as Flatiron Telescope, is designed to enable anyone working in oncology to answer questions faster at any stage of drug development and, ultimately, be more of a “collaborative scientist” that can help expand the type of questions that get asked.   

Life sciences teams could thereby “explore more possibilities, challenge assumptions earlier, and generate better evidence to inform decisions ... that lead to better outcomes for patients,” says Estep. Flatiron Telescope is a platform powered by artificial intelligence (AI) that makes it easier to use complex, real-world oncology data without needing any specialized technical or coding skills, but in the end the beneficiaries are the tens of millions of people worldwide navigating a cancer diagnosis.      

The platform launched in May, bringing together a large volume of representative and timely real-world data (RWD), deep clinical and scientific expertise codified into both the data and product, and natural language analytics, she says. The RWD foundation currently includes over 6 million research-ready patients across the U.S., Germany, Japan, and the U.K., and the goal is to grow that figure to north of 10 million.  

In addition to encoding human knowledge into how data gets curated and analyzed, Flatiron experts are also in the loop wherever they’re needed, Estep says. “Over the last 15 years, Flatiron has been building and validating policies and procedures that help us turn raw clinical context into research-grade variables ... not only how do you turn clinical notes into structured data but also what are all the other edge cases and nuances and missing pieces of clinical data that need to be thought through to have a true representative picture of care.”  

Flatiron Health’s peer-reviewed VALID Framework ensures that data extracted by its large language models match the quality of expert human abstraction (Clinical Cancer Informatics, DOI: 10.1200/CCI-25-00215). “Generic text-to-code tools often fail in oncology because of the clinical nuance,” Estep says, citing as an example a question about how many HER2-positive metastatic breast cancer patients are in a dataset where Telescope was found to deliver a 20 percentage point increase in accuracy.  

The multi-agent architecture of Flatiron Telescope actively prompts users with clarifying questions when it detects missing or ambiguous clinical criteria, she explains. Among the follow-up questions related to biomarker positivity would be whether next-generation sequencing results should be incorporated to reduce missingness, and if the system should be looking for any positive, all positives, or the most recent positive if there are multiple results—exactly what Flatiron analysts would do when clarifying requirements for stakeholders.  

Telescope also has off-ramps where a human on Flatiron’s clinical, scientific, or statistical team takes over to handle “highly specialized bespoke questions,” Estep notes. “Being able to ask Telescope questions in natural language makes the analysis and the data more accessible, but it doesn’t eliminate the need to ask a well-formed question or to understand what you want to get out of the data or to design an appropriate methodological review,” she adds.  

The platform prompts users to structure clear, unambiguous, and context-rich questions as well as ensure the data is fit for purpose. It was built to extract the exact clinical intent behind complex research criteria, Estep says. 

Real-World Use Cases 

The value of Flatiron Telescope lies in the pairing of unprecedented speed to answers (minutes versus weeks or months) with deep, research-grade clinical rigor (data quality), says Estep. It essentially “democratizes” access to Flatiron Health’s oncology-specific RWD platform, which is widely regarded as one of the most rigorous and comprehensive datasets of its type and, importantly, contemporaneous with the standard of care.  

Among the use cases for Telescope is the ability to translate study inclusion and exclusion criteria into an analyzable cohort and, as criteria are refined, see the effect on patient counts and attrition in real time and “essentially derisk potential identification strategies,” she says. “It helps teams visualize those tradeoffs so they can figure out where they might need to adjust without compromising the question or decision at hand.” 

Estep is particularly excited about Telescope’s potential in enabling researchers to ask and study questions around the natural history of disease, biomarker testing, treatment patterns, outcome analyses, and unmet clinical needs—all “extremely critical for clinical development teams to understand as the real world continues to evolve,” she says. “The standard of care in oncology is shifting incredibly fast, and that affects everything from feasibility to assumptions around selecting the right patient populations.”   

The reality of late is that some cancer patients have multiple standard-of-care options available to them, or the standard of care has changed during a clinical trial, Estep continues. Development teams “need a faster way to continuously quantify what is happening in the real world in a trial-like cohort,” a highly curated group of patients who mimic the characteristics, treatment patterns, and outcomes of participants in a formal clinical trial.  

In this way, “as the trial is evolving, they can continually pressure-test design assumptions, look at control arm performance, target cohort size adjustments, and ultimately be able to improve the probability of technical and regulatory success,” says Estep. Telescope allows teams to rapidly and dynamically “answer more questions and make more informed decisions throughout the course of the trial.” 

