What it actually is

SpeechLive Legal AI Assistant is a drafting layer that sits on top of Philips SpeechLive. You speak — live, or by uploading a recording — and instead of a wall of transcribed text, you get a structured draft in your firm’s template.

The distinction matters more than it sounds. Dictation software transcribes. It hands you words, and somebody downstream still has to turn those words into a document: apply the template, format the fees, build the signature block, fix the numbering. If your firm has been absorbing that gap with a secretary’s or paralegal’s time, you have been buying half a solution for years.

This closes that half. And it does it without removing the person who has been closing it for you — which is the part that makes it different from nearly everything else being sold to law firms right now.

How it works

You do not have to dictate in order

This is the part that surprises people who have used voice software before. There are no field prompts, no “next section,” no rigid script. You talk the way you would brief a capable assistant sitting across the desk — including changing your mind halfway through.

That is a real instruction from a Philips demonstration of an engagement letter, spoken mid-flow. The draft came back with the corrected rate and the fees in bold. You are not learning a command language; you are talking, and the corrections you make out loud are simply understood as corrections.

The part that decides whether a small firm says yes

What law firms are creating with it

Summaries deserve a specific mention. They are one of the highest-volume things attorneys ask for and one of the most time-expensive to produce by hand — case summaries, witness note summaries, a client-readable version of something written for a court.

Templates you build by talking

You do not need an IT project to get your firm’s documents in. You ask the AI chat to build a template — create an engagement letter template, create a witness note document — and then edit it.

Templates can carry their own standing instructions. If a section must be reproduced word for word, you can tell the template to treat it as verbatim and leave the generative drafting switched off for that block. That is how you keep your firm’s mandatory language mandatory.

More than one way to get the facts in

Dictation is the primary input, but not the only one. You can attach up to five source files to a single document so the AI can pull in detail you did not say out loud — supported formats include .txt, .docx, .pdf, .jpeg, .png, .bmp and .heif, and that includes handwriting recognition from a photographed page.

Which means a photograph of your handwritten notes from a client meeting can become source material for the draft, rather than something you have to read aloud.

After the draft: keep talking to it

The document is not frozen when it is generated. There is an AI chat over the finished draft, so the revisions you would normally mark up by hand you can simply ask for:

• “Renumber these in sequence” — when everything came back as item one
• “Change the numerals to Roman numerals”
• “Add a signature block at the bottom”
• “Add a short summary at the end with no legalese”

Then export to Word, PDF or plain text — or leave it in the workflow, where your typist sees it in SpeechLive, proofreads, completes and returns it.

Where it fits, and where it does not

We would rather tell you this now than after you have bought it.

What about speech recognition itself?

Legal speech recognition — the vocabulary that handles legal terminology and citations correctly — runs inside the AI Assistant. That is where the legal-trained recognition applies today.

SpeechLive also includes speech recognition at the cursor through a small desktop application, so you can dictate straight into an email or another application. That one currently uses the standard vocabulary rather than the legal one.

Philips has additional capability in development, including deeper Microsoft Word integration. We will describe those here when they ship and not before — a roadmap is not a product, and you should not buy one. If a specific capability is load-bearing for your firm, ask us and we will get you a straight answer on where it actually stands.

Security, compliance and your clients’ data

Built on SpeechLive, which matters commercially

The AI Assistant is a layer on the SpeechLive platform, not a standalone purchase. That means an underlying SpeechLive plan is part of the picture, and existing SpeechLive customers are in a very different position from someone starting fresh — there are several upgrade paths, and what you already own changes the number substantially.

Pricing depends enough on what you have today that quoting a single figure here would mislead you. Tell us what you are running and we will price it properly, including whether any promotional terms currently apply.

Why buy it from Voice Automated

Because we have done this exact migration before, on the clinical side, for two years — walking professionals off end-of-life Nuance products and onto what replaced them. We know which parts break, which parts people miss, and which reassurances turn out to be worth nothing.

We are also not going to sell you this if it is wrong for you. Our comparison of the Dragon Legal alternatives in 2026 names options we make nothing from, including one that may already be sitting in your Microsoft subscription. Read that first if you are still deciding.

Read next