What a small law firm should (and shouldn't) automate with AI

The safest, highest-return place for a small firm to start is the administrative work around the practice of law: intake, filing, capturing deadlines for a person to confirm, and routine client updates. The risky places are anything that produces legal work product without a lawyer's review, and anything that touches the trust account.

The rules that decide where the line sits are ones you already know: competence, confidentiality, and supervision. This guide walks through what to automate, what to leave alone, and how to set it up so it holds up to those duties.

This is practical guidance, not legal or ethics advice. Check your state bar's guidance before adopting any AI tool.

Start with the professional-responsibility questions

In July 2024 the American Bar Association issued Formal Opinion 512 on lawyers' use of generative AI. It applies existing Model Rules rather than creating new ones, and many state bars have published their own guidance since. The themes that matter for automation:

• Competence. You need a reasonable understanding of what a tool can and can't do, and you can't rely on its output without checking it.
• Confidentiality. Before client information goes into any AI tool, assess the risk that it could be disclosed or used to train the tool. The opinion takes the view that tools which learn from your inputs can require the client's informed consent first.
• Supervision. Partners and managers are responsible for firm policies on AI use and for supervising staff and outside vendors who use it.
• Fees. If you bill hourly, you bill for time actually spent. Time an automation saves you isn't billable time.

The cautionary tale is familiar: in 2023, a federal court in New York sanctioned lawyers who filed a brief citing cases that ChatGPT had invented. The problem was filing the tool's output without verifying it, rather than using the tool in the first place.

So automate the work where a mistake is easy to catch and review, keep a person approving anything that leaves the firm, and never let a tool be the final word on law, deadlines, or money.

What to automate

Client intake

A new inquiry arrives through your website form or email. Automation can collect the details, pull the names of the parties together so they're ready for a conflicts search, draft the engagement letter from your template, send it for e-signature, and create the matter folder once it's signed.

Who decides: a lawyer runs and clears the conflict check and reviews the engagement letter before it goes out. The automation does the preparation, and the decisions stay with the lawyer.

Document filing

Pleadings, correspondence, discovery, and signed agreements arrive by email, scan, and upload. Automation reads each one, renames it to your convention, files it to the right matter folder, and tags the document type, date, and parties. Anything it can't match with confidence goes to a review queue instead of a guess.

This is usually the biggest time saver in a small firm, and the lowest risk: a misfiled document is easy to spot and move.

Deadline capture, confirmed by a person

Automation can read court notices, orders, and scheduling letters and propose calendar entries with the source document linked.

Who decides: your docketing person or the responsible lawyer confirms every entry. Deadline calculations depend on court rules, service method, and holidays. Treat automated dates as a second set of eyes that catches notices someone might have missed, and never use them as the docket of record.

Routine client updates

Status emails, "we received your documents," and appointment reminders can be drafted from matter notes and your templates.

Who decides: a paralegal or lawyer approves each draft before it's sent, at least until you're confident in a given template.

Plain-English search across firm files

"The signed settlement agreement in the Alvarez matter" should return the document, not two hundred candidates. Search built on your own document store saves time every day.

Set it up so search respects the same access rules as your files, including ethical walls. Nobody should be able to find a document through search that they couldn't open directly.

Time-entry drafts

Automation can draft time entries from calendar events, emails sent, and documents edited, so lawyers aren't reconstructing a day from memory on Friday.

Who decides: the lawyer edits and approves every entry. The draft is a memory aid, and the bill reflects time actually worked.

What not to automate

Set up confidentiality first

Before any automation touches client files, confirm the following:

• AI tools are on business or enterprise plans whose terms say your data isn't used to train models
• Client files stay in your own document system or cloud storage, not copied into a vendor's
• Each automation can reach only the folders it needs
• Every automated action is logged, and someone reviews the log periodically
• Staff have a written policy on which tools they can use and what can go into them
• You've decided when to tell clients about AI use, and whether any engagement needs consent

If you don't have a written policy yet, our AI use policy template takes an afternoon. For the data questions to put to any vendor, see the owner's guide to AI data safety.

Where to start

Pick intake or filing, whichever takes more of your staff's week. If you're not sure, run a one-week time audit first. Build one workflow, keep the approval steps on, and measure the hours for a month before adding the next.

We set these up for small firms with the approval steps and access controls described above. See what we build for law firms, or book a free call and tell us where your paralegals' time goes.