AI for law firms takes on the groundwork around legal work: sorting new enquiries, a first pass through documents, summarising matters, finding the right precedent and capturing time. We design and build these tools on the systems your practice already runs. They do not give legal advice. A lawyer reviews anything that goes to a client or a court, and stays responsible for it.
Overview
Legal work is careful work, and a large part of a lawyer's day goes on the steps around it: reading through documents, finding the precedent the firm used last time, summarising where a matter stands and remembering to record the time.
Tensor Partners builds agents for that groundwork. They read, sort, summarise and search, working from your firm's own documents and the rules you set, and they hand everything to a lawyer to review. Nothing goes to a client or a court without a lawyer's sign-off. Confidentiality is designed in from the start: agents see only the matters a person is already permitted to see, every action is logged, and where client data lives is agreed before anything is built.
Our live systems were not built for a law firm, but two of them do the same kind of work. The Triage Agent reads a morning inbox, separates signal from noise and turns actionable messages into a prioritised task list. The Reconciliation Agent matches records automatically and flags exceptions for human review instead of guessing. Those are the patterns behind intake triage and filing for a practice.
At a glance
Agents work from the documents and systems your firm already runs, limited to the matters a person may see. Everything they produce goes to a lawyer, who reviews it and stays responsible for anything that goes to a client or court.
What you already run
Client enquiries
Web forms and the firm inbox
Matter documents
Contracts, correspondence and disclosure
Precedents + templates
The firm's own past work
Practice systems
Matter records, email and calendar
The agent
AI agent
Works to the firm's rules
Reads, sorts, summarises and searches
Lawyer review
A lawyer signs off
And stays responsible for what goes to a client or court
What comes out
Intake notes, routed
Passages flagged for review
Matter summaries
Draft time entries
Use cases
New enquiries from your website and inbox sorted by matter type and urgency, the details your team needs for its conflict check captured, and each one routed to the right person. The same pattern as our Triage Agent.
See the Triage Agent →An agent reads contracts, correspondence and disclosure against a checklist the firm defines and flags the clauses and passages worth a lawyer's attention, each flag linked to the source text. It speeds up the first pass; the lawyer does the review and forms the view.
Document processing →A plain summary of where a matter stands, drawn from the file: the parties, key dates, what has been agreed and what is outstanding. Useful for handovers, partner reviews and picking a file back up, and always checked by the lawyer before it is relied on.
Ask a question and get an answer from the firm's own precedents, templates and past work, with every answer citing the document it came from. Access follows your existing permissions, so people only see what they are already allowed to see.
Internal knowledge assistants →Draft time entries built from the emails, documents and calendar entries a lawyer worked on that day, ready to review, adjust and submit. Less time reconstructing the week from memory, and nothing is recorded without the lawyer's approval.
Incoming documents and correspondence matched to the right client and matter and filed, with anything the agent cannot match confidently sent to a person instead of guessed. The same matching-with-exceptions pattern as our Reconciliation Agent.
See the Reconciliation Agent →Case studies
Inbox operations
Not built for a law firm, but the same job as intake triage. An agent that reads the morning inbox, separates signal from noise, turns actionable messages into a prioritised task list and delivers a daily brief of what needs doing and by whom.
Finance operations agent
Built for a finance team, not a practice, but it shows the guardrail that matters in legal work. The agent reads live bank transactions, matches each credit to an invoice, records the payment and flags exceptions for human review instead of forcing a match.
How it works
Every phase ends in something you can see working. You decide at each step whether to go further.
A fixed-price AI opportunity audit. We map how work moves through your firm, from first enquiry to final bill, agree what any model may see, and rank every opportunity on return, feasibility and risk, with the builds priced.
AI consulting + audit →Working software on your stack, tested against your real documents, with lawyer review built in wherever a judgement call sits.
AI agents →Shipped into production with logging, access controls and rollback. We stay on to measure, tune and extend the system as your firm and the models change.
AI automation →Services
Each one can run on its own or as part of a wider engagement.
FAQ
By taking on the groundwork around legal work: sorting new enquiries, a first pass through documents against a checklist, matter summaries, finding precedents and drafting time entries. The aim is more of a lawyer's time on judgement and clients, with a lawyer reviewing everything the agent produces.
No. The tools we build support lawyers; they do not advise clients. Anything that goes to a client or a court is reviewed and approved by a lawyer, who stays responsible for it, as they would for any work prepared for their sign-off.
It is designed in from the start. Agents get the minimum access each task needs, follow your existing matter permissions and log every action they take. What any model may see, and where client data is stored and processed, is agreed during the audit, before anything is built, and nothing ships without your sign-off on data handling.
Usually, yes. We build on top of the systems you already run, connecting through their APIs instead of replacing them. If a system genuinely cannot be connected, the audit will tell you before you commit to a build.
That is why it is built to stop and ask. Agents point to the source behind every summary or flag so a lawyer can check it, send anything uncertain to a person instead of guessing, and keep a record of what they did. Every engagement starts with a fixed-price audit, so you see the risks, the guardrails and the numbers before committing.
Start here
Start here
30 minutes, no pitch deck, no obligation. Tell us what's eating your team's time and we'll tell you straight whether AI is the answer, and roughly what it would take.