AI medical scribe Heidi raises $476M to expand beyond transcription
Australian startup aims to tackle clinician shortages with AI agents handling coding, prescriptions, and follow-ups
1Rapid GP adoption
40% of Australian GPs now use AI transcription tools like Heidi, up from 22% in early 2024. The Melbourne-built platform currently handles 2.8 million weekly patient visits globally, though none of these tools are TGA-approved. Heidi operates in regulatory gray areas by limiting functionality to documentation - transcribing consultations without influencing clinical decisions. This exemption allows rapid scaling while avoiding medical device classification.
2Unicorn funding round
Heidi's $476M Series B led by Blackbird Ventures values the company at $1.26B, making it Australia's newest tech unicorn. General Catalyst provided an additional $240M revenue-based loan. The capital will fund expansion into clinical tasks like medication charts and workers' compensation certificates. Blackbird reports Heidi as its fastest-growing portfolio company, outpacing even Canva's early trajectory with ARR surging from $1M to $50M in two years.
3Regulatory tightrope
Heidi plans to soon handle clinical coding, follow-ups, and drug calculations - functions that may trigger TGA oversight as medical devices. The company avoids EU/UK markets due to stricter regulations. Australia's health regulator completed an AI scribe review noting 'scope creep' and decision-influencing AI, with potential compliance actions coming. 19 practitioner complaints have been filed regarding privacy breaches and consent issues with AI scribes.
4Clinician trust barriers
While ER physician Tatiana Lowe delegates documentation to Heidi, she draws the line at complex diagnoses or end-of-life conversations. Research shows only 50% of patients accept AI scribes, with under 10% comfortable with AI-driven treatment decisions. Errors persist - like recording a Keflex allergy as 'allergic to Catholics' - though they've become rarer with system improvements. Unchecked automation risks embedding mistakes in permanent records.
5Product roadmap
Within a month, Heidi will detail how AI agents will operate under clinician supervision. The system currently requires doctor approval for each automated task. Future capabilities include repeat prescriptions and dose calculations - features CEO Thomas Kelly acknowledges may require medical device classification. The company maintains advance consultation with regulators about planned feature development.
6Regulatory landscape
Australia's TGA maintains its existing framework can accommodate AI healthcare tools, assessing them by intended purpose and risk level. Digital Rights Watch criticizes Australia's weaker protections compared to EU/UK. The Health Practitioner Regulation Agency reports 19 AI-related complaints, mostly about consent and privacy. Regulators are actively examining the line between administrative tools and clinical decision influencers.
7Accuracy challenges
Despite quality improvements, errors still occur in about 1% of Heidi transcriptions according to internal data. The ABC reported a case where a scribe wrongly documented illicit drug use. Consumers Health Forum warns automation can reduce vigilance - clinicians may proofread less as they grow accustomed to AI assistance. CEO Elizabeth Deveny argues some process friction serves as a quality control mechanism.
8Global implications
Heidi's expansion highlights tensions between healthcare innovation and patient safety. As AI assumes more clinical responsibilities, regulators worldwide grapple with classification challenges. The Australian experience demonstrates how rapidly AI can penetrate medicine when operating in regulatory gaps. With 190-country reach across 110 languages, Heidi's decisions will influence global standards for AI in healthcare.