Dictation for doctors: a workflow for clinical notes
Clinicians who dictate well rarely speak a note from top to bottom in one breath. They move through a template field by field, say numbers in a predictable way, and reread a few high-risk items every time before signing. This guide is about that workflow on a general-purpose computer. It is not advice on which software your organisation may use: that decision belongs to your employer, your IT department and the rules where you practise, and you should check with them before dictating anything that identifies a patient.
First, find out what you are allowed to use
Hospitals and practices usually have an approved clinical speech system, often built into the electronic record and trained on medical vocabulary. If yours does, that is normally the right tool for patient notes, and a general dictation app is the wrong one regardless of how well it works. General tools are still useful for everything around the record: referral letters drafted without identifiers, teaching material, emails to colleagues about logistics, research writing.
If you work independently and choose your own tools, read the next sections and the questions for IT before deciding. The broader guide to dictation privacy explains local, cloud and hybrid processing in more detail.
Build templates that you fill by voice
The fastest way to dictate a note is to let the template carry the structure, so your voice only supplies content. Keep the headings in the document or record, click into each field, and dictate a short block. A simple outpatient template might look like this:
- Presenting complaint: one sentence. Example: “Three days of productive cough and fever, worse at night.”
- History: short sentences rather than one long chain. Pause between them so sentence breaks land correctly.
- Examination: dictate findings in a fixed order every time, so missing items are easy to spot.
- Assessment: the working diagnosis and differentials as a short list.
- Plan: one action per line. Say “new line” between items.
Because you click into a field before speaking, an error stays inside one section instead of spreading across the whole note. Templates also make it obvious when a field is empty.
Phrasing that survives speech recognition
Speech models handle everyday language well and struggle most with short, similar-sounding words that carry clinical weight. A few habits reduce the risk:
| Item | Risk | Habit |
|---|---|---|
| Doses | “fifteen” and “fifty” can be confused | Say the number then the unit clearly: “five hundred milligrams twice daily”, then check the digits |
| Negatives | “no” at the start of a phrase can drop out | Use a full sentence: “There is no chest pain” rather than a fragment |
| Laterality | left and right are short words | Say it twice in different forms if needed: “left knee, the left side”, then confirm |
| Drug names | Rare names become common words | Paste names from the record or a list instead of dictating them |
| Abbreviations | Spoken letters are unpredictable | Dictate the full term, abbreviate by keyboard if your local style requires it |
Your organisation may have a list of abbreviations it does not allow in records. Dictating full terms avoids the question.
Correction habits before you sign
Rereading every word of every note is not realistic in a busy clinic. Rereading a small fixed set of items is. Make it a ritual of under a minute:
- Every number: doses, frequencies, dates, measurements, results.
- Every left and right.
- Every sentence containing “no”, “not”, “denies” or “without”.
- Every drug name and allergy.
- The patient's name and identifiers, if the note contains them, against the record header.
If you find the same mistake three times in a week, change the habit rather than the correction: paste that drug name, or say that number differently. Errors cluster around the same words for each speaker.
Privacy questions to ask your IT department
Take these to IT or your information governance contact before using any dictation tool near patient information. Get the answers in writing.
- Which dictation or speech tools are approved for clinical documentation here?
- Is a general-purpose dictation app allowed for non-clinical work on this computer?
- If a tool sends audio to a server, which locations and vendors are acceptable?
- Do we need a contract or agreement with the vendor before staff use it?
- Are tools that process speech only on the device treated differently?
- Who do I tell if dictated text ends up in the wrong field or the wrong application?
Where speech is processed in Voice Monkeys
Voice Monkeys makes no compliance claims, and you should not treat it as approved for patient data unless your organisation says so. What is technically true: on a Mac with Apple Silicon, speech is converted on the Mac itself and works offline after a one-time model download. On Windows, Linux and Intel Macs the subscription version sends audio encrypted to a server in Germany, converts it in memory and discards it without storing audio or text; Windows and Linux can instead use an optional offline model. It types into any field, holds to talk on right Option or right Alt, and Esc cancels a take you do not want. For notes inside a record that has its own clinical speech system, use that system. For letters, teaching and admin writing, you can download Voice Monkeys and decide with your IT team. Built-in tools are covered in the guides for Mac dictation and Windows voice typing.
A sample session: an outpatient letter
Open the letter template. Click into the opening paragraph and say: “Thank you for referring this patient, whom I saw in clinic today regarding persistent knee pain comma left side period.” Click into findings and dictate two or three short sentences. Click into the plan and dictate one item per line. Then run the correction ritual: numbers, sides, negatives, drugs. A one-page letter handled this way takes a few minutes of speaking plus a minute of checking, and the structure is identical every time, which makes the checking faster.
Questions and answers
Is general dictation software allowed for patient notes?
That depends entirely on your employer and the rules where you practise. Ask IT or your governance contact; many organisations require their own approved system.
Does Voice Monkeys claim to be compliant with health privacy laws?
No. It only states where speech is processed: on the device for Apple Silicon Macs, or on a server in Germany with audio and text discarded for the cloud version.
How do I stop dose errors when dictating?
Say number and unit clearly, then check every number before signing. If a dose keeps coming out wrong, type it.
Should I dictate drug names?
Common ones usually work. Rare names are safer pasted from the record or a list, because speech models may turn them into ordinary words.
Can I dictate a note while the patient is in the room?
That is a clinical and policy question for your organisation. Technically, dictation only needs your voice and a microphone close to you.
Sources
- Apple: Use Dictation on Mac (checked November 6, 2026)
- Microsoft: Use voice typing to talk instead of type on your PC (checked November 6, 2026)