Law firms: what to automate and what stays with the lawyer
Intake, file preparation, first drafts, deadlines and time recording: what a law firm can hand to AI, and what must stay with the lawyer.
Thibault Pons de Vincent
Founder of Oliverlist and VivaSync
A law firm sells judgment: advice, strategy, advocacy, negotiation. Yet a large part of the week goes elsewhere. Answering new enquiries, chasing documents, filing, formatting, watching dates, rebuilding timesheets so you can bill. Much of that work can be handed to AI. The advice itself cannot.
Many firms carry extra layers of this work. Clients and counterparties are spread across several countries. Correspondence runs in more than one language. New clients go through KYC checks before a matter opens, and some filings must be translated. None of this is legal judgment, and all of it takes time.
At Oliverlist, the company I founded in 2016, AI has been part of our product since May 2023. By 2025 it was sorting prospect replies in production, writing call scripts and coaching the team. One rule came out of that experience, and I apply it everywhere: AI prepares, and a person approves anything that commits the company. In a law firm, that rule is not up for debate. Whatever leaves the firm carries its liability, and what clients share is protected by confidentiality and privilege. What follows describes possibilities to adapt to your firm, not client case studies.
Intake: sort every enquiry and prepare the first conversation
New matters come in through the website, by email, by phone, through referrals and, more and more, by WhatsApp. Some fall outside your practice areas. Others are urgent and still sit in a shared inbox. AI can read each enquiry as it lands, classify it by practice area and urgency, and route it to the right partner or associate.
It can also prepare what the lawyer needs before the first call: a summary of the request, the parties involved, the documents and KYC information to collect, and a draft reply that acknowledges receipt and offers a time slot. It can search your matter database for the names of the parties and flag a potential conflict. The conflict decision stays with a lawyer.
Set the sorting to prefer a false alarm over a missed emergency. An extra alert costs a few seconds of reading. An urgent enquiry left unanswered can cost you a client, or a deadline.
Files, letters and deadlines: the preparation work
Once a matter is open, part of the work is preparation. Sorting and renaming incoming documents, building a chronology, putting together the bundle and its index. AI does this quickly and in your house formats.
It can draft the repetitive correspondence: acknowledgements, document requests, status updates to the client, cover letters, follow-ups to opposing counsel. It can also produce working translations between languages so the team can read a document on day one, while anything filed still goes through a certified legal translator. A first draft is not a letter: the lawyer reads, corrects and signs. For opinions and pleadings, AI can propose a structure or summarize a large file. But every legal authority it cites must be checked against the source, because AI can quote a judgment that does not exist, with complete confidence.
Deadlines deserve separate treatment. AI can read a judgment, a hearing notice or a letter from the court registry, extract the dates and propose calendar entries with reminders. But a missed procedural deadline cannot always be repaired. I would keep a human check on every date entered, however reliable the tool has proven to be.
Before any client file goes into a tool, get written confirmation of where the data is stored and that it will not be used to train the provider's models. Accounts should be in the firm's name, with access you control.
Time recording: rebuild it instead of typing it
Timesheets are often filled in at the end of the week, from memory, with gaps. AI can propose time entries from everything that leaves a trace: emails sent, documents edited, meetings and calls in the calendar. It links each entry to the right matter and writes a narrative the client can understand. The lawyer approves or corrects instead of reconstructing the week.
The same logic extends to preparing invoices, tracking retainers and chasing unpaid fees, always approved before anything goes out. At Oliverlist, we have automated the entire chain since 2020, from importing contacts to issuing the invoice. What struck me is how similar this work is from one industry to another: a file, a trail, time spent, an invoice.
What stays with the lawyer
Advice and strategy. Conflict decisions. The engagement letter and the fee arrangement. Anything signed or sent in the firm's name. The delicate moments with a client, such as delivering bad news or recommending a settlement. AI can prepare each of these moments. It does not replace them.
One condition frames everything else: each automation has a named owner, who knows what it does, checks its output and can switch it off. That is how you move quickly without losing control of what leaves the firm.
Key takeaways
- The work around the advice can go to AI: intake, file preparation, first drafts, time recording.
- Judgment, signature and the client relationship stay with the lawyer.
- Every deadline entered by AI is checked by a person, and every legal authority is checked against the source.
- Start with a task that takes time without adding anything for the client, such as time recording or acknowledgements.