Legal dictation in Ireland has followed a surprisingly consistent pattern over the past fifty years: each new technology solved some of the problems of the last one while introducing new limitations. The tape recorder freed lawyers from the typewriter. Dragon freed them from the typist. The smartphone freed them from the desk. AI, finally, may free them from the correction cycle.

Each transition seemed like it would be the last. None of them were — until, possibly, now.

The tape-and-typist era (1970s–2000s)

For decades, the dominant model for legal documentation was simple and effective. The solicitor or barrister would dictate into a tape recorder — first full-size cassette, later micro-cassette, eventually digital recorder. The tape would go to a legal secretary or typist who would transcribe it into a typed document. The lawyer would review the typed draft, mark corrections, and the secretary would produce a final version.

This system had real virtues. The lawyer could focus on the substance — narrating the attendance note, structuring the opinion, composing the letter — without worrying about typing, formatting, or spelling. The secretary handled the mechanical aspects and often added formatting knowledge: where to put the heading, how to lay out the case reference, the firm’s standard letter template.

The division of labour was efficient. A senior partner dictating into a recorder between meetings could produce three or four pieces of correspondence in the time it would take to type one. The secretary, working from the recording the next morning, could produce polished documents at typing speed.

But the model had costs. Every law firm needed secretarial staff, typically at a ratio of one secretary to one or two fee earners. The turnaround was overnight at best — dictate in the afternoon, receive the typed draft the next morning. Rush requests disrupted the workflow. And the system depended entirely on the availability and skill of the individual secretary.

For barristers, the model was less reliable. Many barristers at the Irish Bar didn’t have dedicated secretaries. Those in larger sets of chambers might share a clerk, but the clerk’s role was administrative rather than secretarial. Many barristers typed their own work from the beginning — or handwrote it, particularly in court.

Dragon arrives: the promise of self-sufficiency (late 1990s–2010s)

Dragon NaturallySpeaking launched in 1997 and reached the legal market within a few years. The promise was transformative: the lawyer could dictate directly into their computer, bypassing the secretary entirely. No tape, no waiting, no turnaround time. Speak and see the words appear on screen.

For larger solicitors’ firms, particularly in Dublin, London, and the UK, Dragon represented a genuine efficiency gain. Firms invested in licences, trained their solicitors, and some achieved impressive results. Partners who’d spent decades dictating into tape recorders adapted their workflow to dictating into Dragon with reasonable success.

The reality, however, was more complicated than the marketing suggested.

Dragon required hours of voice training before it achieved acceptable accuracy. Each user needed their own voice profile. The software ran only on Windows desktops with significant processing power. It worked best in quiet environments — open-plan offices were problematic. And it required a level of computer literacy and patience that not every practitioner possessed.

For barristers, Dragon was particularly ill-suited. The desktop-only constraint meant it couldn’t be used between courts, in consultations, or anywhere away from a desk. The voice training required a time investment that many barristers couldn’t justify. And the cost — several hundred euro per licence — was a harder sell for sole practitioners than for firms spreading the cost across departments.

Dragon also didn’t solve the formatting problem. It transcribed speech into text in a Word document, but the output was essentially a raw transcript. The lawyer still had to add structure, format headings, and lay out the document correctly. Dragon could do some formatting by voice command (“bold that,” “new paragraph”), but producing a properly structured legal document required significant manual intervention.

The legal secretary didn’t disappear. She just changed roles — from transcriptionist to document formatter and proofreader. And the turnaround remained: the solicitor dictated into Dragon, printed the draft, marked corrections, and the secretary produced the final version. The tape was gone, but the workflow was similar.

The smartphone interlude (2010s–2020s)

When Apple introduced Siri in 2011 and Google improved its voice recognition in the same period, every lawyer suddenly had a dictation device in their pocket. For the first time, dictation was portable. You could dictate a note walking to the car park, standing in a corridor, or sitting on a train.

The legal profession’s response was enthusiastic for about a week, then quietly disappointed.

Phone dictation produced raw text — no formatting, no structure, no legal vocabulary. Case citations were garbled. Latin terms were mangled. Irish names were hopeless. The transcript needed so much correction that many practitioners found it faster to just type.

There were also concerns about where the audio went. Apple and Google both processed dictation through cloud servers, with limited transparency about data location. For lawyers handling privileged material, speaking client details into a phone that sent audio to servers in California felt uncomfortable — even before GDPR made it a regulatory concern.

The smartphone solved the portability problem but solved nothing else. Lawyers who tried mobile dictation in this era mostly concluded that dictation simply didn’t work for legal documents. It was a reasonable conclusion based on the tools available. It was also, as it turned out, a temporary one.

The AI era: 2023–present

The arrival of large language models and transformer-based speech recognition in 2023-2024 changed the dictation landscape in ways that previous incremental improvements had not.

The shift wasn’t just about better transcription accuracy — though that improved dramatically. It was about what happened after transcription. For the first time, a dictation tool could take a stream of speech and produce not just text, but a structured document. An attendance note with the right heading, participants, narrative, and action items. An opinion with proper section headings and paragraph structure. A letter with the correct salutation, reference, and sign-off.

This addressed the fundamental problem that every previous generation of dictation had left unsolved: the gap between a transcript and a usable document.

Transcription accuracy jumped to a level where corrections became the exception rather than the rule. AI models trained on legal texts handle case citations, neutral citations, and Latin terms with far greater reliability than any previous technology. Irish accents, which had been a persistent problem, are handled effectively by models trained on diverse English datasets.

Automatic formatting meant that the lawyer received a structured document, not a wall of text. The AI understood that when a solicitor dictated “attendance note, re Murphy v Dublin City Council, meeting with client on 14 March,” that was a heading, not body text. The output was a document that needed reviewing, not reformatting.

Mobile-first design meant the tools were built for the phone from the start, not adapted from desktop software. Dictating between appointments became not just possible but natural.

EU-only processing became available for the first time, addressing the GDPR concerns that had made security-conscious practitioners uneasy about cloud-based dictation.

The pattern breaks

Every previous generation of dictation technology solved some problems while leaving others untouched:

Tape recorders freed lawyers from the typewriter but required secretaries and overnight turnaround
Dragon freed lawyers from the secretary (partially) but required a desktop, training, and manual formatting
Smartphones freed lawyers from the desk but produced unusable raw text

The AI generation is the first to address all of these simultaneously: it’s mobile, it requires no training, it handles legal vocabulary, and it produces formatted documents. The lawyer’s role shifts from author and typist to reviewer and editor.

This doesn’t mean the technology is perfect. AI dictation still makes mistakes. It still requires review. Complex documents still benefit from careful editing. But the baseline has shifted from “spend 20 minutes correcting a two-minute dictation” to “spend two minutes reviewing a two-minute dictation.”

Where things stand

AI dictation for barristers and solicitors is still early. The accuracy is there, the formatting is genuinely useful, and the mobile-first model fits how lawyers work. But adoption takes time — especially in a profession where the tools you use carry professional liability implications.

What’s different this time is that the barriers that stopped previous generations aren’t present. There’s no voice training period. There’s no desktop requirement. There’s no €500 licence fee to justify before you’ve seen a result. A practitioner can try AI dictation on their phone after their next consultation and know within five minutes whether it’s useful.

The tape recorder had a good run. Dragon had its era. The smartphone was a false start. Whether AI dictation becomes the next standard tool at the Irish Bar and in solicitors’ practices depends on whether it consistently delivers what those previous technologies promised but couldn’t: speak naturally, get a usable document, move on with your day.