Guide · Fundamentals
What is legal automation?
A definition that is useful rather than promotional, the six workflows firms actually automate, where the human has to stay, and what it realistically costs.

Legal automation is software that performs a law firm's repeatable administrative work: opening matters, assembling documents, entering expenses, sending status updates, calculating deadlines, chasing invoices and producing reports. It runs inside the practice management platform the firm already uses, and it leaves every judgment call to a person.
That last part is the whole definition, really. Automation takes the work that has a right answer and does it the same way every time. Anything that requires a lawyer to decide something stays with the lawyer.
What it is not
- It is not AI writing your legal work. Drafting arguments, advising clients and exercising judgment are not what this is for, and a firm that automates those is buying a malpractice problem rather than a productivity gain.
- It is not replacing your staff. It removes the re-typing, the chasing and the remembering, which is the part of the job your best people like least.
- It is not a product you buy off a shelf. Every firm's intake questions, folder structure, document set and billing conventions differ. The platforms provide the surface; the workflows have to be built to match how your office actually runs.
The six things firms actually automate
Across practice areas, the same handful of workflows come up again and again.
- Intake to matter. A qualified consult becomes a matter with contacts, custom fields, a folder structure and a conflict-check task, without anyone re-entering what the client already told you.
- Document assembly. Engagement letters, estate plans and standard filings built from firm templates with the conditional logic already inside them, so the document knows whether the client is married, which pronouns apply and which provisions belong.
- Expense and invoice entry. Vendor invoices from process servers, filing services and experts turned into expenses on the right matter, with the receipt filed and referenced.
- Client communication. Status updates sent at real milestones rather than when someone remembers, which retires most of the "what is happening with my case" calls.
- Deadlines and task chains. Stage-driven task lists assigned to the right person with the right dates, and on platforms that support it, court deadline chains calculated from a trigger event.
- Reporting. Trust balances, work in progress, collections and the intake funnel compiled on a schedule rather than rebuilt by hand every week.
If a task has a right answer and someone does it more than once a week, it is a candidate. If it needs judgment, it is not.
Judgment calls that change with the client and the facts. Automation has no business here.
Advising on strategyJudgment work with repeatable scaffolding. Automate the assembly, a person approves the result.
Drafting from templatesRare and simple. The build costs more than the task ever will.
One-off filing quirkSame answer every time, done constantly. This is where the hours come back.
Expense entry, status updates, reportingWhere the human stays
Every automation worth building has a review step for anything it is not certain about. A document goes to an attorney before it is sent. An expense the system cannot confidently match to a matter waits in a queue rather than being posted to the wrong client. Filing privileged material into the wrong matter is a confidentiality problem, not a bug report, and the design has to reflect that.
The practical rule we use: deterministic rules handle routing, a model is only used as a tie-breaker when the rules are genuinely ambiguous, and anything below a confidence threshold goes to a person. Thresholds are raised only for rules that have measured clean in production.
What it costs and how long it takes
Scoped builds in this market generally run from a few thousand dollars for a single workflow to the low five figures for a connected set, with ongoing monitoring priced monthly. Two to six weeks is a realistic build window for most scopes. The variable that moves timelines most is not engineering: it is platform access. On some platforms API credentials are issued per firm by the vendor, and that request can take longer than the work.
How to tell if your firm is ready
You do not need to be a large firm. You need three things: a platform with an API worth building on, someone in the office who can describe how the work actually happens rather than how the manual says it should, and a willingness to review a queue for the first few weeks while the rules earn their thresholds.
If you want to see what your specific platform supports, we have written up both in detail: the MyCase API and the Clio API.
Keep reading
Wondering what to automate first?
Thirty minutes with the person who runs your office. We will tell you honestly if your platform already does what you need.