The Question That Never Gets Asked
It is 6:40 pm in a busy outpatient clinic. A physician is about to add a second agent for a patient with diabetes and reduced kidney function, and she wants to confirm the dose adjustment. The product monograph is a 40-page PDF somewhere in her inbox. The rep visited three weeks ago. Nine patients are still waiting.
So she goes with memory and moves on.
This is not a rare failure. It is the normal pattern, and researchers have measured it.
The "Half-Unanswered" Problem, in Numbers
In a systematic review in JAMA Internal Medicine, Del Fiol and colleagues pooled 72 studies of how clinicians handle questions during patient care:
| Measure | Finding |
|---|---|
| Questions raised | 0.57 per patient seen |
| Questions pursued | 51% |
| Answers found, when pursued | 78% |
The authors note that this pattern had held for decades, even though online evidence resources were widely available. The problem was never whether answers existed. It was the effort needed to reach them at the moment of care.
From Bookshelf to Answer Engine
How doctors find information has changed in four clear stages:
| Era | How doctors got answers | The limit |
|---|---|---|
| Bookshelf | Textbooks, formularies, journals | Slow and quickly out of date |
| PDF and rep | Monographs, reprints, rep visits | Approved, but hard to search and easy to ignore |
| Search box | Online references, search engines | Still too slow for a 10-minute consult |
| Answer engine | Conversational AI | Fast, but only as reliable as its source |
The fourth era is the first one where getting an answer takes about as much effort as asking the question.
What Doctors Are Doing With AI Today
Global Adoption
The AMA's 2026 Physician Survey on Augmented Intelligence, based on responses from about 1,700 physicians, found that:
- 81% of physicians use AI professionally, more than double the 2023 figure.
- The average physician now uses 2.3 AI use cases, up from 1.1.
- 39% use AI to summarise research and standards of care, the most common use.
Purpose-built clinical AI tools are spreading fastest. NBC News reported in May 2026 that one AI medical search tool reached roughly 65% of US doctors in a single month, according to company figures.
In WhatsApp-First Markets
In South Asia, MENA and Southeast Asia, the same behaviour runs through a different channel: the messaging app already on every doctor's phone.
In a national survey of 718 physicians in Pakistan, WhatsApp was the app used most often each week for medical purposes. It was also the most common digital channel pharma companies used to reach those doctors, ahead of websites, apps and e-detailing. A 2026 study of healthcare workers at three tertiary hospitals in Peshawar again found WhatsApp to be the most used platform.
Doctors in these markets are ready for AI answers. They want them in the channel they already use.
Where AI Still Gets Medicine Wrong
Speed is not the same as safety, and the evidence shows it.
A 2023 JAMA Network Open study had 33 specialists grade a general chatbot's answers to 284 medical questions. Accuracy was high on average, but the authors reported important limitations.
Newer research highlights a deeper issue. A 2025 study in Nature Communications tested twelve large language models on medical questions. All of them showed significant metacognitive deficiencies: they failed to recognise the limits of their knowledge and gave confident answers even when no correct option was available.
For a doctor checking a dose, a confident wrong answer is worse than no answer. For a pharma company, open-web AI creates three risks:
Risk 1 – Label Mismatch
The answer may reflect another country's approval, an older label or a withdrawn indication.
Risk 2 – Off-Label Drift
A general model has no concept of what the company is permitted to say.
Risk 3 – Lost Signal
The doctor's question, one of the most valuable data points a brand can get, disappears into a third-party tool.
Governed AI – The Model That Fits Pharma
The fix is not to keep doctors away from AI. It is to give them AI that follows the same rules as every other company communication.
Industry codes already set that bar. The IFPMA Code of Practice, which member companies follow worldwide, requires promotional information to be accurate, not misleading and consistent with approved product information, and it restricts off-label promotion. In the US, the FDA's January 2025 final guidance on scientific information about unapproved uses follows the same logic, although it is not yet in effect pending federal paperwork review. National regulators add their own rules on top.
| Trait | Open-web AI | Governed pharma AI |
|---|---|---|
| Source of answers | Whatever the model learned | Only the company's approved content |
| Off-label questions | Unpredictable | Blocked, or escalated to Medical Affairs |
| Who can ask | Anyone | Invited healthcare professionals |
| What the company learns | Nothing | Every question, by region and product |
How WaaPharma Brings Approved Clinical AI to Doctors
WaaPharma was built for this shift. We give pharma brands approved, AI-powered clinical answers from trusted pharma content, delivered inside the app doctors already open dozens of times a day: WhatsApp.
How It Works
- A rep sends the doctor a personalised WhatsApp invite with a "Launch WaaPharma" button.
- The doctor chooses a therapeutic area, such as Cardiometabolic, Nephrology or Neurology.
- The doctor asks about the portfolio, molecules, dosing, interactions, evidence or guidelines, 24/7.
- Every answer comes only from approved content, with no off-label responses.
- The pharma team tracks queries, sample requests and escalations on a real-time dashboard.
Because WaaPharma runs on Meta WhatsApp Flows, there is no app to download and no login to remember.
Back to the physician at 6:40 pm. With WaaPharma, her dose question takes one message, and the answer she gets is the approved one.
Results From Live Deployments
- 72.3% doctor usage rate
- 4.14 queries per doctor
- 33.9% invite-to-active onboarding
- 48 hours to go live for a new product launch
Frequently Asked Questions
How is AI changing the way doctors access medical information?
It replaces searching through documents with asking direct questions. The AMA reports that 81% of US physicians use AI at work, and research and guideline summaries are the top use.
How often do doctors have unanswered clinical questions?
A JAMA Internal Medicine systematic review found that clinicians raise about 0.57 questions per patient and pursue only about half of them.
Is general AI reliable enough for medical questions?
Not on its own. A 2025 Nature Communications study found that all twelve models tested gave confident answers even when they lacked the correct information. AI restricted to approved sources is safer for clinical and pharma use.
Do doctors in Pakistan use WhatsApp for medical purposes?
Yes. In a national survey of 718 Pakistani physicians, WhatsApp was the most frequently used app each week for medical purposes.
What is WaaPharma?
WaaPharma is a WhatsApp-based clinical AI platform that lets pharma companies answer HCP questions 24/7 from approved content, with real-time analytics on every interaction.
See Approved Clinical AI in Action
Your doctors are already asking AI about your medicines. Make sure the answer is yours. Book a 30-minute WaaPharma demo to see approved medical information delivered to HCPs inside WhatsApp, live in 48 hours.
Book a DemoReferences
- Del Fiol G, Workman TE, Gorman PN. Clinical questions raised by clinicians at the point of care: a systematic review. JAMA Internal Medicine. 2014;174(5):710–718.
- American Medical Association. 2026 Physician Survey on Augmented Intelligence. March 2026.
- Griot M, et al. Large language models lack essential metacognition for reliable medical reasoning. Nature Communications. 2025;16:642.
- Goodman RS, et al. Accuracy and reliability of chatbot responses to physician questions. JAMA Network Open. 2023;6(10):e2336483.
- Pharmaceutical digital marketing and its impact on healthcare physicians of Pakistan: a national survey. Cureus. 2018.
- Ali I, et al. Digital professionalism and social media use among healthcare workers in tertiary care hospitals of Peshawar, Pakistan. Khyber Medical University Journal. 2026;18(1):100–107.
- IFPMA. IFPMA Code of Practice 2019.
- US FDA. Communications From Firms to Health Care Providers Regarding Scientific Information on Unapproved Uses. Federal Register, January 7, 2025.
- NBC News. Most U.S. doctors are quietly using this AI tool. May 2026.