Heidi has shown how much value can be unlocked by reducing documentation burden. Over 15 million UK sessions. An estimated 4 million clinical hours returned. Used by one in two UK GPs. The numbers demonstrate that when AI solves an obvious daily pain point — the time spent typing, formatting, and completing notes — adoption can be rapid, widespread, and operationally meaningful.
The next layer of clinical AI is what happens around that documentation: answering clinical questions, checking guidance, supporting prescribing decisions, logging CPD, and helping clinicians learn.
Heidi and the Documentation Layer
Heidi's ambient voice technology captures the consultation — listening in the background while the clinician focuses on the patient — and generates structured clinical documentation for review. The clinician checks, edits, and approves. The note enters the record. The value is time saved, reduced after-hours work, better patient eye contact, and lower cognitive load from administrative tasks.
This is genuinely important. Documentation is foundational to safe care — it supports continuity, communication between clinicians, prescribing decisions, referral quality, medico-legal defensibility, and clinical audit. Poor documentation can increase medical errors, misdiagnosis, and inappropriate treatment. A tool that makes documentation faster and more complete — while keeping the clinician in the review loop — improves a critical clinical process.
Why Documentation Is Foundational but Not the End of the Workflow
The clinical consultation generates multiple outputs beyond the note. A management plan that should align with current guidelines. A prescribing decision that should be safe for this specific patient. A referral that should include the information the receiving clinician needs. Safety-netting advice that should be specific, time-bound, and presentation-appropriate. A learning point — something the clinician encountered for the first time, or where they were uncertain, or where the guideline had changed since they last checked.
The documentation captures what happened. The clinical knowledge layer supports what should happen — and why.
The Clinical Questions That Arise After the Note
Consider the workflow of a GP who has just finished a consultation and is reviewing the AI-generated note.
"The note says I prescribed amoxicillin — but is that the first-line recommendation for this presentation, or should I have used a different antibiotic per the local formulary?" This is a guideline and formulary question.
"The note documents chest pain with a musculoskeletal assessment — but should I have calculated a Wells score or used the HEART pathway to rule out higher-risk causes?" This is a risk stratification question.
"The note includes safety-netting advice — but is it specific enough? What red flags should I have included for this particular presentation?" This is a clinical safety question.
"I was unsure about the management of this patient's comorbidity combination — I want to check what NICE says and record this as a learning point for my appraisal." This is a CPD question.
None of these questions are answered by the documentation tool. They require a different capability: cited clinical information retrieval, calculators, guideline-grounded reasoning, and learning capture.
Guideline-Grounded Search as the Next Layer
The documentation layer captures the consultation as it happened. The guideline layer helps the clinician verify that what happened aligns with current best practice. These are complementary — and both are necessary for safe, high-quality care.
A clinician who documents efficiently but never checks guidelines may produce a clean record of suboptimal care. A clinician who checks guidelines obsessively but cannot document efficiently may fall behind, carry admin home, and burn out. The clinical AI stack needs both layers — and the best workflow integrates them.
How iatroX Supports Clinicians Beyond Documentation
On iatroX, the emphasis is not ambient note generation. It is the next step: cited clinical answers, clinical calculators, exam Q-banks, and CPD workflows for clinicians who need to turn a clinical question into a defensible action.
Ask iatroX answers the guideline question. Calculators answer the risk stratification question. CPD captures the learning question. Q-banks reinforce the knowledge base that underpins all three.
The workflow is complementary: use Heidi (or any AI scribe) to document the consultation. Use iatroX to verify the clinical reasoning, check the guideline, calculate the score, and save the learning.
Why the Future Is Not One AI Tool but a Clinical AI Stack
Heidi's own report describes ambient voice technology as "the first step in a broader evolution towards clinical decision support and agentic AI workflows." The trajectory is clear: AI in clinical practice will not be one tool doing everything. It will be a stack of specialised layers — documentation, evidence retrieval, guideline verification, risk calculation, learning, and governance — each serving a different clinical moment.
The tools that earn daily use will be those that fit specific clinical workflows without requiring the clinician to abandon their existing practice patterns. Heidi fits the documentation workflow. iatroX fits the clinical knowledge workflow. Together, they represent two layers of the same shift: practical, clinician-facing AI that makes the working day safer, faster, and more sustainable.
Use iatroX alongside documentation tools for guideline-grounded answers, calculators, and CPD →