For in-house legal teams
A legal department answers to the whole company, and the volume arrives without a front door. This is the page I would send to a general counsel with ten minutes: which workflow to start with, and what a pilot measures.
The queue is the problem before the drafting is.
A Tuesday
The mailbox has nineteen new requests and four of them are urgent according to the person who sent them. Two lawyers have separately started on the same vendor question. A business team is waiting on an NDA that has been sitting for six days behind something that turned out not to matter.
A notice arrived on Friday with a deadline that nobody has stamped. Somebody from finance wants to know what the team spent its time on last quarter, and the only honest answer is a shrug. None of that is a failure of lawyering. It is a queue with no front door.
The three questions I am asked
The first is whether it will actually save time. The honest answer is that a pilot is how you find out, with your own baseline. I have no way of knowing what your queue looks like before I read two weeks of it.
The second is where the data goes. That answer is on the security and data page: your environment or a region you choose, sources fixed in advance, and the training question settled in writing before the pilot starts.
The third is who is accountable. The answer is a named reviewer, always, with a mandate that says what the agent could read and do, and a record that says who signed.
Start with intake and triage
Not because it is the most impressive, but because it is the one that produces the baseline the other three are later measured against. Once every request is a named file with a type, an owner and a clock, you can see the shape of the work for the first time. The argument about what to automate next then stops being an argument.
- 01Legal intake and triageIn-house
Requests from anywhere in the company land in a single queue and are opened as a named file.
- 02Contract first-pass reviewIn-house
Standard contracts are read against the review playbook and every deviation is marked by clause.
- 03Regulatory and compliance operationsIn-house
Notices, applications and regulator correspondence run in a single list set to the tightest deadline.
- 04AI governance and release gateIn-house
Every AI workflow has a register entry and every change passes through a gate.
The strip after each name shows which parts of the standard that workflow leans on hardest, in the order intake, mandate, review, record.
Contract first-pass review comes second because it needs your review playbook in writing first. If the playbook is not written, the first job is to write it, and that output alone can be worth more than the workflow. If a specific queue is the one actually hurting, whether that is contract first-pass review, the compliance clock or an AI register nobody has kept, we start there instead. One workflow, not the right workflow chosen in advance by me. A second workflow is not opened until the first has passed its evaluation gate; measuring two at once is measuring neither.
What an engagement looks like
A call of about an hour
The workflow, the team, the systems, the reviewer. No charge, no obligation.
A written scope
Stages, mandate, record format, the one measure, the dates and the price.
A supervised pilot
A baseline first, nothing unattended, and a pass or fail on a date fixed at the start.
What is measured during the pilot
The measure is yours. These are proposals, none of them a promised result: things to measure, not numbers to reach. The periods are written into each scope.
- Baseline: the current state is measured. The chosen measure is counted as it stands before the workflow exists, and the pilot does not start until that number is written down.
- Month one: the pilot runs supervised. Every output passes a lawyer’s approval, the measure is read weekly and the number of events in the critical error class is counted.
- Month two: the evaluation set is rerun. The gate’s result is recorded as a pass or a fail; if it failed, the reason is written down.
- End date: two numbers sit side by side, the baseline and the end. The decision is yours: the workflow continues, changes or stops. All three are legitimate outcomes.
At the end you hold the workflow definition, the routing rules where they apply, the playbook, the mandate register, the evaluation set and every run record, in an exportable format.
I am not going to tell you what the numbers will be. Anyone who does, before reading your queue, is describing a brochure.
What this is not
It is not a licence, not a platform, and not a replacement for your contract system, your document system or the assistant you already pay for. Workflow, copilot or CLM draws that line. LEXSAS is not a law firm and gives no legal advice.
Three short answers
The workflow runs in your own environment, the sources the agent may read are fixed in advance, and your organisation remains the data controller. The design brief names the provider terms and account settings that keep your content out of model training, and the pilot does not start until they are in place. The detail is set out, item by item, on the security and data handling page.
The lawyer who signs. The agent proposes, the decision is a person’s, and the record shows whose decision it was. That is not a rule the workflow invents; it is where professional responsibility already sits. What the workflow adds is that the responsibility can be shown.
Your tool answers; the workflow carries the work. The tool you have is used inside the workflow as a tool. LEXSAS sells nothing in its place.
Start with one workflow from the workflows page, or read the operating standard first. Get in touch