AI in healthcare and life sciences 2026: from pilots to everyday practice

Executive summary

AI has moved from pilots into the daily work of clinicians and researchers, faster than many organisations have put governance around it. The evidence of benefit is strongest for narrow, regulated uses such as imaging and for reducing documentation time; it is thinner for outcomes such as burnout and late-stage drug development. This report sets out what the latest evidence shows and what regulators now expect.

Clinicians have adopted AI faster than their organisations have governed it. 81% of US physicians surveyed by the American Medical Association in early 2026 used AI in their practice, up from 38% in 2023, with clinical documentation and summarising research the most common uses [1].
Regulated AI is concentrated in imaging. The FDA’s list of AI-enabled medical devices passed 1,600 authorisations by September 2026, most of them in radiology [2]. The FDA notes that the list is not comprehensive [2].
Ambient documentation is the first large-scale clinical use of generative AI, and the evidence is mixed. In a study across six health systems, burnout among 263 clinicians fell from 51.9% to 38.8% after 30 days of using an ambient scribe, though the study had no control group [3]. A randomised trial of two ambient scribes among 238 outpatient physicians reported benefits for documentation time and well-being, alongside occasional clinically significant inaccuracies in notes [4].
In drug discovery, AI speeds early stages but has not yet changed late-stage success. AI-discovered molecules passed Phase I at 80% to 90%, against historical averages of roughly 40% to 65%, while Phase II success, at about 40%, is in line with industry norms [5]. One of the most advanced examples, rentosertib, reported phase 2a results in 2025 [6].
Regulators are setting expectations. In January 2025 the FDA published draft guidance on a risk-based approach to establishing the credibility of AI models used to support regulatory decisions on drugs and biological products [7]. In the EU, AI that is a medical device or a safety component of one is high-risk under the AI Act [8], with those obligations now applying from 2 August 2028 [9].

Key findings

0181% of US physicians used AI in their practice in 2026, up from 38% in 2023

02The FDA lists more than 1,600 AI-enabled medical devices authorised for the US market

03Burnout fell from 51.9% to 38.8% after 30 days with an ambient AI scribe in one six-system study, which had no control group

US physicians using AI in their practice

2023

38%

2026

81%

Source: American Medical Association, 2026

Clinician burnout before and after 30 days with an ambient AI scribe

Before

51.9%

After 30 days

38.8%

Source: JAMA Network Open, 2025 (no control group)

What leaders should do

Put clinician-led governance around the AI tools already in use, starting with documentation and decision support.
Measure ambient documentation against agreed outcomes, including note accuracy, not only time saved.
Map each AI-enabled device and AI component in regulated products to its FDA and EU AI Act obligations before 2028.

What this report covers

Clinician adoption
Regulated AI devices
Ambient documentation: time saved and open questions
AI in drug discovery and development
Regulation in the US and EU
What to do in 2027

Methodology

We reviewed public sources published between 2024 and October 2026, favouring material from 2025 and 2026. These include professional surveys (American Medical Association), regulatory publications (FDA and the European Union), and peer-reviewed or preprint studies of ambient documentation and AI-discovered drugs. Several ambient scribe studies are short or lack control groups, and we say so where it matters. Each figure is cited to a dated source. Where we relied on a secondary report of a primary document, the source list says so. Surveys are self-reported and should be read as indicators, not audited results. We did not survey organisations directly.