AI Scribe vs Dictation vs Human Transcription (UK, 2026)
UK clinicians turn consultations into notes three ways in 2026. Ambient AI scribes (Motics, Heidi, and peers) listen to the consultation with consent and draft the structured note — least clinician effort, newest model. Speech-recognition dictation (Lexacom Echo, Dragon Medical One) types what you dictate — mature, fast, but you still compose every sentence. Human transcription (typists or services like Lexacom Scribe) converts your dictation later — familiar, with per-minute costs and turnaround delays. The incumbents themselves signal where this is heading: Lexacom now sells its own Ambient AI, and Microsoft cross-promotes Dragon Copilot — its ambient AI product — from the Dragon Medical One page.
The three documentation models compared
| Ambient AI scribe | Speech-recognition dictation | Human transcription | |
|---|---|---|---|
| Who composes the note | AI drafts from the consultation; clinician reviews and edits | Clinician composes aloud; software types it | Clinician composes aloud; a person types it later |
| Clinician time per note | Review only — typically the biggest saving | Dictation + correction time | Dictation time now, review later |
| Turnaround | Seconds–minutes after the session | Real-time | Hours–days, depending on service |
| Structure (SOAP, letters) | Drafted automatically to template | Whatever you dictate | Whatever you dictate, formatted by the typist |
| Patient audio captured? | Yes (ambient mode) — consent required; dictation mode available | No — clinician voice only | No — clinician voice only |
| Example tools (UK) | Motics, Heidi, Preve, TORTUS (NHS) | Lexacom Echo, Dragon Medical One | Lexacom Scribe, in-house secretaries |
| Cost model | Free tiers; per-seat £15–£65/user/month billed yearly; usage credits (Motics) | Licence pricing on request (Lexacom, Dragon Medical One) | Per audio minute or salaried staff |
The real difference: who does the composing
Strip the branding away and clinical documentation has one bottleneck: composing the note — deciding what to say and saying it in clinical language. Human transcription and speech recognition both leave composing with the clinician; they only change who types. That's why a GP who dictates fluently saves real time over typing, but still spends cognitive effort narrating every letter. Ambient AI moves the composing itself: the consultation happens naturally, the AI drafts the structured note — subjective, objective, assessment, plan, or a referral letter — and the clinician's job collapses to review and correction. The saving isn't typing speed; it's not having to narrate your day twice.
That framing also explains the failure modes. Ambient drafts occasionally include errors a composer wouldn't make (omissions, misattributed speakers), which is why clinician review is mandatory. Dictation produces exactly what you said — including everything you forgot to say. Human transcription adds turnaround delay and a per-minute meter. None of the three removes the clinician's responsibility for the filed record.
The UK incumbents — and which way they're pointing
Lexacom (Banbury, 25+ years, around 25,000 users) is the UK healthcare dictation incumbent — digital dictation and workflow (Lexacom 3), speech recognition (Lexacom Echo), and human transcription on a pay-per-audio-minute model (Lexacom Scribe), with a strong NHS and primary-care base. The strategic tell: Lexacom's unified platform now includes its own Ambient AI product — the incumbent itself treats ambient as the destination. Dragon Medical One, the medical speech-recognition standard originally from Nuance and now sold by Microsoft, remains a mature, capable dictation product for UK clinicians — and Microsoft's own UK product page cross-promotes Dragon Copilot, its ambient AI documentation product, from the same page. Neither incumbent publishes UK pricing; both quote per deployment.
Read the direction honestly: the companies that built their businesses on dictation are building ambient products. That doesn't make dictation obsolete — it makes it a mode rather than a strategy.
When dictation still wins
- Patients who can't consent to recording. Ambient capture needs informed consent; for patients lacking capacity (community, neuro, elderly-care caseloads), Mental Capacity Act principles point to not recording at all. Dictating findings after the session documents the care without capturing patient audio — which is why a dictation mode should be a hard requirement in any AI scribe you buy.
- Acoustically hostile sessions. Gym floors, hydrotherapy pools, group classes, multiple simultaneous conversations — environments where ambient capture degrades. Dictating afterwards is more reliable than transcribing chaos.
- Clinicians who compose best aloud. Some clinicians produce their tightest clinical reasoning by narrating it. For them, dictation into an AI that structures the output (rather than raw speech-to-text) is the sweet spot — modern scribes' dictation modes do exactly this.
- Medico-legal caution about recordings. Some clinics prefer not to create consultation recordings at all as a matter of policy. Dictation workflows sidestep the question entirely.
