A reporter's instinct, rebuilt as infrastructure
Every weekday morning, a system in New York wakes up and reads more than seven thousand updates to clinical trials around the world. It does not get bored at update four hundred. It does not skim. By the time most people have finished their coffee, it has flagged the sixty to eighty events that actually matter and written them up - a trial quietly slowing down, an enrollment target missed, a site going dark. The company is called AppliedXL. The person who decided a database deserved the same suspicion a good reporter brings to a press release is Francesco Marconi.
Marconi runs AppliedXL as co-founder and chief executive. The pitch sounds almost contradictory: take the patience and skepticism of investigative journalism, hand it to machine learning, and use the pair to catch consequential events before they are announced. AppliedXL has a word for that window - the gap between when something is true and when it becomes public. They call it pre-news.
It is a strange place for a journalist to end up, and that is rather the point. Marconi did not drift into artificial intelligence as a trend. He spent years inside two of the most important newsrooms in the world building the tools first, then left to build a company around the idea that the most valuable journalism might be the kind that happens before anyone files a story.
From raw signal to flagged event
The mechanics are less mysterious than the marketing. AppliedXL treats public data sources the way a beat reporter treats a beat - continuously, with memory, and with a nose for what has changed. Strip away the jargon and it runs like a newsroom that never sleeps.
The volume is the part that humbles you. A team of human analysts could not read 7,000 trial updates a day without missing the one that mattered. A pure algorithm could read them all and still miss it, because it would not know what mattered. Marconi's bet is that the interesting work lives in the seam between the two.
The scale of a quiet engine
That last bar is the whole business model. Roughly one percent of what comes in is worth telling someone about. Finding that one percent reliably, every single day, is the difference between noise and intelligence.
Two newsrooms, one obsession
Long before AppliedXL, Marconi was inside the institutions trying to figure out what AI meant for the people who write the news. At the Associated Press he co-led the strategy for content automation - early, unglamorous work teaching wire-service systems to produce stories from structured data. Corporate earnings reports were among the first beats handed to algorithms, freeing reporters to chase the stories that needed a human.
Then The Wall Street Journal made him its first ever R&D Chief, where he led a team of data scientists and computational journalists building tools for the newsroom. The title was new because the role was new. Nobody had quite drawn the org chart for a reporter who also ships software.
In 2020 he put it all in a book. Newsmakers: Artificial Intelligence and the Future of Journalism, published by Columbia University Press, walks through newsrooms using algorithms to write stories, reporters mining enormous public datasets, and outlets deciding by machine where their content should go. Its argument is calm and slightly contrarian for its moment: the future of news is not a faster writer, it is a machine that notices, with editors still firmly in charge. He has been teaching the same idea ever since, including a stint as an adjunct instructor at Columbia Journalism School.
The career, in stops
Economics first, journalism later
Here is the giveaway about how Marconi thinks. He did not start in a newsroom. He earned a bachelor's degree in economics in Portugal and a master's in economics in Italy before he ever filed a story. Only then did he cross the Atlantic for a master's in journalism and business at the University of Missouri, followed by postgraduate work at Columbia and Harvard.
An economist who became a reporter who became an engineer who became a CEO. It explains the through-line. To Marconi a clinical-trial database is not a spreadsheet and not a story - it is a market of signals, and somebody ought to be reading it like one. He is Italian-Portuguese by background, which means the man building a global signal-detection engine grew up bilingual and has been crossing borders, literal and disciplinary, his whole life.
When journalism becomes an edge
Point a pre-news engine at politics and you have a faster newswire. Point it at biotech and you have something investors will pay for. AppliedXL's first commercial beat was not the obvious one. It was clinical trials - the slow, technical, high-stakes machinery of drug development, where a single delayed trial can move a company's stock and a missed enrollment target can hint at trouble months before the press release.
That is why partners matter here. STAT, the respected health and medicine publication, brought editorial credibility. Bloomberg brought the Terminal, where the world's traders live. The same event-detection that started as a journalism experiment now sits inside the workflows of people managing biotech risk for a living. Consultancies have studied the approach as a way to quantify execution risk in biopharma pipelines. The journalist's question - what changed, and who needs to know - turns out to be worth money.
Along the way the field has noticed him. MediaShift listed him among its top twenty digital media innovators. Editor & Publisher named him to its 25-under-35 next generation of publishing leaders. He has been a Digiday Future Leader finalist, a Tow Fellow at Columbia, and an affiliate researcher at the MIT Media Lab's Laboratory for Social Machines - the rare operator who keeps one foot in the lab.
What he is really building
Strip AppliedXL down and the ambition is bigger than biotech. Marconi wants to prove that the discipline of journalism - verify, contextualize, decide what is worth saying - can be encoded and run at machine scale across any domain dense with signals. Clinical trials are the proving ground because the data is public, structured, and consequential. If it works there, the method travels.
What he refuses to give up is the editor in the loop. For all the talk of automation, his entire body of work argues the opposite: keep human judgment at the center, and let the machine do the reading no human could survive. It is an unfashionably patient idea in an industry that loves to promise full autonomy. It is also, so far, the one that has shipped.