IOLTA and AI Don't Mix Well — Unless NJ Solo Attorneys Follow These Rules First

Trust accounting is the one place in your practice where a mistake doesn't just embarrass you — it ends your career. Yet more and more solo attorneys in New Jersey are quietly letting AI-assisted billing platforms, practice management tools, and automated reconciliation features touch their IOLTA accounts without pausing to ask whether that's actually safe.

It should give you pause. Here's why — and what to do about it.

---

The Automation Creep Nobody Talks About

When most lawyers think about "AI in the practice," they picture drafting tools or research assistants. What they don't picture is the billing module in their practice management software quietly using machine learning to categorize transactions, suggest trust ledger entries, or auto-populate client cost advances.

But that's exactly what's happening in platforms like Clio, MyCase, and several competitors. AI-assisted features are being layered into the same interfaces where NJ attorneys manage their IOLTA funds. These aren't hypothetical future features — they're live, default-on in some cases, and most users haven't read the release notes.

The problem isn't that automation is inherently wrong. It's that NJ's IOLTA rules — governed by the New Jersey Court Rules, specifically Rule 1:21-6 — impose strict, personal obligations on the attorney of record. No software agreement, however sophisticated, transfers that responsibility anywhere else.

---

What Rule 1:21-6 Actually Demands

NJ Court Rule 1:21-6 requires attorneys to maintain detailed records for every client trust account, including:

• A cash receipts and disbursements journal for each account
• A separate ledger for each client matter
• Reconciliation of ledger balances with bank statements at least monthly
• Retention of all records for seven years

The rule is explicit: the attorney is personally responsible for these records. That means if an AI categorization feature in your practice management tool misclassifies a client cost advance as an operating expense — or worse, misattributes a deposit to the wrong client ledger — you own that error under NJ ethics rules, not the vendor.

This is not a theoretical risk. It is the kind of mistake that triggers an OAE (Office of Attorney Ethics) inquiry.

---

Three Specific AI Features to Audit Right Now

If you use any AI-assisted features in your billing or practice management software, run through this audit:

1. Auto-categorization of transactions. Many platforms now suggest or auto-assign expense categories using ML models trained on aggregate data. Before trusting these suggestions for trust account entries, verify whether the model distinguishes between operating funds and client funds. Most general-purpose categorization engines do not — they were trained on business accounting, not legal trust accounting.

2. Automated reconciliation alerts. Some tools use AI to flag reconciliation discrepancies. This is genuinely useful — if you understand what the model flags and what it silently passes. Read your platform's documentation on reconciliation logic. "AI-assisted" does not mean "complete." Know the edge cases it misses.

3. Invoice generation tied to trust draws. If your platform auto-generates invoices and pulls payment from trust automatically upon approval, audit the approval workflow. Is there a human checkpoint before funds are disbursed? Under Rule 1:21-6, disbursements from trust require attorney authorization — a rubber-stamp click-through in an automated workflow may not satisfy that standard in an OAE review.

---

The Vendor Contract Problem

Here is what most attorneys miss when they sign up for these platforms: the standard terms of service for practice management software disclaim liability for financial errors almost universally. You are licensing a tool, not hiring a fiduciary.

When AI features touch your IOLTA account — even indirectly — your vendor agreement should, at minimum:

• Specify which features interact with trust account data
• Clarify whether transaction data is used to train shared AI models (a confidentiality issue under NJ RPC 1.6 that compounds the trust accounting concern)
• Confirm data residency and retention policies consistent with your seven-year record-keeping obligation

Most off-the-shelf agreements don't cover any of this. That's your job to negotiate or at least document as a known limitation.

---

A Practical Stance for NJ Solo Attorneys

None of this means you should avoid automation in trust accounting. Manual reconciliation is error-prone in its own right, and good software genuinely reduces mistakes. The point is to be the last set of eyes, not a passive approver.

A reasonable approach looks like this: use your platform's AI features for flagging and suggestion, but treat every trust account entry as requiring a deliberate human decision before it's recorded. Monthly reconciliations should be done by you — not simply accepted from a dashboard summary. And before you add any new AI feature to a workflow that touches client funds, ask your platform's support team one direct question: Does this feature interact with trust account data?

If they can't answer clearly, that's your answer.

The OAE doesn't grade on a curve because your software made a compelling UI. Competence under NJ RPC 1.1 has always included understanding your tools — and in 2025, that means understanding exactly where the algorithm ends and your professional obligation begins.