Executive summary
In March 2026 the American Medical Association reported that 81 percent of physicians use AI in practice. That's more than double the 38 percent it recorded in 2023. The leading application isn't diagnosis. It's summarizing medical research and standards of care. In April 2026 a survey of 367 medical affairs professionals across 48 countries found 27 percent using AI tools for key opinion leader engagement. Roughly a third of their organizations operate under a written AI policy. Those two findings describe an asymmetry, not a problem. Clinicians have already built the habit of reading synthesized evidence. Medical affairs teams that build a governed, always-on read of the literature, congress coverage and trade press gain something specific. Their field teams walk into every KOL conversation holding the current state of the record. This brief covers the evidence, the January 2026 FDA and EMA principles, and four moves for the fourth quarter.
The evidenceYour audience adopted first
Start with the clinicians, because they moved first.
The AMA has fielded its Physician Survey on Augmented Intelligence every year since 2023. The 2026 wave, released in March, found 81 percent of physicians using AI professionally, against 38 percent three years earlier (AMA, 2026). The average physician now reports 2.3 distinct use cases, up from 1.1 (AMA, 2026).
Adoption on its own would be a modest finding. The composition is the useful part. The leading use case is summarizing medical research and standards of care, cited by 39 percent of respondents (AMA, 2026). That single use rose 33 percentage points from the 2023 survey. More than three-quarters now say AI improves their ability to care for patients, up from 65 percent in 2023 (AMA, 2026).
So the specialist you brief on Thursday has probably already read a machine-written précis of your therapeutic area. Not instead of the literature. Before it.
InterpretationThe gap is the opening
Medical affairs adoption hasn't moved at the same speed, and the numbers say so plainly.
A global cross-sectional survey published in Cureus in April 2026 polled 367 professionals across 48 countries (Kraemer, 2026). Respondents spanned medical science liaisons, MSL leadership and executive medical affairs leaders. Twenty-seven percent reported using AI tools for KOL engagement. Literature review led the applications at 22 percent, followed by data analysis at 20 percent and presentation preparation at 16 percent (Kraemer, 2026).
Among those already using it, the reported benefits were concrete rather than aspirational. Forty-two percent cited improved efficiency. Eighteen percent cited better preparation for KOL engagements (Kraemer, 2026).
Now read the two datasets together. Four in five of the people in the room have the habit. Roughly one in four of the people walking into the room have the tooling. That gap closes either way. The only real question is whether you close it deliberately, with governance attached, or by accident, one enterprising MSL at a time.
Regulatory contextThe governance vocabulary already exists
Sentiment inside the function supports the move. In the same survey, 74 percent expected AI to be impactful or highly impactful in medical affairs (Kraemer, 2026). Another 87 percent called learning about AI important or very important to staying competitive. About one-third of organizations had a written policy governing MSL use of AI (Kraemer, 2026).
Regulators handed you a starting frame in January. On January 14, 2026, the FDA and EMA jointly published the Guiding Principles of Good AI Practice in Drug Development (FDA and EMA, 2026). The ten high-level principles cover nonclinical, clinical, post-marketing and manufacturing use. They aren't binding guidance. They read as a governance checklist: human-centric design, risk-based validation, data governance and documentation, and lifecycle management.
Field medical isn't drug development, and nobody should pretend otherwise. But the compliance colleague reviewing your first MSL AI policy will reach for the nearest published regulatory vocabulary. That vocabulary now exists, and it's eight months old.
| Finding | Figure | Source and year | Method note |
|---|---|---|---|
| Physicians using AI professionally | 81% | AMA, March 2026 | Annual survey since 2023; 2023 baseline 38% |
| Top physician use: summarizing medical research | 39% | AMA, March 2026 | Up 33 percentage points from 2023 |
| Average AI use cases per physician | 2.3 | AMA, March 2026 | Up from 1.1 in 2023 |
| Physicians saying AI improves patient care | >75% | AMA, March 2026 | 65% in 2023 |
| Medical affairs using AI for KOL engagement | 27% | Cureus, April 2026 | n=367, 48 countries, non-probability convenience sample |
| Organizations with a written MSL AI policy | ~33% | Cureus, April 2026 | Same sample |
| Say AI literacy matters to stay competitive | 87% | Cureus, April 2026 | Same sample |
The Ringer anglePreparation is an input problem
Better KOL preparation isn't a drafting exercise. It's a question of what reaches the MSL before the meeting.
