Pilot
How to run a legal AI pilot
Three weeks, on contracts that actually reach you, inside your own account. This is the method we run, written out in full, including the part most pilots leave undefined: what counts as finished.
Talk about a pilotThe exchange, stated plainly
Most pilot offers are vague about what the vendor wants back, which is how a three-week trial becomes a three-month one. Here is both halves.
You get
- Three weeks of the full product, free.
- Your playbook and your house forms taught from documents you have already written.
- Every piece of setup done for you, on a call, on the first day.
- A written report at the end of what it found.
You give
- One real matter. Nothing confidential you are not comfortable sharing.
- About two hours of lawyer time, across three weeks.
- One 30-minute call at the end, recorded with your permission.
- Permission to quote you, anonymously unless you later agree otherwise.
Neither of us owes the other anything at the end. No auto-renewal, no conversion clause, no obligation to buy. If it is not right, you stop on any day and nothing is owed.
The three weeks
Pilots start on a Monday, because the measurement needs two distinct calendar weeks and three weeks leaves exactly one week of slack.
| When | What happens | Who |
|---|---|---|
| Day 0, Monday | A 45-minute setup call. Your account is created, your prior documents are loaded to teach the playbook and the forms, and the first matter is set up live on the call. | We drive, you watch |
| Days 1 to 2 | The first matter goes through end to end. The target is work product in your hands inside 48 hours. A pilot that has produced nothing by the end of week one rarely recovers. | We assist |
| Day 5, Friday | A ten-minute check. What was wrong with the output, and what can be fixed inside the pilot rather than promised for later. | Us |
| Week 2 | You run one yourself, unaided. We do not touch it. This is the week that decides the pilot, and it is the honest read on whether the thing works without its author in the room. | You |
| Week 3 | The wider surface: one question run across the whole document set, and a draft produced from your own form. | You, we are on call |
| Day 21, Friday | Exit interview, 30 minutes, recorded. The questions are below, and we do not change them to suit the result. | Both |
What "complete" means, and why it is not our opinion
This is the question to press every vendor on, ours included. If a pilot ends with the vendor telling you it went well, you have bought an impression.
A pilot here is complete when two things are true in the database, and both are machine-checked rather than judged: work product was actually delivered, and the product was used in two distinct calendar weeks.
The second one is the one that matters. A single enthusiastic afternoon is not adoption. Two separate weeks means somebody came back to it when the novelty had worn off, which is the smallest honest test of whether a tool has entered the work.
Five questions to ask before you start
These are the ones we ask ourselves before offering a pilot, because a pilot that cannot succeed wastes your three weeks more than it wastes ours. They work on any vendor, not just this one.
Ask us something elseIs the paper actually negotiable?
If the contracts you receive are take-it-or-leave-it, a redline has little value to you no matter how good it is. This is the question most likely to disqualify a pilot, and the most expensive one to discover in week three.
Is there a document set, or a single file?
One contract tests a commodity. A matter with a base agreement, its amendments and its order forms tests whether the tool can hold the whole thing at once, which is where the difference between tools shows up.
Is this work a named line of business?
If contract work is a bolt-on rather than something your week is built around, the volume will not be there to tell you anything by day 21.
Can you name a matter you could start on Monday?
"Something will come in" is not a pilot. A named matter, already on your desk, is the difference between a three-week evaluation and a three-week intention.
What do you use today?
If you already run something you are happy with, say so at the start. The honest comparison is against your current tool, not against doing it by hand, and a vendor who does not want to hear the answer is telling you something.
The exit interview
Thirty minutes on day 21, recorded with permission. Writing the questions down before the pilot starts is the point: a vendor who chooses the questions afterwards will choose the flattering ones.
- What did it get right that you did not expect?
- What did it get wrong, and what did that cost you?
- Where did you stop trusting it?
- What did you do with the work product: send it, rework it, or throw it away?
- If it vanished tomorrow, what would you miss?
Question three is the one we care about most. A tool that is wrong occasionally is workable. A tool you have stopped trusting is finished, whatever the other four answers say.
Don't take our word for it
Run the method above rather than taking anyone's word, ours included. Three weeks on your own paper answers the question. A demo does not.