Flatiron’s AI-Driven Telescope Makes Data Access More User-Friendly

For years now, data from Flatiron Health have been hard to miss at the annual meeting of the American Society of Clinical Oncology (ASCO). Launched 15 years ago to organize data for community cancer centers and provide them with analytical tools to improve research and treatment, Flatiron has become a powerhouse for real-world data research, fueled by its user base of 4700 providers across 1600 sites of care.¹

More than 4 dozen studies at ASCO 2026 relied on Flatiron data,² with countless more mentions through individual slides or talks. But the big news was in the exhibit hall, where Flatiron leaders were rolling out Telescope, the artificial intelligence (AI)-powered analytics platform that the company says represents its most direct move yet to put its data in the hands of researchers without forcing them to go through analyst teams or write lines of code.¹

The platform is built on Flatiron's longitudinal database, which is now approaching 6 million patient journeys, said Conal Scanlon, senior director for product management, who oversees Telescope’s development. Scanlon offered a demonstration of Telescope for The American Journal of Managed Care® during ASCO.

Telescope, he said, is built with 2 core principles in mind: The first is to take those millions of patient journeys “and make them more actionable” for users. The second is “to build the 15 years of clinical and scientific expertise that we have analyzing our data into the process.”

Instead of asking users to write lengthy instructions to generate the right data, Telescope aims to put the complex part of answering queries “under the hood,” so that user questions can be sparingly simple. Yet, because Flatiron wants users to trust the process, Telescope lets the user follow the steps of building an answer in a fully transparent way—the AI version of when your high school math teacher made you show your work.

The central design principle behind Telescope is natural language. As Scanlon demonstrated, users can type plain-English queries, and the platform interprets the request by asking questions drawn from the same scientific best practices that Flatiron's own research teams apply.

When Scanlon typed, “TNBC patients“ for patients with triple negative breast cancer, he explained how this triggered multiple steps from Telescope: “If it has enough information to be able to find the patients in our data set, then it will go and do that. If it doesn't, then it's going to come back and ask me a set of clarifying questions,” such as, “How do you want to define HER2 status? Do you want to use IHC and IHS testing?” for immunohistochemistry.

For next-generation sequencing (NGS), Telescope will ask, “Should NGS results be incorporated to reduce missingness? Should the system use the most recent test or any negative result?”

Once those parameters are confirmed, the platform translates them into structured query language, or SQL, and then executes the request, pulling the filtered patient cohort from the underlying database.

“The platform exposes its reasoning at each step, showing both a plain-language summary of the query logic and the raw SQL, so users with varying levels of technical background can verify what the system actually did,” Scanlon explained.

Telescope then generates baseline characteristics tables, treatment patterns, and survival analyses—the same outputs that anchor a formal observational study.

Painstaking Process to Improve Accuracy

Scanlon was candid about the technical challenge the team faced in building the product. Out-of-the-box large language models produced accurate queries only about 50% to 60% of the time when first applied to a problem, a failure rate he described bluntly as unacceptable.

“So, we went through several rounds of iterations to improve the accuracy by adding in custom reasoning, and then by injecting our clinical and scientific best practices, like those questions that the model is asking back. And we were able to get up to over 90% accuracy,” he said.

That pain is all gain now, because for end users without analytics expertise, Telescope opens new doors.

"Things that used to take a couple of days and have a few handoffs involved—going to the technical team and then waiting for someone to pick it up—they're able to do those questions now in under an hour themselves," Scanlon said.

The intended users are primarily life sciences partners and academic researchers, with potential expansion into commercial teams and early clinical development groups that historically lacked the analytic resources to work directly with real-world databases.¹ Regulatory applications, including external control arms for submissions to the FDA or international bodies such as the National Institute for Health and Care Excellence in the United Kingdom, represent a downstream aspiration. However, Scanlon noted that pharmaceutical sponsors will likely need time to grow comfortable submitting AI-generated analyses as part of formal dossiers.

At present, Flatiron’s database captures records from approximately 2 in 5 patients with cancer in the United States and has expanded globally to include the UK, Germany, and Japan.¹

The company sees Telescope as a step toward the goal of democratizing access to its data, which will increase both the volume and speed of decisions its partners can make. For all the studies at ASCO with Flatiron data, there could be more, Scanlon said.

“We really expect that by using a tool like this, you'll be able to make better usage of our data, derive more insights from it, and ultimately have more of those publications,” he said.

References

1. Flatiron Health launches Flatiron Telescope, a new AI platform delivering oncology insights in minutes. News release. BusinessWire. May 19, 2026. Accessed June 20, 2026. https://www.businesswire.com/news/home/20260519371207/en/Flatiron-Health-Launches-Flatiron-Telescope-a-New-AI-Platform-Delivering-Oncology-Insights-in-Minutes

2. Flatiron Health brings 49+ research acceptances and next-generation capabilities to ASCO 2026. News release. Flatiron Health. May 21, 2026. Accessed June 20, 2026. https://resources.flatiron.com/press/flatiron-health-brings-25-research-acceptances-and-next-generation-capabilities-to-asco-2026