An MSL walking into a conversation needs to know what changed in the therapeutic area since the last one. Which abstracts landed. Which congress sessions drew argument rather than applause. Which trade outlets covered the data, and in what register. Which investigators are being quoted, and on what. All of that exists on the public record. It's scattered across journals, congress programs, trade press and social posts, and no one reads all of it on a Tuesday.
Media intelligence, always on, across the therapeutic area
Pulse reads earned, owned and social coverage continuously, then tells you what moved and why it matters. For medical affairs that means field teams arrive holding the current state of the record, not the version captured at the last quarterly review.
Human-led, AI-powered, with the provenance trail your medical and legal reviewers will ask for. See how Pulse works →
RecommendationFour moves for the fourth quarter
None of these need a platform decision this month.
- Write the policy before you scale the tools. Two-thirds of organizations in the Cureus sample had no written AI policy for MSLs. Borrow the structure from the January principles: risk tiering, validation, documentation, human review. A three-page policy beats a nine-month working group.
- Separate synthesis from generation. Summarizing published evidence for internal preparation and drafting external scientific content carry different risk. Name that line in the policy, and your reviewers will stop treating every use as the riskiest one.
- Instrument the therapeutic area, not the brand. KOLs discuss mechanism, comparators, guidelines and unmet need. A monitoring set scoped to your product name will miss most of the conversation your MSLs are about to walk into.
- Measure preparation, not activity. The Cureus respondents already using AI reported better KOL preparation as a benefit. Ask your field teams to score preparedness before and after. Six weeks of that data will tell you more than a vendor evaluation.
Physicians did the hard part. They built the habit, and they told a national survey they trust the results more than they did three years ago. Meeting a fluent audience with equal fluency is a smaller project than it sounds. It's the one your field teams can finish before the January congress season.
Methodology note
Physician figures come from the American Medical Association Center for Digital Health and AI, Physician Survey on Augmented Intelligence, released March 12, 2026. Nearly 1,700 physicians responded, spanning a range of specialties, practice settings and career stages. It is an annual instrument fielded since 2023; comparisons to 2023 are drawn from the AMA's own reporting of prior waves.
Medical affairs figures come from a global cross-sectional survey published in Cureus on April 16, 2026, with 367 participants from 48 countries across pharmaceutical, biotechnology, medical device, diagnostic and other healthcare organizations. The authors used non-probability convenience sampling distributed through professional networks, which limits generalizability and may over-represent AI-interested respondents.
The two instruments measure different populations over different periods and are not directly comparable. They are presented here as adjacent readings of the same market, not as a matched pair. No causal claim is made or implied, and nothing here should be read as guidance on any product, indication or clinical decision.
Sources
- American Medical Association, “AMA: AI usage among doctors doubles as confidence in technology grows,” March 12, 2026 · ama-assn.org
- American Medical Association, Physician Survey on Augmented Intelligence, March 2026 · PDF
- American Medical Association, “More than 80% of physicians use AI professionally,” 2026 · ama-assn.org
- Kraemer J., “Perceptions, Utilization, and Impact of Artificial Intelligence on Medical Science Liaisons: A Global Cross-Sectional Survey of Medical Affairs Professionals,” Cureus 18(4), April 16, 2026 · cureus.com
- FDA and EMA, Guiding Principles of Good AI Practice in Drug Development, January 14, 2026 · fda.gov
- McGuireWoods, “FDA and EMA Provide Guiding Principles for AI in Drug Development,” January 27, 2026 · mcguirewoods.com
- FDA Center for Drug Evaluation and Research, “Artificial Intelligence for Drug Development” · fda.gov
The brands that win know something others don't.
Your clinical audience is already fluent. Matching that fluency is an intelligence question before it's a tooling question. Ringer Sciences delivers always-on intelligence across your audience, your narrative, and your market: Pulse reading the conversation around your therapeutic area, Echo auditing how AI describes your science, and Trace mapping the voices that move a field. Human-led, AI-powered.