Lawyers write. A lot. All day. Letters, briefs, pleadings, memos, meeting notes, client updates, emails to opposing counsel — written output is the core of the profession. And yet, in most firms, that output is still produced at 40 words per minute on a keyboard, the same way it was in the 1990s.
This is not a technology gap. It is a trust gap. Voice dictation has long carried two dealbreakers for the legal profession: insufficient accuracy on legal vocabulary, and an unacceptable risk to client confidentiality. Both of those locks have just been picked. And the productivity stakes — recovering two to three hours a day — have become impossible to ignore.
This article explains how modern voice dictation actually works inside a law practice in 2026, what it changes in concrete terms, and the conditions under which it can be adopted without conflicting with the duty of confidentiality.
Why dictation is an ergonomics question, not a gadget
Let’s start by looking at what really happens in a lawyer’s day.
A partner in a commercial firm produces, on average, between 6,000 and 10,000 words a day across all formats. A litigation associate often produces more. Roughly half of it is “creative” output — briefs, legal memos, drafting clauses. The other half is operational communication: emails to the client, letters to the other side, meeting summaries, internal instructions.
At 40 words per minute on a keyboard, 8,000 words is more than three hours of pure typing — before edits, before proofreading, before the friction of switching from one matter to the next. In practice, a lawyer spends four to five hours a day with a finger on a keyboard. Over a ten-hour day, that is half the available time, gone to mechanical input rather than to the client.
Modern voice dictation captures 150 to 180 words per minute — close to four times faster than typing. The theoretical gain is enormous. The real-world gain, once the learning curve is behind you, lands around two hours a day for a lawyer who dictates most of their correspondence and the first draft of their written work. Over a month, that is roughly 40 hours recovered — a full working week. Over a year, more than 500 hours: time returned to the matter, to the client, or to a life outside the office.
That gain alone is what justifies looking seriously at the question. Everything else — the quality of the text, the confidentiality of the data — determines whether the tool gets adopted or quietly abandoned.
The historical problem: legal vocabulary, badly understood
For years, voice dictation stumbled over legal vocabulary. Latin terms (a fortiori, res judicata, prima facie, ratione materiae), citations (statute sections, case names, court abbreviations), proper nouns (courts, judges, parties), and acronyms — generic transcription engines mangled all of it.
The result: raw text that was unreadable, a correction time that wiped out the speed advantage, and the lasting impression that the technology would never deliver on its promise.
In 2026, the situation is different, for two reasons. First, transcription models have made a dramatic leap in quality. OpenAI’s Whisper, released in 2022, and the European models that followed, pushed the accuracy standard from around 90% to well above 97% on everyday language — technical vocabulary included.
Second, and more importantly, modern tools no longer just transcribe. They process the text with a language model that understands context. If you dictate “the Court of Appeal held that,” the system writes it correctly. If you say “ratio nay materiae,” it returns “ratione materiae.” If you dictate “section twelve forty of the civil code,” it writes “section 1240 of the civil code.” That layer of comprehension changes everything.
The concrete result: a lawyer can now dictate a brief or a client letter with enough accuracy that review becomes a simple check rather than a rewrite.
The confidentiality lock — and why it just broke
For the profession, confidentiality is not a convenience. It is a duty — the attorney-client privilege, the rules of professional conduct, and the obligations every bar imposes on its members. Any tool that sends voice recordings or legal text to foreign servers, without guarantees about how that data is handled, poses a real problem — not a theoretical one.
And that is exactly what most legacy solutions did, and still do. Dragon NaturallySpeaking, long the market reference, is a US product whose servers sit in the United States. Microsoft 365 dictation routes through Azure. And even Apple Dictation, while it processes locally on recent Apple Silicon chips, is not auditable and remains a black box from the firm’s standpoint.
GDPR compliance, the fallout from the Schrems II ruling, and growing awareness of the US CLOUD Act have changed the calculus. A lawyer can no longer route privileged content through a service whose data sovereignty they cannot guarantee. For a matter involving litigation strategy, settlement discussions, medical records, trade secrets, or privileged communications, that risk can be disqualifying on its own.
This is what makes European solutions relevant. Dikto, for example, uses Mistral AI — a French model, hosted in Europe, governed exclusively by European law. Transcriptions are not retained after processing, are not used to train models, and are not accessible to foreign authorities.
