Corti's emergency-call AI bet: from Copenhagen's 112 lines to 100M patients a year

An AI copilot for healthcare and public safety: it listens to patient interactions in real time, triages, documents, assigns billing codes and flags what clinicians or emergency call handlers may have missed.

An AI trained on real emergency calls can hear what human dispatchers miss — and listening to the encounter, not replacing the professional, is the product.

$27M Series A (2021); $60M Series B co-led by Prosus Ventures and Atomico (September 2023) with Eurazeo, EIFO and Chr. Augustinus Fabrikker — roughly $87M total

Andreas Cleve and Lars Maaløe (with Michael Boesen, who later left) founded Corti in Copenhagen in 2016. The first product listened to the city's official 112 calls, trained on tens of thousands of archived recordings, to recognize the words and background sounds — like irregular breathing — that accompany cardiac arrest. In the Copenhagen trial, dispatchers recognized cardiac arrests about 73% of the time; Corti's AI spotted them 95% of the time. In 2018 the European Emergency Number Association took it to trials in four more European cities.

The company then widened the wedge: the same listening engine became a general copilot for clinical encounters — triage, real-time documentation, billing codes, post-visit quality feedback, even a second opinion that could contradict the clinician. Maaløe insisted on building its own models and deliberately hired no in-house physicians, arguing a staff clinician would bias the system; outside researchers were invited to stress-test it instead. Early meetings met pushback about job replacement, and The Verge's 2018 review pressed on real weaknesses: the system could not explain its decisions, and the full study behind the headline numbers — 161,650 calls — had not been published, leaving false positives unquantified. ChatGPT changed the temperature: by 2023, Maaløe said, meetings had become radically easier, and the goal was to "make the AI term boring." Usage grew from 15M consultations a year in 2021 to 100M patients a year, 150,000 uses a day, by September 2023, when Prosus Ventures and Atomico co-led a $60M Series B.

Training data was the moat: Copenhagen's archived 112 calls gave the model an acoustic signature of cardiac arrest that no general-purpose system had.

The head-to-head numbers — 95% detection against a human 73% — were a demonstration no sales deck could match, and EENA turned them into a four-city European trial.

Refusing to hire in-house physicians was a stated design principle against bias, and it differentiated Corti from GPT-wrapper competitors when they arrived.

Documentation and coding monetize everyday clinical work rather than rare emergencies, which is how usage scaled from 15M consultations a year in 2021 to 100M patients a year in 2023.

Specialized beats general in high-stakes domains: Corti trained on one city's emergency-call archive and beat humans at one recognition task — then widened that wedge into everyday clinical work.

As of the September 20, 2023 Series B announcement, Corti remains a private Copenhagen company scaling across Europe and the US, with emergency-dispatch and hospital customers, ~150,000 daily uses and a stated aim of making AI assistance in healthcare unremarkable. The company had not disclosed its valuation in this round.