October 23, 2026

AI & Healthcare: The Future of Healthcare

Every few generations, medicine changes so much that a doctor from before and a doctor from after would struggle to recognise each other's work. Anaesthesia did it. Antibiotics did it. Imaging did it. Artificial intelligence is now doing it again, and the change is no longer a promise in a conference talk. It is showing up in hospitals, in screening programmes and in the consultation room.

The latest sign came from Sweden. A trial of more than 105,000 women, published in The Lancet in January 2026, found that breast screening with AI support was as safe as having two radiologists read every image. Cancers that surfaced between screening rounds were no more common with AI (1.55 per 1,000 women against 1.76 with standard double reading), more cancers were picked up at screening (sensitivity 80.5% against 73.8%), and false alarms did not rise (specificity 98.5% in both groups) (Gommers et al., The Lancet, 202602464-X)).

To see why this matters, it helps to look back first: at the handful of breakthroughs that built modern medicine, then at what AI is doing today, and finally at what the people building it expect next.

Medicine has been reinvented before

Healthcare as we know it is the product of a few dozen decisive breakthroughs. Most of them began in Europe.

Around the 5th century BCE: clinical observation. The writings attributed to Hippocrates of Kos moved medicine from ritual towards observation, classification and cause and effect (World History Encyclopedia).
1543: anatomy. Andreas Vesalius's De humani corporis fabrica, printed in Basel and built on his teaching at Padua, challenged Galen with the dissected human body (Caius College, Cambridge).
1628: circulation. William Harvey's treatise on the motion of the heart and blood overturned some 1,500 years of accepted belief (Life of Harvey).
1796: vaccination. Edward Jenner's smallpox vaccine began the road to the only human disease eradicated so far, certified by the World Health Assembly in 1980 (WHO).
1846: anaesthesia. The first public demonstration of ether at Massachusetts General Hospital, on 16 October, made modern surgery bearable (Smithsonian).
1860s to 1880s: germ theory and antisepsis. Louis Pasteur's work on fermentation inspired Joseph Lister's antiseptic surgery, published in The Lancet in 1867 (Lister, 1867). Robert Koch announced the tuberculosis bacillus in 1882 (Nobel Prize).
1895: X-rays. Wilhelm Röntgen's discovery, in Würzburg, gave medicine its first look inside the living body (Nobel Prize).
1928 to 1945: antibiotics. Alexander Fleming observed penicillin in 1928; Florey and Chain turned it into a medicine, and all three shared the 1945 Nobel Prize (Nobel Prize).
1953: the structure of DNA. Watson and Crick's double helix, published in Nature in April 1953, opened molecular medicine (Nobel Prize).
1954: organ transplantation. Joseph Murray's team in Boston transplanted a kidney between identical twins (Nobel Prize).
1970s: CT and MRI. Godfrey Hounsfield's scanner in London and the later work of Lauterbur and Mansfield let doctors see inside the body without surgery (Nobel Prize, 1979).
1992: evidence-based medicine. An article in JAMA by the Evidence-Based Medicine Working Group gave a movement its name and put clinical research ahead of intuition (JAMA, 1992).
1990 to 2003: the Human Genome Project. Launched in October 1990, declared complete in April 2003 (NHGRI).
2020 to 2023: mRNA vaccines. The UK authorised the BioNTech and Pfizer vaccine in December 2020 (GOV.UK); Katalin Karikó and Drew Weissman received the 2023 Nobel Prize for the discovery that made it possible (Nobel Prize).

Each of these changed what a doctor could see, prevent or treat. AI is the first to change something else as well: how much of a doctor's day goes to paperwork.

How AI is changing healthcare today

Biology: AlphaFold and a Nobel Prize

In October 2024, the Nobel Prize in Chemistry went half to David Baker for computational protein design and half to Demis Hassabis and John Jumper of Google DeepMind for AlphaFold, the AI system that predicts protein structures (Royal Society; Google DeepMind). Knowing a protein's shape is the first step in designing a drug that fits it.

Regulated devices: more than 1,500 and counting

By the end of March 2026, the US Food and Drug Administration's public list of AI-enabled medical devices held 1,524 authorisations, roughly three in four of them in radiology (IntuitionLabs, from the FDA list). The FDA itself says the list is not comprehensive.

