They say you are what you eat.
Well, your legal AI is only as smart as the data you feed it. And right now, it's starving.
Every few weeks, it feels like another legal AI tool launches into the legal software market, from assistants built into your practice management software, like Clio Work and Smokeball’s Archie, to heavy hitters like Harvey and Legora. Unlike a generic ChatGPT search, legal AI is built around how legal work actually happens.
These tools are positioned as junior lawyers who can read more documents faster than any person could. You can type or talk to them the way you’d ask a paralegal at the desk next to you to “draft this follow-up,” “build a chronology,” or “summarize what we agreed to on this matter.” No prompt engineering degree required.
There’s just one problem, most of these “juniors” are walking into the office having only read half the file.
In this article, we’ll unpack what legal AI can and can’t read, why phone calls and meetings are the missing context, and how to connect your AI notetaker and phone system to your legal software so your AI can finally work from what was actually said.
What your legal AI can read
Before we get to what’s missing, let’s give your legal AI its due. It has a lot to work with. Today, the typical legal AI can read:
- Emails, if your Outlook or Gmail is connected.
- Documents stored against the matter in your practice management system, or in a document management system like NetDocuments or iManage.
- File notes saved to the matter, whether you, your team, or your AI notetaker wrote them up.
- Calendar events like meetings, deadlines, and court dates.
- Time entries logged against the matter.
- Matter details like parties, status, key dates, and billing info.
This is the written layer of legal work: if it exists as text in a connected system, your AI can read it.
What your legal AI can’t read
If it isn’t written down, your AI can’t read it. That rules out phone calls, voicemails, and meetings, whether they happen in person or online. These are some of the most common and important channels in legal practice.
Research by Clio found that 68% of clients initially reach out by phone, rather than by email or a web form. Those first calls are your firm’s front door. If they never make it into your legal software, your AI misses the conversation where most of your matters begin.
Calls and meetings are also where much of the work on a matter actually gets done. It’s where clients give instructions, lawyers give advice, and scope and settlement positions get negotiated. Yet most of those conversations never reach the matter file.
If a conversation isn’t in the system, your AI doesn’t know it happened. Ask it, “What did we agree about the deposit?” and it’ll answer from the emails and documents it can read, even if you made the real agreement on a call last week. Ask it to draft a follow-up, and it’ll leave out whatever was discussed out loud. Your AI isn’t failing at its job; it’s working with half the facts, so at best its answers are incomplete or out of date, and at worst they’re wrong.
Why phone calls and meetings are hard to capture
Emails and documents are already in writing, so they’re easy to save to your legal software. Calls and meetings aren’t. They happen on the fly, and they leave no record beyond whatever you scribble down afterward (if there’s time).
To get a conversation from a call or meeting into your legal software, someone has to record it, accurately transcribe it, find the right matter, and save it in a form a colleague could actually use if they picked up the file later. That’s four manual steps for every call and meeting. Multiply that by everything your firm handles in a week, and it’s no wonder so few conversations end up on file.
The average business phone system isn’t much help, either. Some log that a call happened, and a few can record or transcribe it. But those records stay inside the phone system, disconnected from the matter, the client, your legal software, and the AI that’s supposed to make sense of it all.
The conversation happened, but as far as your legal software is concerned, it might as well not have. You end up with two separate sources of truth and an AI that can only read one of them. That’s the gap a legal-specific phone integration is built to close.
The best way to get phone calls and meetings to your legal AI
The fix isn’t another task for someone to remember. It’s capture that happens automatically, in a system already connected to your legal software. And because these are client conversations, it can’t be just any tool. It needs to be closed, secure, and built for legal work, with your data staying in one place instead of training someone else’s model.
Get that right, and your legal AI finally has the full picture. The instructions, advice, and agreements from your calls and meetings end up written down and saved to the matter, right alongside your emails and documents.
That’s what VXT does. VXT is legal AI too, but with a different job: turning conversations into completed legal paperwork. During calls and meetings, it takes notes so you can focus on your client. Afterward, it summarizes and saves those notes, along with a time entry, to the right matter in your legal software, and drafts a follow-up email based on the conversation. Those notes then become context your legal AI can draw on. VXT captures what was said, and your legal AI puts it to work.
Don’t let your legal AI starve
When calls stay in phone logs, voicemails get forgotten, and texts live on someone’s phone, your legal AI only has half the picture. Its answers are less reliable, and less useful, as a result.
VXT closes the loop. VXT Phone is the phone system your firm uses to make and receive calls, and VXT Meet is the AI notetaker for your in-person and video meetings. Both save the conversation straight to the matter, so your legal AI stops guessing. It works from what was said, not just what was written.
Legal AI tools will keep getting better, but they’ll only ever be as good as the information they can read. The firms that get the most out of them are the ones feeding their AI everything that matters, not just what’s easy to capture.
Your legal AI is hungry. Feed it the whole conversation.
