Introduction: the impossible task
The National Institute for Health and Care Excellence (NICE) has published well over 20,000 pages of evidence and recommendations. Add to that the entirety of the British National Formulary (BNF), the comprehensive Clinical Knowledge Summaries (CKS), and a constant stream of MHRA alerts, and you have a mountain of data that no single human can be expected to memorise and recall with perfect accuracy.
This is the clinician's daily dilemma. In the middle of a busy clinic or a high-pressure on-call shift, you need instant, precise, and up-to-date answers. There is no time to sift through dozens of search results or lengthy PDF documents. You need the right information, right now.
This raises a crucial question about the role of artificial intelligence in medicine. It’s not just "can AI find information?", but "can AI understand and reliably relay complex UK clinical guidance?" This is the exact problem iatroX was built to solve. Let's take a walk through the engine that powers it.
Beyond a simple search: how modern clinical AI works
To understand what makes iatroX different, it helps to use an analogy.
A standard search engine is like a library index card. It’s useful – it tells you which book on the shelf might contain the answer you're looking for. But the hard work is still yours. You have to find the book, locate the right chapter, read it, and hope you've interpreted the correct part.
iatroX is like having a specialist clinical librarian. This librarian has already read every single book in a curated, UK-specific medical library. They have understood the content, cross-referenced it, and when you ask a question, they can give you the precise paragraph you need instantly, citing the exact source, page, and edition.
This advanced capability is made possible by a process called Retrieval-Augmented Generation (RAG), which we’ve tailored specifically for the demands of UK healthcare.
Inside the engine part I: the "walled garden" of knowledge (retrieval-augmented generation)
Step 1: building a trusted library
The single most important safety feature of the iatroX engine is that it searches its own exclusive, built-in library. This is its "walled garden," a secure space that protects it from the unreliable, unvetted, and often US-centric information of the open internet.
This is the 'Retrieval' part of RAG. When you ask iatroX a question, its first step is to retrieve all the relevant passages from within this trusted library, which contains only verified UK clinical sources like:
- NICE guidelines
- Clinical Knowledge Summaries (CKS)
- The British National Formulary (BNF and BNFc)
- MHRA drug safety updates
- SIGN and RCGP guidelines
- And other authoritative UK sources
Once the AI has retrieved the correct, relevant passages, it then uses its advanced language skills to generate a clear, natural-language answer. This is the 'Augmented Generation' part. Critically, the answer is "augmented" by, or "grounded" in, the facts it has just found. Our engine is strictly forbidden from making things up or "hallucinating" – it can only use the evidence it has just retrieved from its trusted library.
Inside the engine part II: understanding the clinician's intent
Step 2: it's not just what you ask, it's why you ask
A simple search engine looks for keywords. But in medicine, context is everything. A question like "What about metformin?" can mean dozens of different things. This is where the next layer of our engine comes into play.
We have developed proprietary classifiers that instantly analyse your query to understand the probable clinical intent. Is this a dosing question? A query about side effects? A question about its use in pregnancy or renal impairment?
By understanding the intent, the engine retrieves the most relevant piece of evidence to answer your specific need. A query about metformin and renal impairment will therefore prioritise the dosing tables and recommendations from the BNF, not a general summary of the drug's mechanism of action. Further proprietary algorithms then work to ensure the final answer is presented with absolute clarity, removing ambiguity so you can make decisions with confidence.
The ultimate test: a commitment to accuracy
Putting our engine to the test: reliability you can trust
We believe that in clinical practice, trust is not given; it is earned through performance and transparency. That's why we are committed to rigorously testing and validating our engine's performance.
In rigorous internal testing of our latest version, shipped in May 2025, the iatroX engine demonstrated an accuracy rate exceeding 99% for primary care and medication-related queries.
What do we mean by "accuracy"? We define it as the substance of the answer generated by the AI directly and correctly reflecting the guidance and context from the trusted source material it has cited.
This industry-leading standard is not maintained by algorithms alone. It is upheld by our dedicated internal team of UK clinicians who perform continuous validation, quality assurance, and review of our engine's outputs. This human oversight ensures the iatroX engine remains one of the safest and most reliable clinical information tools available.
Conclusion: from information overload to intelligent augmentation
The iatroX method is fundamentally different from a generalist AI. By combining a curated, "walled garden" library with an intelligent understanding of clinical intent, we deliver answers with a proven, near-perfect accuracy rate.
This transforms the AI from a simple search tool into a professional instrument. It is designed to augment your own expertise and clinical judgment, freeing you from the immense cognitive load of information recall. This allows you to focus your time, attention, and skills on what truly matters: the patient in front of you.
Call to action
We invite you to experience the speed, reliability, and safety of the iatroX engine for yourself. Download the app and see how quickly you can get trusted answers to your next clinical question.