Screening: Sweden and the NHS

The Swedish trial above is the second result from the same programme. Its earlier analysis found that AI-supported screening detected 6.4 cancers per 1,000 women against 5.0 with standard reading, and cut radiologists' screen-reading workload by about 44% (Hernström et al., The Lancet Digital Health, 2025).

In the United Kingdom, the EDITH trial, backed by £11 million from the National Institute for Health and Care Research, is testing AI in breast screening for nearly 700,000 women across 30 NHS sites (GOV.UK).

Documentation: the burden AI scribes target

The paperwork problem was measured before AI arrived. A time-and-motion study of 57 US physicians in four specialties found that for every hour of direct patient contact, nearly two further hours went to the electronic health record and desk work, with doctors reporting another one to two hours of EHR work after clinic hours (Sinsky et al., Annals of Internal Medicine, 2016).

Ambient AI scribes, which listen to the consultation and write the note, are now being tested at scale:

An NHS England-sponsored evaluation led by Great Ormond Street Hospital followed almost 17,000 patient interactions at nine London sites, from hospitals to GP surgeries and mental health services. Clinicians spent 23.5% more of the appointment interacting directly with patients (report coverage).
In July 2026, NHS England backed the rollout of integrated ambient voice technology as part of a £10 billion technology investment (Digital Health).

The brain: Neuralink and decoding thought

At the edge of medical AI sits the brain-computer interface. Neuralink, Elon Musk's company, implanted its first human participant, Noland Arbaugh, in January 2024. Arbaugh, who is paralysed, went on to control a computer cursor by thought (Scientific American).

The implant carries 1,024 electrodes on 64 threads. A software decoder, calibrated to each user's own neural patterns, translates the recorded motor-cortex signals into cursor movements (Fierce Biotech). Some threads retracted after the first surgery, and the company responded by changing its recording and decoding algorithms, so the decoder is what made up for the lost signal.

Studies have since opened outside the United States. University College London Hospitals reported seven patients taking part in the UK study by 29 January 2026 (UCLH), and trials have also been registered in Canada and the United Arab Emirates (trial tracker). In May 2025 the company said the FDA had granted Breakthrough Device designation to its speech-restoration device, a status that speeds review and is not an approval (Reuters, via Tuoi Tre). Most results so far come from the company's own updates, not peer-reviewed trials.

The rules: Europe is writing them

In Europe, the question is as much regulation as technology. A WHO Europe survey of 50 of the Region's 53 countries found 32 (64%) already using AI-assisted diagnostics, yet only four had a health-specific national AI strategy (WHO Europe, 2025). WHO Europe's director, Dr Hans Kluge, put it this way: "AI is already a reality for millions of health workers and patients across the European Region" (WHO Europe).

The EU AI Act's high-risk rules were due to apply in August 2026. The Digital Omnibus on AI, Regulation (EU) 2026/1744, published in the Official Journal on 24 July 2026, moved them to 2 December 2027 for stand-alone high-risk systems and 2 August 2028 for AI built into regulated products such as medical devices. Transparency obligations were not deferred (National Law Review; Pearl Cohen).

What comes next, in the words of those building it

Predictions are cheap. Here is what some of the people and institutions involved have actually said and done.

**Eric Topol, cardiologist and author of Deep Medicine,** argues that the greatest opportunity of AI in medicine is not fewer errors or lighter workloads but restoring the human connection between patients and doctors, with AI taking over clerical work to give clinicians time (Scripps Research).

Jeremy Farrar, then WHO's Chief Scientist, said when WHO issued its January 2024 guidance on large multimodal models that generative AI can improve healthcare only if those who develop, regulate and use it account for the risks (TechTarget).

Demis Hassabis has made Isomorphic Labs's goal public: a 2026 interview was headlined his rush to "solve all disease" (Yahoo Finance). Isomorphic, which has sites in London and Lausanne, raised a $2.1 billion round in May 2026 (Reuters, via WHTC).