How the Food and Drug Administration (FDA) responds to a standard of care that changes in the middle of an ongoing clinical trial depends on whether study results remain generalizable and clinically meaningful to contemporary medical practice and don’t automatically shut down the trial, Estep adds. But the FDA is likely to put the onus on sponsors to assess how shifts in treatment standards may affect the interpretation of trial data.  

Over the longer term, Estep says, she imagines Flatiron Telescope becoming a “continuous intelligence engine for scientists.” This is where she foresees a fundamental change in how research gets done. 

Scaling the Network  

The RWD foundation of Flatiron Health is primarily derived from electronic health records (EHRs) of patients, and it is experiencing an impressive growth spurt powered by geographic expansion and AI. For the immediate future anyway, the focus will be on enlarging the network in the U.S., Germany, Japan, and the U.K., says Estep. Most recently, Flatiron Health announced its first partnership in Scotland with NHS Greater Glasgow and Clyde, the country’s largest health board. 

Scaling of the global network has enabled Flatiron to curate data from those site partners into disease-specific longitudinal datasets reflective of real-world patients, treatment practices, and therapeutic outcomes, she says. To that end, the company has announced a series of major expansions to its AI-powered panoramic real-world oncology datasets for breast cancer and non-small cell lung cancer (NSCLC).  

This year, these have included a data model covering more than 17 biomarkers across 345,000-plus NSCLC patients and the launch of multinational breast cancer and NSCLC datasets across the U.K., Germany, and Japan that follow an unveiling in prostate cancer in January. Flatiron’s global breast cancer panoramic dataset also surpassed 1 million patients, making it the largest curated dataset of its kind inclusive of hormone receptor-positive, HER2-positive, and HER2-low status patients, among whom more than 81,000 are tracked from early to metastatic settings.  

By layering Telescope atop these interoperable, disease-specific datasets, study teams can now interrogate the information and learn from patients in those geographies “at the pace that care is evolving,” says Estep. The datasets will continue to expand to provide the depth and the data needed to answer the questions users are trying to answer.   

Flatiron additionally offers real-world evidence insights, analytics in support of treatment patterns, natural history, and comparative effectiveness studies. Life sciences teams and researchers can also opt to study the deidentified, EHR-sourced longitudinal clinical datasets on their own. 

RWE and RWD are “two sides of the same coin,” points out Estep. “We want to answer every single question that someone may ask about oncology and to do that we need the most rigorous and deep and representative real-world data foundation ... [as well as] the ability to generate evidence and insights continuously and quickly from that in partnership with researchers to answer these pressing questions.”  

Broad Utility 

Flatiron Telescope has broad utility across pharma companies, since it was built to answer questions across the oncology development and commercialization life cycle, Estep says. Beyond clinical development use cases, health economics and outcomes research teams are using the platform “to evaluate whether a dataset is fit for purpose, build cohorts, examine real-world treatment patterns, and run outcome studies.”  

Medical affairs teams, meanwhile, are using Telescope “to better understand evidence gaps, [evaluate] biomarker testing and trends, [and understand] variation in clinical practice across geographies,” says Estep. With commercial teams the goals of users include “exploring therapeutic utilization, understanding real-world treatment adoption and market dynamics across geographies, [and] further interrogating qualitative insights, like the rationale behind certain prescribing decisions.” 

In addition to its life sciences partners, Flatiron Health also works in the same way with the National Comprehensive Cancer Network and several large academic partners to conduct research. Telescope has a role at these healthcare institutions, although the primary research goals and operational use cases tend to be different than those at pharma companies.  

The focus areas, Estep says, include real-world effectiveness of treatments and patterns of care, the patient experience, and clinical trial matching. The platform is sometimes more useful in academia because of the no-code interface that allows researchers to start on certain projects without having to set up localized data infrastructure, write complex code, or build analytics tools from scratch, she adds.  

For life science partners that have licensed Flatiron RWD, Telescope is now available to use on top of those datasets, Estep reports. “We’re also seeing Telescope reduce barriers to engagement for teams and companies that haven’t worked with Flatiron historically or licensed real-world data ... maybe [because] they had a quick question, but it wasn’t worth licensing a dataset for. We’re seeing a bunch of those types of projects surface now.” 

Flatiron is also now working with a handful of large life science partners to embed Telescope’s agents within their own data and analysis ecosystems and workflows, says Estep. “Every biopharma customer is also thinking about how they’re going to unify their data, analysis, and model embeddings ... they don’t want people hopping across six different front ends. They want an integrated view, and so we’ve been talking about how to work [together] to make Telescope’s underlying context available within the AI operating systems biopharma teams are already building.”