How we chose
Vendor facts were verified on 10 June 2026 against lexacom.co.uk (products and positioning; Lexacom publishes no pricing) and Microsoft's UK Dragon Medical One page (which itself cross-promotes Dragon Copilot; no public UK pricing). AI scribe pricing references our cost guide, where every figure was read from vendors' public pricing pages. We've avoided circulating-but-unverifiable claims about the dictation market (including usage-share figures we couldn't pin to a current primary source).
Motics is our product and sells the ambient-plus-dictation model this page concludes in favour of — weigh that, and note the conclusion is also the incumbents' own: Lexacom sells Ambient AI and Microsoft sells Dragon Copilot. Lexacom and Dragon are trademarks of their owners; Motics is not affiliated with or endorsed by either.
Frequently asked questions
What's the difference between an AI scribe and dictation software?
Dictation software (Lexacom Echo, Dragon Medical One) types what you say — you still compose every sentence of the note aloud. An ambient AI scribe listens to the consultation itself (with patient consent) and drafts the structured note for your review — the composing moves to the AI. That's the time difference: dictation saves typing; ambient saves narrating your day twice. Most modern AI scribes also include a dictation mode, so the choice in practice is a tool that does both versus one that only types.
Is an AI scribe better than Dragon Medical One?
Different jobs, converging vendors. Dragon Medical One is mature medical speech recognition — excellent at turning fluent dictation into text in real time, licensed per deployment (UK pricing on request from Microsoft). It doesn't draft notes from the consultation itself. Microsoft's own UK page points users toward Dragon Copilot, its ambient product, for that. If your clinicians dictate happily and only want typing removed, DMO remains a strong choice; if the goal is not composing the note at all, that's the ambient scribe category — from Microsoft, the NHS-focused tools, or the clinic suites like ours.
Does Lexacom have an AI scribe?
Yes — Lexacom's unified platform now includes an Ambient AI product alongside its established lines: Lexacom 3 (digital dictation and workflow), Lexacom Echo (speech recognition), and Lexacom Scribe (human transcription billed per audio minute). Pricing is on request. For NHS practices already running Lexacom workflows, it's the lowest-friction route into ambient documentation; private allied-health clinics typically compare it against the scribes built for their PMS ecosystems (Cliniko, Nookal, Meddbase) and their note styles.
Is human transcription still worth using in 2026?
In shrinking niches. Human transcription (services like Lexacom Scribe, or in-house secretaries) still makes sense for long-form complex correspondence where a skilled medical secretary adds judgement, and for clinicians who want zero workflow change. Against it: per-audio-minute costs that scale linearly with caseload, turnaround measured in hours or days rather than seconds, and the same composing burden as dictation. Most clinics moving off transcription jump straight to ambient AI rather than to speech recognition.
How do the costs compare across the three models?
Structurally differently. Human transcription bills per audio minute (rates vary by service) — costs scale with every consultation, forever. Dictation software licenses per user, with UK pricing on request at both Lexacom and Dragon Medical One. AI scribes are the most price-transparent category: free tiers (Heidi, Preve), per-seat plans at £15–£65/user/month billed yearly, or usage-based credits (Motics — plans sized to what the clinic documents, shared across unlimited users). For a clinician documenting a full diary, any of the AI options typically costs less per month than a few hours of per-minute transcription.
When should a clinic choose dictation over ambient recording?
Four cases: patients who can't give informed consent to recording (capacity — Mental Capacity Act principles point to dictation, not recording); acoustically hostile sessions like gym floors and group classes; clinicians whose clinical reasoning is genuinely sharper when narrated; and clinics with a policy preference against creating consultation recordings. The practical conclusion isn't choosing dictation as your tool — it's requiring dictation mode in whatever ambient scribe you buy, so the same templates and review workflow cover the whole caseload.
Do I still need to check notes if a human typed them or an AI drafted them?
Yes, identically. The clinician who files the note is responsible for it under HCPC record-keeping standards and CQC's accurate-complete-contemporaneous expectation, regardless of whether it came from an AI draft, speech recognition, or a medical secretary. The review burden differs — AI drafts need checking for omissions and occasional misattribution, dictation output for recognition errors, human-typed letters for transcription slips — but the accountability is constant. Tools change who types; they don't change whose name is on the record.
Sources
- Lexacom — products (Lexacom 3, Echo, Scribe, Connect)
- Lexacom — unified platform including Ambient AI
- Microsoft — Dragon Medical One (UK)
- Mental Capacity Act 2005 — legislation.gov.uk
- HCPC — Standards of conduct, performance and ethics (records)
- Heidi — UK pricing (per-seat reference points)
- Motics — pricing