Projects under way

NHS England's £10 billion technology programme puts ambient voice technology and AI-based triage at the centre (Digital Health).
The European Health Data Space (Regulation (EU) 2025/327) came into force in March 2025 and phases in over several years; its rules on reusing health data for research apply from 26 March 2029 (A&O Shearman; Kennedys).
Google's AMIE, a conversational diagnostic AI. A Lancet feasibility study at Beth Israel Deaconess Medical Center, published on 8 October 2026, had 98 patients complete a pre-visit conversation with it under physician supervision; no conversation needed to be stopped for safety. It was a single-centre, single-arm study, and the authors call for larger trials (Google Research; ARISE).
UK Biobank, whose proteomics project measures proteins in blood samples from its 500,000 volunteers, funded by 14 pharmaceutical companies (UK Biobank).

What this means for a doctor's day

Most of what is described above will reach patients slowly, through trials, regulators and hospital budgets. One part is already here for any doctor who wants it: the note. Nearly two hours of record-keeping for every hour with a patient is the burden that wears clinicians down most, and it is the one AI can lift today. When the note writes itself, the doctor looks at the patient instead of the screen, and nothing said in the room is lost because nobody had time to type it.

That is the job Cognivolt was built for.

Cognivolt is an AI medical scribe with clinical decision support and an EMR built in. You can record the consultation live, dictate, upload audio, type notes, or photograph rough notes and get them back as a clean, structured note. It works in most languages and for all specialties, from psychiatry and psychology to cardiology, dermatology and general practice.

There is nothing to set up or learn first: press record and the note is ready when you stop. The patient record is on every plan from $39 a month, so there is no separate EMR to buy, and decision support is on the Standard plan at $89. If your clinic already has an EHR, the Super Extension puts the approved note into it in one click, on every plan.

Doctors in Europe work in several languages, under data rules that are tightening, in systems that often do not talk to each other. A scribe that writes the note and keeps the record it belongs to removes one of those problems. There is a 14-day free trial with no card needed: start here.

Frequently asked questions

How is AI used in healthcare today?
AI reads medical images, supports breast cancer screening, predicts protein structures for drug discovery, and writes clinical notes from consultations. More than 1,500 AI-enabled devices are on the FDA's list, about three in four in radiology, and a Swedish randomised trial of more than 105,000 women found AI-supported mammography screening as safe as double reading by two radiologists.

What is the future of AI in healthcare?
The direction set by current projects is less paperwork, earlier detection and faster drug discovery. The NHS is backing ambient voice technology and AI triage, Isomorphic Labs, with sites in London and Lausanne, is raising billions to design drugs with AI, and the European Health Data Space will open health data for research from 2029. Experts such as Eric Topol argue the real prize is giving doctors back time with patients.

What does the EU AI Act mean for medical AI?
The high-risk rules were due in August 2026 but were postponed by the Digital Omnibus on AI, which was published on 24 July 2026. They now apply from 2 December 2027 for stand-alone systems and from 2 August 2028 for AI built into regulated products such as medical devices. Transparency obligations were not delayed.

Can AI replace doctors?
Not on the evidence so far. The studies above keep a clinician in the loop: radiologists read the images, supervising physicians watched AMIE's conversations, and doctors review every note an AI scribe drafts. WHO Europe's message is that people and patients must stay at the centre of AI decisions.

How much time do AI scribes save doctors?
With an accurate AI scribe, the writing itself almost disappears: the note is drafted from the consultation, and what is left is a quick check. With Cognivolt, the time spent writing notes drops by 90 to 100%, depending on how much you choose to edit. The other gains matter as much: your attention stays on the patient (an NHS evaluation of almost 17,000 interactions found clinicians spent 23.5% more of the appointment interacting directly with patients), and fewer details are missed, because the whole conversation is captured rather than what you managed to jot down.

What is the best AI scribe for doctors in Europe?
Cognivolt is an AI scribe with built-in clinical decision support and an EMR. You can record, dictate, upload audio, type notes or photograph rough notes. It works in most languages and for all specialties, and the Super Extension puts the note into the EHR you already use, on every plan.

What is the best AI scribe for a solo doctor?
Cognivolt. Press record and get an accurate note from the first session, with nothing to set up first. The patient record is already there at the same price, from $39 a month, so there is no separate EMR to buy.