Straight answers about agents in your firm.
The questions firms actually ask us in evaluation — autonomy, models, cost, data and migration.
Common questions
Ten answers, no hedging.
Is this just another legal chatbot?
No. Chat is one interface into an operating system. A chatbot returns text; LawPortal agents read your matters, prepare real work and execute permitted actions inside the system — with approvals and an audit trail.
How much autonomy do the agents actually have?
Exactly what you configure. Authority is set per action type across five levels — suggest, prepare, approve, execute, record. Firms usually start with prepare-and-approve everywhere and widen it as trust builds.
Whose AI model does it use?
Yours, where supported. You connect an approved provider and, in most configurations, your own credentials. You also decide which models may handle which categories of work.
Who pays for AI usage?
If you bring your own provider account, usage is billed by that provider directly to your firm and stays visible to you. Where LawPortal supplies capacity, it is described in your plan. We do not hide model cost inside opaque credits.
Is our data used to train models?
Firm data is used to operate your firm's environment. We do not treat client matter data as training material, and provider-side training controls are part of the model policy you configure.
How is our data separated from other firms?
Each firm operates in a logically isolated environment, and agent retrieval is scoped to what the directing user is permitted to see.
Do we have to replace our current system on day one?
No. Firms typically move one workflow or matter group first and run LawPortal alongside their existing environment while results are measured.
What happens when an agent gets something wrong?
Consequential work stops for human review before it takes effect, and every action is recorded with its basis, so mistakes are caught at review and traceable afterwards.
Which practice areas are supported today?
Personal injury, immigration and workers' compensation are configured deepest. Other contingency and volume practices are evaluated case by case.
How long does onboarding take?
Founding-firm configurations typically run two to four weeks for a first workflow, including authority rules and approval gates.
Still have a question we did not answer?
Send it over. We would rather answer it directly than have you guess from a website.