That does not relieve the firm of its own obligations — a record of processing activities, a mention in the engagement letter where appropriate, informing the client. But it makes the solution usable within a demanding ethical framework, instead of disqualifying it before the conversation even starts.
Three uses where dictation changes the game
Not every legal task lends itself equally well to dictation. Here are three concrete uses where the payoff is immediate.
1. Letters to the client and to opposing counsel
This is the fastest win to capture. A typical firm letter runs 200 to 500 words. On a keyboard, that is ten to twenty minutes depending on complexity. By voice, three to five minutes, proofreading included.
The advantage goes beyond time saved. When you dictate a letter instead of typing it, you write it the way you would say it — in a tone that is more natural, more direct, more human. Many lawyers find their letters get clearer once they switch to voice, because speaking aloud exposes the tangled sentences the keyboard lets through.
In concrete terms, a lawyer who sends 15 to 20 letters a day can bring their correspondence down from three hours to one.
2. The first draft of briefs and memos
Dictation is not the tool for producing the final version of a brief. The work of argument, structure, and citing authority remains intellectual labor that demands reading and editing at the keyboard. But it is a formidable tool for producing the first draft — the one where you lay out the facts, the central argument, the early citations.
The most effective method: prepare a quick outline by hand or at the keyboard (section headings, essential citations), then dictate the content of each section, following that outline. What would take two hours to write at the keyboard takes forty to sixty minutes. The editing that follows is shorter, too, because the structure is already there.
3. Meeting and call summaries
A lawyer walks out of a 90-minute client meeting. The right reflex — the one that gets lost for lack of time — is to write the summary immediately. At the keyboard, that is thirty to forty-five minutes on top of the ninety already spent. The real-world practice: it gets put off, details are lost, and the summary often never gets written at all.
By voice, the same summary takes ten to fifteen minutes, right after the meeting, while the memory is fresh. The file stays current. Time recording is more accurate. Handing the matter to a colleague is immediate.
On this use alone, voice dictation pays for itself within the first two weeks.
Which tool for a law firm?
Three categories of solution are relevant today.
Dragon Legal remains the historical reference. Very complete, with a deep legal dictionary. But: a prohibitive price (around $700 per seat, plus upgrades), no macOS version since 2018, and data held in the United States. For a Windows-only firm that can afford it, it is still an option. For Mac users, it no longer is.
Apple Dictation built into macOS is free and runs locally on Apple Silicon Macs. Adequate for short notes or messages. Inadequate for professional output: no text cleanup, no rephrasing, no handling of hesitations, and only middling accuracy on legal vocabulary.
Dikto is a European solution built for exactly this kind of use. Transcription by Whisper, processing by Mistral AI hosted in Europe. The output is cleaned, punctuated, and structured. Dictation works in any macOS application — Word, Outlook, Pages, Slack, Safari, or directly inside a practice management tool. The price is €9 a month, or €89 a year. For a firm, that is the investment with the best return in the last decade.
The app is available at dikto.ai and sets up in under five minutes — one keyboard shortcut, and dictation is available everywhere.
Rolling out dictation in a firm: the method
Three pieces of advice for a successful rollout, drawn from the experience of lawyers who use it.
First, start with a narrow use. Do not try to dictate everything on day one. Pick one simple use — client correspondence, for example — and stick to it for two weeks. The habit forms, the tone settles, the reflexes take hold.
Second, structure your dictation. Do not launch into a long text without a plan. Say the structure out loud (“I’m going to write a letter in three paragraphs: subject, statement of facts, request”), then dictate each section. The output is cleaner, and the review faster.
Third, fold dictation into your existing workflow. There is no need to change your practice management tool, your word processor, or your email client. A good dictation solution works across every application. Yours simply shows up as a new input method, complementary to the keyboard.
The bottom line
Voice dictation is not a passing fad or a toy for early adopters. It is a structural shift in professional practice, comparable to the move from paper to email correspondence in the 2000s. The first to adopt it gain a real edge over those who wait.
For lawyers, the stakes go beyond individual time savings. It is also the ability to keep matters better updated, to send correspondence faster, to increase the volume each associate can handle without degrading quality, and to reduce the cognitive fatigue that comes from four hours of daily keyboard input.
On one condition: that the chosen solution respects the duty of confidentiality. That is now possible with European tools, GDPR-compliant, whose use is compatible with the firm’s professional obligations.
A serious trial takes two weeks. The return on investment takes two days.