The promise vs. the reality

General-purpose AI tools - ChatGPT, Claude, Gemini - are remarkably capable. They can draft emails, summarize documents, and answer legal questions with surprising competence. Many solo family law attorneys have started using them for quick tasks.

But there’s a gap between “generally useful” and “reliable for legal work.” For family law attorneys, that gap is where the risk lives.

Problem 1: No jurisdiction awareness

Ask a general AI tool to calculate the deadline for filing a motion in Santa Clara County, and you’ll get an answer. It might even be close. But it probably won’t:

Distinguish between court days and calendar days per CCP § 1005
Know that Santa Clara County observes different court holidays than Alameda County
Adjust for service method (mail adds calendar days; electronic adds court days)
Account for the specific year’s holiday schedule

The deadline it returns might be off by one or two days. In most contexts, that’s a rounding error. In litigation, it’s a missed filing.

Problem 2: No source accountability

When a general AI generates a case summary or fills information into a document, there’s no way to trace where each fact came from. It might be from a document you uploaded. It might be from the model’s training data. It might be fabricated entirely.

For family law work - where financial declarations, custody evaluations, and disclosure requirements carry legal consequences - every fact needs a source. If you’re signing a declaration that includes AI-generated content, you need to know that the AI pulled from your client’s actual documents, not from a similar case it was trained on.

Problem 3: No matter context

General AI tools process each conversation in isolation. They don’t know that the email you’re asking about relates to the Garcia matter, that the financial documents are for an FL-150, or that opposing counsel’s name is the same across two of your cases (which creates an email routing problem).

Legal work is context-heavy. Every email, document, and deadline exists within the context of a specific matter, with specific parties, specific deadlines, and specific jurisdictional requirements. A tool that can’t maintain that context forces you to re-explain the situation every time.

Problem 4: No approval gate

When you draft an email using a general AI tool, it appears in a chat window. You copy it, paste it into your email client, and send it. There’s no system ensuring that you reviewed the content before it went out. There’s no record that you approved it.

For low-stakes communication, this workflow is fine. For legal correspondence - demand letters, disclosure requests, settlement offers - the absence of a structured review process is a liability.

Problem 5: No billing integration

General AI tools don’t track what work they did, for how long, or on which matter. If you spent 20 minutes going back and forth with ChatGPT refining an FL-150, that time exists only in your memory. There’s no draft billing entry, no matter association, no professional narrative.

For solo attorneys who already lose 15-20% of their billable time to poor tracking, adding untracked AI interactions makes the problem worse.

What “purpose-built” means

The difference between a general AI tool and a legal-specific one isn’t just the training data. It’s the architecture:

Matter-centric organization: Every action happens within the context of a specific case
Jurisdiction awareness: Court rules, holidays, and procedures are built into the system, not guessed at
Source accountability: Every fact the AI presents is linked to the document it came from
Approval gates: Nothing is filed, sent, or changed without explicit attorney review
Billing integration: Work is tracked automatically and tied to the right matter

These aren’t features you can bolt onto a general chatbot. They require a different foundation - one built around how legal work actually flows.

So what should you use?

General AI tools have a place in a family law attorney’s toolkit. They’re useful for brainstorming, drafting initial outlines, and quick research on non-critical questions.

But for the core operational work of running a family law practice - deadline management, form preparation, email triage, document review, billing - the tool needs to understand your jurisdiction, maintain your matter context, and ensure you have the final say on everything that matters.

That’s not a criticism of general AI. It’s a recognition that legal work has specific requirements that generic tools weren’t designed to meet.